CN112035635A - Medical field intention recognition method, device, equipment and storage medium - Google Patents

Medical field intention recognition method, device, equipment and storage medium Download PDF

Info

Publication number
CN112035635A
CN112035635A CN202010884353.8A CN202010884353A CN112035635A CN 112035635 A CN112035635 A CN 112035635A CN 202010884353 A CN202010884353 A CN 202010884353A CN 112035635 A CN112035635 A CN 112035635A
Authority
CN
China
Prior art keywords
entity
medical
preset
coarse
grained
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Pending
Application number
CN202010884353.8A
Other languages
Chinese (zh)
Inventor
原丽娜
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Kangjian Information Technology Shenzhen Co Ltd
Original Assignee
Kangjian Information Technology Shenzhen Co Ltd
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Kangjian Information Technology Shenzhen Co Ltd filed Critical Kangjian Information Technology Shenzhen Co Ltd
Priority to CN202010884353.8A priority Critical patent/CN112035635A/en
Publication of CN112035635A publication Critical patent/CN112035635A/en
Priority to PCT/CN2021/084659 priority patent/WO2022041730A1/en
Pending legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
    • G06F16/33—Querying
    • G06F16/332—Query formulation
    • G06F16/3329—Natural language query formulation or dialogue systems
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
    • G06F16/36—Creation of semantic tools, e.g. ontology or thesauri
    • G06F16/367—Ontology
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/24—Classification techniques
    • G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
    • G06F18/2415—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on parametric or probabilistic models, e.g. based on likelihood ratio or false acceptance rate versus a false rejection rate
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F40/00—Handling natural language data
    • G06F40/20—Natural language analysis
    • G06F40/237—Lexical tools
    • G06F40/247—Thesauruses; Synonyms
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F40/00—Handling natural language data
    • G06F40/20—Natural language analysis
    • G06F40/279—Recognition of textual entities
    • G06F40/289—Phrasal analysis, e.g. finite state techniques or chunking
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F40/00—Handling natural language data
    • G06F40/20—Natural language analysis
    • G06F40/279—Recognition of textual entities
    • G06F40/289—Phrasal analysis, e.g. finite state techniques or chunking
    • G06F40/295—Named entity recognition
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F40/00—Handling natural language data
    • G06F40/30—Semantic analysis
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/045—Combinations of networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/049—Temporal neural networks, e.g. delay elements, oscillating neurons or pulsed inputs
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Engineering & Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Artificial Intelligence (AREA)
  • Computational Linguistics (AREA)
  • Data Mining & Analysis (AREA)
  • General Health & Medical Sciences (AREA)
  • Health & Medical Sciences (AREA)
  • Mathematical Physics (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Evolutionary Computation (AREA)
  • Audiology, Speech & Language Pathology (AREA)
  • Biomedical Technology (AREA)
  • Computing Systems (AREA)
  • Biophysics (AREA)
  • Software Systems (AREA)
  • Molecular Biology (AREA)
  • Databases & Information Systems (AREA)
  • Probability & Statistics with Applications (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Evolutionary Biology (AREA)
  • Animal Behavior & Ethology (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Human Computer Interaction (AREA)
  • Medical Treatment And Welfare Office Work (AREA)
  • Information Retrieval, Db Structures And Fs Structures Therefor (AREA)

Abstract

The invention relates to the field of artificial intelligence, and discloses a medical field intention identification method, a device, equipment and a storage medium, which are applied to the field of intelligent medical treatment and improve the accuracy of an intention identification result in the medical field. The method comprises the following steps: acquiring an initial question sentence from a terminal; calling a preset identification model to perform entity identification on the initial question sentence to obtain an entity identification result; entity linkage is carried out on the plurality of coarse-grained entity labels according to a preset medical entity synonym table, and linked entity labels are obtained; performing intention recognition on the initial question sentence according to a preset intention recognition model, an entity recognition result and the linked entity label to obtain a candidate medical intention; generating a knowledge graph query statement according to the candidate medical intention; and inquiring the preset medical knowledge map based on the knowledge map inquiry statement to obtain a knowledge map inquiry result, generating a corresponding target language according to the knowledge map inquiry result and sending the target language to the terminal.

Description

Medical field intention recognition method, device, equipment and storage medium
Technical Field
The present invention relates to the field of medical data, and in particular, to a method, an apparatus, a device, and a storage medium for medical field intention recognition.
Background
With the application of computer technology in the medical field, the on-line inquiry mode gradually breaks through the traditional medical treatment limitation, brings convenience and efficient medical experience for users, can meet the medical treatment requirements of the users without leaving the home, avoids the troubles of remote routes, registration and queuing and the like, saves medical resources and improves the inquiry efficiency. With the development of natural language processing technology, the on-line inquiry system gradually develops towards intellectualization, for example, an intelligent inquiry-answering engine is introduced into the inquiry system, so that a doctor can be replaced to answer user questions in the inquiry process, and auxiliary decision support can be provided for the doctor, so that the inquiry process is more efficient.
In the traditional medical question-answering system, a large number of manually corrected question answers are used as a knowledge base, and the question answers with the highest matching and user question similarities are fed back to the user in a text similarity mode. However, because the description modes of the diseased groups and the disease symptoms, the corresponding treatment modes and the like in the treatment process have diversity and specificity, fixed question-answering knowledge cannot be covered, an inference mechanism cannot be formed, and a large amount of labor cost is needed for maintaining a knowledge base, the template-based medical question-answering system adopts a rule matching or sentence matching mode to identify the intentions, cannot fully cover various question description modes, and has low accuracy in identifying the intentions in the medical field.
Disclosure of Invention
The invention provides a method, a device, equipment and a storage medium for identifying medical field intentions, which solve the problem of low accuracy in identifying medical field intentions.
A first aspect of an embodiment of the present invention provides a medical field intention identification method, including: acquiring an initial question sentence from a terminal, wherein the initial question sentence is a question sentence input by a target user in a medical intelligent question-answering system; calling a preset identification model to perform entity identification on the initial question statement to obtain an entity identification result, wherein the entity identification result comprises a plurality of coarse-grained entity labels and a plurality of entity relations; entity linkage is carried out on the plurality of coarse-grained entity labels according to a preset medical entity synonym table, and linked entity labels are obtained; performing intention recognition on the initial question sentence according to a preset intention recognition model, the entity recognition result and the linked entity label to obtain a candidate medical intention; generating a knowledge graph query statement according to the candidate medical intention; and inquiring the medical knowledge map in a preset medical knowledge map based on the knowledge map inquiry statement to obtain a knowledge map inquiry result, generating a corresponding target language according to the knowledge map inquiry result and sending the target language to the terminal.
Optionally, in a first implementation manner of the first aspect of the embodiment of the present invention, the invoking a preset recognition model to perform entity recognition on the initial question statement to obtain an entity recognition result, where the entity recognition result includes a plurality of coarse-grained entity labels and a plurality of entity relationships, and includes: calling a first preset identification model to perform entity identification on the initial question statement to obtain a plurality of coarse-grained entity labels; calling a second preset identification model to extract the relation of the initial question statement to obtain a plurality of entity relations; and generating an entity identification result according to the coarse-grained entity labels and the entity relations.
Optionally, in a second implementation manner of the first aspect of the embodiment of the present invention, the invoking a first preset recognition model to perform entity recognition on the initial question statement to obtain a plurality of coarse-grained entity labels includes: calling a first preset identification model to perform entity identification on the initial question statement according to the fine granularity to obtain a plurality of fine-granularity entity labels; and calling a first preset identification model to perform entity identification on the fine-grained entity labels according to the coarse granularity to obtain a plurality of coarse-grained entity labels.
Optionally, in a third implementation manner of the first aspect of the embodiment of the present invention, the invoking a first preset recognition model to perform entity recognition on the initial question statement according to a fine granularity to obtain a plurality of fine-grained entity tags includes: extracting a plurality of characteristic dimension vectors for the initial problem according to fine granularity, wherein the characteristic dimension vectors comprise word vectors, word label vectors, word position vectors and part-of-speech feature vectors; inputting the characteristic dimension vectors into a BilSTM layer of a first preset identification model to obtain a plurality of intermediate vectors output by the BilSTM layer; and inputting the plurality of intermediate vectors into a CRF layer of a first preset identification model to generate a plurality of fine-grained entity labels.
Optionally, in a fourth implementation manner of the first aspect of the embodiment of the present invention, the invoking a first preset identification model to perform entity identification on the fine-grained entity tags according to coarse granularity to obtain a plurality of coarse-grained entity tags includes: calling a first preset recognition model to recognize the fine-grained entity labels according to the coarse granularity to obtain a plurality of narrow-sense entity characteristics and a plurality of limited entity characteristics, wherein the narrow-sense entity characteristics comprise symptoms, diseases, parts, medicine, examination and treatment, and the limited entity characteristics comprise time, frequency, degree, negative words, description and numerical values; combining the plurality of narrow-sense entity features and the plurality of limited entity features according to preset rules to generate a plurality of generalized entity features, wherein the generalized entity features comprise generalized symptoms, generalized inspection, generalized treatment and generalized drugs; a plurality of generalized entity features is determined as a plurality of coarse-grained entity labels.
Optionally, in a fifth implementation manner of the first aspect of the embodiment of the present invention, the performing entity linking on the multiple coarse-grained entity labels according to a preset medical entity synonym table to obtain linked entity labels includes: searching a plurality of standard medical terms corresponding to a plurality of coarse-grained entity labels in a preset medical entity synonym table, wherein each coarse-grained entity label corresponds to one standard medical term, and the coarse-grained entity labels and the standard medical terms are synonyms; fusing the plurality of coarse-grained entity labels to obtain a plurality of fused coarse-grained entity labels; and carrying out entity linking operation on the fused coarse-grained entity labels and the standard medical terms to generate linked entity labels.
Optionally, in a sixth implementation manner of the first aspect of the embodiment of the present invention, the performing, based on the knowledge-graph query statement, a knowledge-graph query on a preset medical knowledge graph to obtain a knowledge-graph query result, generating a corresponding target utterance according to the knowledge-graph query result, and sending the target utterance to the terminal includes: inquiring in a preset medical knowledge map based on the knowledge map inquiry statement to obtain a knowledge map inquiry result, wherein the knowledge map inquiry result comprises the relation of a target entity, the attribute of the target entity and a plurality of entities; and generating a corresponding target language operation according to the relation of the target entity and the attribute of the target entity, and sending the target language operation to a terminal.
A second aspect of an embodiment of the present invention provides a medical field intention identifying apparatus, including: the system comprises a statement acquisition module, a question generation module and a question generation module, wherein the statement acquisition module is used for acquiring an initial question statement from a terminal, and the initial question statement is a question statement input by a target user in a medical intelligent question-answering system; the entity recognition module is used for calling a preset recognition model to perform entity recognition on the initial question statement to obtain an entity recognition result, and the entity recognition result comprises a plurality of coarse-grained entity labels and a plurality of entity relations; the entity linking module is used for carrying out entity linking on the plurality of coarse-grained entity labels according to a preset medical entity synonym table to obtain linked entity labels; the intention identification module is used for carrying out intention identification on the initial question sentence according to a preset intention identification model, the entity identification result and the linked entity label to obtain a candidate medical intention; the statement generation module is used for generating a knowledge graph query statement according to the candidate medical intention; and the map query module is used for querying a knowledge map in a preset medical knowledge map based on the knowledge map query statement to obtain a knowledge map query result, generating a corresponding target language according to the knowledge map query result and sending the target language to the terminal.
Optionally, in a first implementation manner of the second aspect of the embodiment of the present invention, the entity identifying module includes: the entity identification unit is used for calling a first preset identification model to perform entity identification on the initial question statement to obtain a plurality of coarse-grained entity labels; the relation extraction unit is used for calling a second preset identification model to extract the relation of the initial question statement to obtain a plurality of entity relations; and the generating unit is used for generating an entity identification result according to the plurality of coarse-grained entity labels and the plurality of entity relations.
Optionally, in a second implementation manner of the second aspect of the embodiment of the present invention, the entity identifying unit includes: the first identification subunit is used for calling a first preset identification model to perform entity identification on the initial question statement according to the fine granularity to obtain a plurality of fine-granularity entity labels; and the second identification subunit is used for calling the first preset identification model to perform entity identification on the fine-grained entity labels according to the coarse granularity to obtain a plurality of coarse-grained entity labels.
Optionally, in a third implementation manner of the second aspect of the embodiment of the present invention, the first identification subunit is specifically configured to: extracting a plurality of characteristic dimension vectors for the initial problem according to fine granularity, wherein the characteristic dimension vectors comprise word vectors, word label vectors, word position vectors and part-of-speech feature vectors; inputting the characteristic dimension vectors into a BilSTM layer of a first preset identification model to obtain a plurality of intermediate vectors output by the BilSTM layer; and inputting the plurality of intermediate vectors into a CRF layer of a first preset identification model to generate a plurality of fine-grained entity labels.
Optionally, in a fourth implementation manner of the second aspect of the embodiment of the present invention, the second identification subunit is specifically configured to: calling a first preset recognition model to recognize the fine-grained entity labels according to the coarse granularity to obtain a plurality of narrow-sense entity characteristics and a plurality of limited entity characteristics, wherein the narrow-sense entity characteristics comprise symptoms, diseases, parts, medicine, examination and treatment, and the limited entity characteristics comprise time, frequency, degree, negative words, description and numerical values; combining the plurality of narrow-sense entity features and the plurality of limited entity features according to preset rules to generate a plurality of generalized entity features, wherein the generalized entity features comprise generalized symptoms, generalized inspection, generalized treatment and generalized drugs; a plurality of generalized entity features is determined as a plurality of coarse-grained entity labels.
Optionally, in a fifth implementation manner of the second aspect of the embodiment of the present invention, the entity linking module is specifically configured to: searching a plurality of standard medical terms corresponding to a plurality of coarse-grained entity labels in a preset medical entity synonym table, wherein each coarse-grained entity label corresponds to one standard medical term, and the coarse-grained entity labels and the standard medical terms are synonyms; fusing the plurality of coarse-grained entity labels to obtain a plurality of fused coarse-grained entity labels; and carrying out entity linking operation on the fused coarse-grained entity labels and the standard medical terms to generate linked entity labels.
Optionally, in a sixth implementation manner of the second aspect of the embodiment of the present invention, the atlas query module is specifically configured to: inquiring in a preset medical knowledge map based on the knowledge map inquiry statement to obtain a knowledge map inquiry result, wherein the knowledge map inquiry result comprises the relation of a target entity, the attribute of the target entity and a plurality of entities; and generating a corresponding target language operation according to the relation of the target entity and the attribute of the target entity, and sending the target language operation to a terminal.
A third aspect of embodiments of the present invention provides a medical field intention identifying device, a memory having instructions stored therein and at least one processor interconnected by a line; the at least one processor invokes the instructions in the memory to cause the medical field intent recognition device to perform the medical field intent recognition method described above.
A fourth aspect of an embodiment of the present invention provides a computer-readable storage medium storing instructions that, when executed by a processor, implement the steps of the medical field intention identifying method according to any one of the above-mentioned embodiments.
According to the technical scheme provided by the embodiment of the invention, an initial question sentence is obtained from a terminal, and the initial question sentence is a question sentence input by a target user in a medical intelligent question-answering system; calling a preset identification model to perform entity identification on the initial question statement to obtain an entity identification result, wherein the entity identification result comprises a plurality of coarse-grained entity labels and a plurality of entity relations; entity linkage is carried out on the plurality of coarse-grained entity labels according to a preset medical entity synonym table, and linked entity labels are obtained; performing intention recognition on the initial question sentence according to a preset intention recognition model, an entity recognition result and the linked entity label to obtain a candidate medical intention; generating a knowledge graph query statement according to the candidate medical intention; and inquiring the preset medical knowledge map based on the knowledge map inquiry statement to obtain a knowledge map inquiry result, generating a corresponding target language according to the knowledge map inquiry result and sending the target language to the terminal. According to the embodiment of the invention, the deep learning model fusing the multi-dimensional characteristics is adopted to separately perform entity identification and relationship extraction, and meanwhile, the coarse-grained entity identification is adopted to optimize the fine-grained entity identification result, so that the interference of error transmission and redundant information in the entity extraction process is reduced, the accuracy of the entity identification result is improved, and the accuracy of the intention identification result in the medical field is further improved.
Drawings
FIG. 1 is a schematic diagram of an embodiment of a medical field intention identification method according to an embodiment of the invention;
FIG. 2 is a schematic diagram of another embodiment of a medical field intention identification method according to an embodiment of the invention;
FIG. 3 is a schematic diagram of one embodiment of a medical field intention identifying apparatus in an embodiment of the invention;
FIG. 4 is a schematic diagram of another embodiment of a medical field intention identifying apparatus in an embodiment of the invention;
fig. 5 is a schematic diagram of an embodiment of a medical field intention identifying device in the embodiment of the invention.
Detailed Description
The invention provides a medical field intention identification method, a medical field intention identification device, equipment and a storage medium, which are used for reducing the interference of error transmission and redundant information in the entity extraction process, improving the accuracy of an entity identification result and further improving the accuracy of a medical field intention identification result.
In order to make the technical field of the invention better understand the scheme of the invention, the embodiment of the invention will be described in conjunction with the attached drawings in the embodiment of the invention.
The terms "first," "second," "third," "fourth," and the like in the description and in the claims, as well as in the drawings, if any, are used for distinguishing between similar elements and not necessarily for describing a particular sequential or chronological order. It will be appreciated that the data so used may be interchanged under appropriate circumstances such that the embodiments described herein may be practiced otherwise than as specifically illustrated or described herein. Furthermore, the terms "comprises," "comprising," or "having," and any variations thereof, are intended to cover non-exclusive inclusions, such that a process, method, system, article, or apparatus that comprises a list of steps or elements is not necessarily limited to those steps or elements expressly listed, but may include other steps or elements not expressly listed or inherent to such process, method, article, or apparatus.
Referring to fig. 1, a flowchart of a method for identifying an intention in a medical field according to an embodiment of the present invention specifically includes:
101. and acquiring an initial question sentence from the terminal, wherein the initial question sentence is the question sentence input by the target user in the medical intelligent question-answering system.
The server acquires an initial question sentence from the terminal, wherein the initial question sentence is a question sentence input by a target user in the medical intelligent question-answering system. The initial question sentence is a medical knowledge question that the user wants to know, for example, "do you have cephalo and drink? "," which department should the muscle soreness visit? "this embodiment does not limit the consultation field of the initial question sentence as long as it is related to medical treatment.
It is to be understood that the executing subject of the present invention may be a medical field intention identifying apparatus, and may also be a server, which is not limited herein. The embodiment of the present invention is described by taking a server as an execution subject.
102. And calling a preset identification model to perform entity identification on the initial question statement to obtain an entity identification result, wherein the entity identification result comprises a plurality of coarse-grained entity labels and a plurality of entity relations.
Specifically, the server calls a first preset identification model to perform entity identification on the initial question statement to obtain a plurality of coarse-grained entity labels; calling a second preset identification model to extract the relation of the initial question statement to obtain a plurality of entity relations; and generating an entity identification result according to the plurality of coarse-grained entity labels and the plurality of entity relationships.
The server can call a BILSTM layer of a second preset identification model to extract the context of the initial question statement to obtain a plurality of time sequence vectors, and the time sequence vectors are used for indicating the context; inputting a plurality of time sequence vectors into an Attention attribute layer of a second preset identification model to generate a plurality of sentence characteristic vectors, wherein the sentence characteristic vectors are used for indicating entity relationships; the Attention layer calculates the weight of each time sequence vector, then takes the weighted sum of all time sequence vectors as a characteristic vector, and then carries out softmax classification.
Optionally, the server calls a first preset recognition model to perform entity recognition on the initial question statement to obtain a plurality of coarse-grained entity labels, and the method specifically includes:
the server calls a first preset identification model to perform entity identification on the initial question statement according to the fine granularity to obtain a plurality of fine-granularity entity labels; and the server calls a first preset identification model to perform entity identification on the fine-grained entity labels according to the coarse granularity to obtain a plurality of coarse-grained entity labels.
In the embodiment, entity recognition and relation extraction are separately performed by adopting the deep learning model fusing the multi-dimensional characteristics, so that the interference of error transmission and redundant information is reduced, and meanwhile, the fine-grained entity recognition result is optimized by adopting coarse-grained entity recognition, so that the recognition accuracy can be further improved.
103. And carrying out entity linkage on the plurality of coarse-grained entity labels according to a preset medical entity synonym table to obtain linked entity labels.
And the server performs entity linkage on the plurality of coarse-grained entity labels according to a preset medical entity synonym table to obtain linked entity labels. Specifically, the server searches a plurality of standard medical terms corresponding to a plurality of coarse-grained entity labels in a preset medical entity synonym table, each coarse-grained entity label corresponds to one standard medical term, and the coarse-grained entity labels and the standard medical terms are synonyms; the server fuses the plurality of coarse-grained entity labels to obtain a plurality of fused coarse-grained entity labels; and the server performs entity linking operation on the plurality of fused coarse-grained entity labels and the plurality of standard medical terms to generate linked entity labels.
In this embodiment, some users mainly express a spoken medical word, and perform entity linking operation to link to a standard medical term, for example, the user describes "post-abortion", the corresponding medical term is "post-abortion", the "post-abortion" is linked to "post-abortion", and for example, the user describes "lower abdominal distending pain" and needs to link to the standard term of "lower abdominal distending pain"; both "pregnancy 34 +" and "pregnancy 40 +" correspond to the standard medical data "late pregnancy", thus linking both "pregnancy 34 +" and "pregnancy 40 +" to "late pregnancy".
It should be noted that, before entity linking, the coarse-grained entity labels may also be fused to obtain fused coarse-grained entity labels, for example, "pregnant 34 +" and "pregnant 40 +" both belong to late pregnancy, "pregnant 34 +" and "pregnant 40 +" may be fused to "pregnant 34 to 40 weeks," pregnant 34 to 40 weeks "are fused coarse-grained entity labels, and then the fused coarse-grained entity labels are linked to standard medical terms.
104. And performing intention recognition on the initial question sentence according to a preset intention recognition model, an entity recognition result and the linked entity label to obtain a candidate medical intention.
And the server performs intention identification on the initial question sentence according to a preset intention identification model, the entity identification result and the linked entity label to obtain a candidate medical intention.
The intention recognition model is a deep learning model and consists of an input layer, a BERT word vector layer, a BilSTM layer, an Attention layer and a Softmax classification layer; because the problem intention is greatly related to the entity and the entity tag, the entity recognition result and the linked entity tag are also used as the input of the intention recognition model in the embodiment, and the initial problem statement, the recognition result and the linked entity tag are jointly used as the sentence input of the input layer in the embodiment.
The BERT word vector layer generates word vectors from input sentences, and the output of the BERT word vector layer is used as the input of the BilSTM layer; taking the full-connection output of the BilSTM layer as the input of the Attention layer; and finally classifying the intention label of the output of the Attention layer by adopting a Softmax classifier to obtain candidate medical intentions, wherein the intention types comprise: cause, explanation, complications, mode of transmission, treatment, related examinations, disease diagnosis, precautions, efficacy, side effects/hazards, method of operation, method of use/administration, amount of use, dietary recommendations, whether or not, etc.
In the embodiment, the deep learning model is adopted for identifying the user intention, so that the number of templates is reduced, the coverage rate and accuracy of the question and answer condition in the real conversation are improved, and the maintenance cost is reduced.
105. And generating a knowledge graph query statement according to the candidate medical intention.
The server generates a knowledge graph query statement according to the candidate medical intention.
In this embodiment, the entity identification result and the intention identification result of the initial question statement are combined to perform query mapping of the knowledge graph, and a knowledge graph query statement is generated, where a query object may be a relationship between entities or an attribute of an entity.
106. And inquiring the preset medical knowledge map based on the knowledge map inquiry statement to obtain a knowledge map inquiry result, generating a corresponding target language according to the knowledge map inquiry result and sending the target language to the terminal.
Specifically, the server queries a preset medical knowledge graph based on a knowledge graph query statement to obtain a knowledge graph query result, wherein the knowledge graph query result comprises the relation of target entities, the attributes of the target entities and a plurality of entities; and generating a corresponding target language operation according to the relation of the target entity and the attribute of the target entity, and sending the target language operation to the terminal. .
For different knowledge graph query results, namely different entity types are not queried, the entity types comprise the relationship of the entities, the attributes of the entities and the entities, and the specific process is as follows:
and if the knowledge graph query result is the relationship of the query target entity, the relationship of the query target entity is obtained through the knowledge graph query statement. For example, when the user's initial question sentence is "what complications are in cirrhosis? When "the entity was identified as" cirrhosis: the Disease "is intended to be identified as" Complication ", the corresponding knowledge graph query statement is" match (n: Disease) - [ r: compatibility ] - (m: Symptom) where n.name is "cirrhosis" return m.name ", the name attributes of the nodes connecting the labels of the cirrhosis Complication relations as symptoms are combined to generate the target operation" cirrhosis complications include impaired liver function, portal hypertension, gastrointestinal hemorrhage, hepatic encephalopathy, peritonitis, and the like. ", sending the target utterance to the terminal.
And if the knowledge graph query result is the attribute of the query target entity, the attribute of the query target entity is queried through the knowledge graph query statement. For example, when the initial question sentence of the user is "what side effect is with fibrate lipid lowering drugs? The entity extraction result is' fibrate lipid-lowering drugs: the medicine "is intended to identify that the result is" side effect/harm ", and the corresponding map query statement is" match (n: Drug) where n.name is "fibrate lipid-lowering medicine" return n.harm ", so that the target" fibrate lipid-lowering medicine adverse reaction "is generated according to the side effect attribute of the fibrate lipid-lowering medicine, and the target" fibrate lipid-lowering medicine adverse reaction is gastrointestinal discomfort, rash, alopecia, headache, hyposexuality and the like. ", and sending the target language to the terminal;
if the knowledge-graph query result is to query multiple entities, for example, the initial question sentence of the user is "what need to be noticed by farrow during pregnancy? ", entity extraction results are" pregnancy: special periods, hip pain: symptom ", the intention recognition result is" notice ", and the corresponding graph query statement is: "match (n: professional period { name:" pregnancy period ") - [: MultiConditionRestriction ] - > (p: spandex), (m: symmetry { name:" flatus pain "-) - [: MultiConditionRestriction ] - > (p: spandex) return p.attribute", is determined, attention item attribute values of blank nodes having a relation with both pregnancy period and flatus pain are determined, and target words "flatus pain of pregnant women can be hot-compressed on the pain place by using a hot towel and a hot water bag for about half an hour, pain feeling can be relieved by a few", and the target words are sent to a terminal.
It can be understood that the server performs knowledge graph query statement conversion by combining entity types, makes a personalized language and feeds back results to a terminal used by a user, and can provide auxiliary decision support for doctors in online inquiry application, so that the inquiry process is more efficient.
According to the embodiment of the invention, the deep learning model fusing the multi-dimensional characteristics is adopted to separately perform entity identification and relationship extraction, and meanwhile, the coarse-grained entity identification is adopted to optimize the fine-grained entity identification result, so that the interference of error transmission and redundant information in the entity extraction process is reduced, the accuracy of the entity identification result is improved, and the accuracy of the intention identification result in the medical field is further improved. And this scheme can be applied to in the wisdom medical treatment field to promote the construction in wisdom city.
Referring to fig. 2, another flowchart of the method for identifying an intention in a medical field according to an embodiment of the present invention specifically includes:
201. and constructing a preset medical knowledge map.
The server constructs a preset medical knowledge map. The method specifically comprises the following steps:
(1) the server obtains a plurality of data sources including structured medical data, semi-structured medical data, and online medical interrogation session data.
The structured medical data mainly come from data related to diseases, medicines and examination and inspection in a relational database which is stored in business, the semi-structured medical data mainly come from medical data of Wikipedia and medical data of Baidu encyclopedia, and the data are stored as the semi-structured data after being clear. The text content of the structured semi-structured data is long, the speciality is high, and the text content is not easy to understand by a user, so that when a medical knowledge map is constructed, problem answer knowledge (namely on-line medical inquiry dialogue data) generated in an on-line inquiry dialogue after being corrected by a doctor is also used as one of data sources, the scheme of the application is more prone to simulating a real inquiry scene dialogue, and the inquiry experience of the user is optimized.
(2) And the server performs entity extraction on the plurality of data sources to obtain a plurality of entities and a plurality of entity relationships, and sets entity attributes corresponding to the plurality of entities and relationship attributes corresponding to the plurality of entity relationships.
In the embodiment, a top-down mode is adopted for map construction, namely, an entity identification and relationship extraction method based on a deep learning model is adopted for carrying out entity identification and relationship extraction on structured medical data and semi-structured medical data, and the entity identification and relationship extraction method is added into a knowledge map.
Optionally, the step (2) specifically includes:
the server performs entity recognition and relation extraction on the structured medical data based on the deep learning model; the server performs entity identification and relation extraction on the semi-structured medical data by adopting a deep learning model; the server generates a plurality of entities and a plurality of entity relationships; the server sets corresponding attributes for each entity respectively to obtain a plurality of entity attributes, and sets corresponding attributes for each entity relationship to obtain a plurality of entity relationship attributes.
Wherein the plurality of entities comprise departments, diseases, symptoms, medicines, treatment means, foods and health products, and the entity relationship comprises the treatment departments, relevant symptoms, suitable medicines and complications. Different types of entities or relationships can be set with different attributes, for example, the entity "disease" corresponds to attributes such as "explanation", "etiology", "morbidity", etc., the entity "drug" corresponds to attributes such as "specification", "efficacy", "contraindication", etc., and the entity relationship "complication" corresponds to "shock", "infection", etc.
(3) And the server adopts a preset deep learning model to construct an initial knowledge map according to the entities, the entity attributes corresponding to the entities, the entity relationships and the relationship attributes corresponding to the entity relationships.
(4) And the server performs entity alignment and relationship fusion on the initial knowledge graph to generate a preset medical knowledge graph.
The purpose of entity alignment and relationship fusion is to discover and merge multi-source heterogeneous entities which have different entity names but represent the same concept and object in different data sources, and merge attributes and relationships of the entities. The entity alignment is a commonly used entity alignment method based on attribute similarity score, and is not described herein for the prior art.
202. And acquiring an initial question sentence from the terminal, wherein the initial question sentence is the question sentence input by the target user in the medical intelligent question-answering system.
The server acquires an initial question sentence from the terminal, wherein the initial question sentence is a question sentence input by a target user in the medical intelligent question-answering system. The initial question sentence is a medical knowledge question that the user wants to know, for example, "do you have cephalo and drink? "," which department should the muscle soreness visit? "this embodiment does not limit the consultation field of the initial question sentence as long as it is related to medical treatment.
It is to be understood that the executing subject of the present invention may be a medical field intention identifying apparatus, and may also be a server, which is not limited herein. The embodiment of the present invention is described by taking a server as an execution subject.
203. And calling a first preset identification model to perform entity identification on the initial question statement to obtain a plurality of coarse-grained entity labels.
Specifically, the server calls a first preset identification model to perform entity identification on the initial question statement according to the fine granularity to obtain a plurality of fine-granularity entity labels; and the server calls a first preset identification model to perform entity identification on the fine-grained entity labels according to the coarse granularity to obtain a plurality of coarse-grained entity labels.
Optionally, the step of calling, by the server, the first preset identification model to perform entity identification on the initial question statement according to the fine granularity, and obtaining the plurality of fine-granularity entity tags specifically includes:
the server extracts a plurality of characteristic dimension vectors from the initial problem according to the fine granularity, wherein the plurality of characteristic dimension vectors comprise word vectors, word label vectors, word position vectors and part-of-speech feature vectors; the server inputs the characteristic dimension vectors into a BilSTM layer of a first preset identification model to obtain a plurality of intermediate vectors output by the BilSTM layer; and the server inputs a plurality of intermediate vectors into a CRF layer of the first preset identification model to generate a plurality of fine-grained entity labels.
The word label vector is a word label coded by BIOES, the word position characteristic vector is a position vector of a word cut by the jieba word segmentation tool, and the part-of-speech characteristic vector is a part-of-speech vector of the word marked by the jieba word segmentation tool.
It should be noted that the chinese words have no clear boundary information, and the same words constitute different semantics of words in different sequences, such as "lying-in woman" and "obstetrics and gynecology" in which "lying-in woman's belly pain should go to the medical visit of obstetrics and gynecology immediately, the former is labeled as" crowd ", and the latter is labeled as" department ", so that the position information of the words can be used as an effective feature. The part of speech is an important attribute of a word, so that more abstract word characteristics can be expressed, the structural relation of a sentence is further discovered, entity labels such as 'disease', 'symptom', 'crowd' and the like are nouns, and the part of speech and a named entity have strong association relation, so that the performance of entity identification can be further improved by adding part of speech information into a model. After the experiment comparison shows that the word position and the part-of-speech characteristic are added, the recognition accuracy of the preset recognition model is improved by 5 percent.
Optionally, the step of calling the first preset identification model by the server to perform entity identification on the fine-grained entity labels according to the coarse granularity to obtain the plurality of coarse-grained entity labels specifically includes:
the server calls a first preset recognition model to recognize a plurality of fine-grained entity labels according to the coarse granularity to obtain a plurality of narrow-sense entity characteristics and a plurality of limited entity characteristics, wherein the narrow-sense entity characteristics comprise symptoms, diseases, parts, medicine, examination and treatment, and the limited entity characteristics comprise time, frequency, degree, negative words, description and numerical values; the server combines the plurality of narrow-sense entity characteristics and the plurality of limited entity characteristics according to preset rules to generate a plurality of generalized entity characteristics, wherein the generalized entity characteristics comprise generalized symptoms, generalized inspection, generalized treatment and generalized drugs; the server determines a plurality of generalized entity characteristics as a plurality of coarse-grained entity labels.
For example, the user question sentence is "what reason did i ask from morning pain to evening in the latest time for the doctor's hello? "head" is a body part according to fine-grained entity recognition, "pain" is a descriptive phrase, "early" is time, "late" is time, and "head from early pain to late" is recognized as a generalized symptom according to a coarse-grained entity recognition rule.
204. And calling a second preset identification model to extract the relation of the initial question statement to obtain a plurality of entity relations.
And the server calls a second preset identification model to extract the relation of the initial question statement to obtain a plurality of entity relations. The server can call a BILSTM layer of a second preset identification model to extract the context of the initial question statement to obtain a plurality of time sequence vectors, and the time sequence vectors are used for indicating the context; inputting a plurality of time sequence vectors into an Attention attribute layer of a second preset identification model to generate a plurality of sentence characteristic vectors, wherein the sentence characteristic vectors are used for indicating entity relationships; the Attention layer calculates the weight of each time sequence vector, then takes the weighted sum of all time sequence vectors as a characteristic vector, and then carries out softmax classification.
205. And generating an entity identification result according to the plurality of coarse-grained entity labels and the plurality of entity relationships.
And the server generates an entity identification result according to the plurality of coarse-grained entity labels and the plurality of entity relationships.
In the embodiment, the server separately performs entity identification and relationship extraction by adopting the deep learning model fusing the multi-dimensional characteristics, so that the interference of error transmission and redundant information is reduced, and meanwhile, the coarse-grained entity identification is adopted to optimize the fine-grained entity identification result, so that the identification accuracy can be further improved.
206. And carrying out entity linkage on the plurality of coarse-grained entity labels according to a preset medical entity synonym table to obtain linked entity labels.
And the server performs entity linkage on the plurality of coarse-grained entity labels according to a preset medical entity synonym table to obtain linked entity labels. Specifically, the server searches a plurality of standard medical terms corresponding to a plurality of coarse-grained entity labels in a preset medical entity synonym table, each coarse-grained entity label corresponds to one standard medical term, and the coarse-grained entity labels and the standard medical terms are synonyms; the server fuses the plurality of coarse-grained entity labels to obtain a plurality of fused coarse-grained entity labels; and the server performs entity linking operation on the plurality of fused coarse-grained entity labels and the plurality of standard medical terms to generate linked entity labels.
In this embodiment, some users mainly express a spoken medical word, and perform entity linking operation to link to a standard medical term, for example, the user describes "post-abortion", the corresponding medical term is "post-abortion", the "post-abortion" is linked to "post-abortion", and for example, the user describes "lower abdominal distending pain" and needs to link to the standard term of "lower abdominal distending pain"; both "pregnancy 34 +" and "pregnancy 40 +" correspond to the standard medical data "late pregnancy", thus linking both "pregnancy 34 +" and "pregnancy 40 +" to "late pregnancy".
It should be noted that, before entity linking, the coarse-grained entity labels may also be fused to obtain fused coarse-grained entity labels, for example, "pregnant 34 +" and "pregnant 40 +" both belong to late pregnancy, "pregnant 34 +" and "pregnant 40 +" may be fused to "pregnant 34 to 40 weeks," pregnant 34 to 40 weeks "are fused coarse-grained entity labels, and then the fused coarse-grained entity labels are linked to standard medical terms.
It can be understood that entity normalization and entity fusion operations are performed on the same entities of different data sources, and a synonym table of the medical entity is maintained, so that a redundancy-removing conflict-removing medical knowledge graph is constructed, and the question-answering system is guaranteed to have higher-quality data support.
207. And performing intention recognition on the initial question sentence according to a preset intention recognition model, an entity recognition result and the linked entity label to obtain a candidate medical intention.
And the server performs intention identification on the initial question sentence according to a preset intention identification model, the entity identification result and the linked entity label to obtain a candidate medical intention.
The intention recognition model is a deep learning model and consists of an input layer, a BERT word vector layer, a BilSTM layer, an Attention layer and a Softmax classification layer; because the problem intention is greatly related to the entity and the entity tag, the entity recognition result and the linked entity tag are also used as the input of the intention recognition model in the embodiment, and the initial problem statement, the recognition result and the linked entity tag are jointly used as the sentence input of the input layer in the embodiment.
The BERT word vector layer generates word vectors from input sentences, and the output of the BERT word vector layer is used as the input of the BilSTM layer; taking the full-connection output of the BilSTM layer as the input of the Attention layer; and finally classifying the intention label of the output of the Attention layer by adopting a Softmax classifier to obtain candidate medical intentions, wherein the intention types comprise: cause, explanation, complications, mode of transmission, treatment, related examinations, disease diagnosis, precautions, efficacy, side effects/hazards, method of operation, method of use/administration, amount of use, dietary recommendations, whether or not, etc.
In the embodiment, the deep learning model is adopted for identifying the user intention, so that the number of templates is reduced, the coverage rate and accuracy of the question and answer condition in the real conversation are improved, and the maintenance cost is reduced.
208. And generating a knowledge graph query statement according to the candidate medical intention.
The server generates a knowledge graph query statement according to the candidate medical intention.
In this embodiment, the entity identification result and the intention identification result of the initial question statement are combined to perform query mapping of the knowledge graph, and a knowledge graph query statement is generated, where a query object may be a relationship between entities or an attribute of an entity.
209. And inquiring the preset medical knowledge map based on the knowledge map inquiry statement to obtain a knowledge map inquiry result, generating a corresponding target language according to the knowledge map inquiry result and sending the target language to the terminal.
Specifically, the server queries a preset medical knowledge graph based on a knowledge graph query statement to obtain a knowledge graph query result, wherein the knowledge graph query result comprises the relation of target entities, the attributes of the target entities and a plurality of entities; and generating a corresponding target language operation according to the relation of the target entity and the attribute of the target entity, and sending the target language operation to the terminal. .
For different knowledge graph query results, namely different entity types are not queried, the entity types comprise the relationship of the entities, the attributes of the entities and the entities, and the specific process is as follows:
and if the knowledge graph query result is the relationship of the query target entity, the relationship of the query target entity is obtained through the knowledge graph query statement. For example, when the user's initial question sentence is "what complications are in cirrhosis? When "the entity was identified as" cirrhosis: the Disease "is intended to be identified as" Complication ", the corresponding knowledge graph query statement is" match (n: Disease) - [ r: compatibility ] - (m: Symptom) where n.name is "cirrhosis" return m.name ", the name attributes of the nodes connecting the labels of the cirrhosis Complication relations as symptoms are combined to generate the target operation" cirrhosis complications include impaired liver function, portal hypertension, gastrointestinal hemorrhage, hepatic encephalopathy, peritonitis, and the like. ", sending the target utterance to the terminal.
And if the knowledge graph query result is the attribute of the query target entity, the attribute of the query target entity is queried through the knowledge graph query statement. For example, when the initial question sentence of the user is "what side effect is with fibrate lipid lowering drugs? The entity extraction result is' fibrate lipid-lowering drugs: the medicine "is intended to identify that the result is" side effect/harm ", and the corresponding map query statement is" match (n: Drug) where n.name is "fibrate lipid-lowering medicine" return n.harm ", so that the target" fibrate lipid-lowering medicine adverse reaction "is generated according to the side effect attribute of the fibrate lipid-lowering medicine, and the target" fibrate lipid-lowering medicine adverse reaction is gastrointestinal discomfort, rash, alopecia, headache, hyposexuality and the like. ", and sending the target language to the terminal;
if the knowledge-graph query result is to query multiple entities, for example, the initial question sentence of the user is "what need to be noticed by farrow during pregnancy? ", entity extraction results are" pregnancy: special periods, hip pain: symptom ", the intention recognition result is" notice ", and the corresponding graph query statement is: "match (n: professional period { name:" pregnancy period ") - [: MultiConditionRestriction ] - > (p: spandex), (m: symmetry { name:" flatus pain "-) - [: MultiConditionRestriction ] - > (p: spandex) return p.attribute", is determined, attention item attribute values of blank nodes having a relation with both pregnancy period and flatus pain are determined, and target words "flatus pain of pregnant women can be hot-compressed on the pain place by using a hot towel and a hot water bag for about half an hour, pain feeling can be relieved by a few", and the target words are sent to a terminal.
It can be understood that the server performs knowledge graph query statement conversion by combining entity types, makes a personalized language and feeds back results to a terminal used by a user, and can provide auxiliary decision support for doctors in online inquiry application, so that the inquiry process is more efficient.
According to the embodiment of the invention, the deep learning model fusing the multi-dimensional characteristics is adopted to separately perform entity identification and relationship extraction, and meanwhile, the coarse-grained entity identification is adopted to optimize the fine-grained entity identification result, so that the interference of error transmission and redundant information in the entity extraction process is reduced, the accuracy of the entity identification result is improved, and the accuracy of the intention identification result in the medical field is further improved. And this scheme can be applied to in the wisdom medical treatment field to promote the construction in wisdom city.
In the above description of the method for recognizing medical field intention according to the embodiment of the present invention, referring to fig. 3, a medical field intention recognition apparatus according to an embodiment of the present invention is described below, and an embodiment of the medical field intention recognition apparatus according to an embodiment of the present invention includes:
a sentence acquisition module 301, configured to acquire an initial question sentence from a terminal, where the initial question sentence is a question sentence input by a target user in a medical intelligent question-answering system;
an entity identification module 302, configured to invoke a preset identification model to perform entity identification on the initial question statement, so as to obtain an entity identification result, where the entity identification result includes a plurality of coarse-grained entity labels and a plurality of entity relationships;
the entity linking module 303 is configured to perform entity linking on the plurality of coarse-grained entity labels according to a preset medical entity synonym table to obtain linked entity labels;
an intention identifying module 304, configured to perform intention identification on the initial question statement according to a preset intention identifying model, the entity identifying result, and the linked entity tag, so as to obtain a candidate medical intention;
a statement generation module 305 for generating a knowledge graph query statement according to the candidate medical intention;
the map query module 306 is configured to perform a knowledge map query on a preset medical knowledge map based on the knowledge map query statement to obtain a knowledge map query result, generate a corresponding target language according to the knowledge map query result, and send the target language to the terminal.
According to the embodiment of the invention, the deep learning model fusing the multi-dimensional characteristics is adopted to separately perform entity identification and relationship extraction, and meanwhile, the coarse-grained entity identification is adopted to optimize the fine-grained entity identification result, so that the interference of error transmission and redundant information in the entity extraction process is reduced, the accuracy of the entity identification result is improved, and the accuracy of the intention identification result in the medical field is further improved. And this scheme can be applied to in the wisdom medical treatment field to promote the construction in wisdom city.
Referring to fig. 4, another embodiment of the medical field intention recognition apparatus according to the embodiment of the present invention includes:
a sentence acquisition module 301, configured to acquire an initial question sentence from a terminal, where the initial question sentence is a question sentence input by a target user in a medical intelligent question-answering system;
an entity identification module 302, configured to invoke a preset identification model to perform entity identification on the initial question statement, so as to obtain an entity identification result, where the entity identification result includes a plurality of coarse-grained entity labels and a plurality of entity relationships;
the entity linking module 303 is configured to perform entity linking on the plurality of coarse-grained entity labels according to a preset medical entity synonym table to obtain linked entity labels;
an intention identifying module 304, configured to perform intention identification on the initial question statement according to a preset intention identifying model, the entity identifying result, and the linked entity tag, so as to obtain a candidate medical intention;
a statement generation module 305 for generating a knowledge graph query statement according to the candidate medical intention;
the map query module 306 is configured to perform a knowledge map query on a preset medical knowledge map based on the knowledge map query statement to obtain a knowledge map query result, generate a corresponding target language according to the knowledge map query result, and send the target language to the terminal.
Optionally, the entity identifying module 302 includes:
an entity identifying unit 3021, configured to invoke a first preset identification model to perform entity identification on the initial question statement, so as to obtain a plurality of coarse-grained entity tags;
a relation extracting unit 3022, configured to invoke a second preset identification model to perform relation extraction on the initial question statement, so as to obtain a plurality of entity relations;
a generating unit 3023, configured to generate an entity identification result according to the plurality of coarse-grained entity labels and the plurality of entity relationships.
Optionally, the entity identifying unit 3021 includes:
a first identifying subunit 30211, configured to invoke a first preset identifying model to perform entity identification on the initial question statement according to the fine granularity, so as to obtain a plurality of fine-granularity entity tags;
a second identifying subunit 30212, configured to invoke a first preset identifying model to perform entity identification on the fine-grained entity tags according to the coarse granularity, so as to obtain a plurality of coarse-grained entity tags.
Optionally, the first identifier unit 30211 is specifically configured to:
extracting a plurality of characteristic dimension vectors for the initial problem according to fine granularity, wherein the characteristic dimension vectors comprise word vectors, word label vectors, word position vectors and part-of-speech feature vectors; inputting the characteristic dimension vectors into a BilSTM layer of a first preset identification model to obtain a plurality of intermediate vectors output by the BilSTM layer; and inputting the plurality of intermediate vectors into a CRF layer of a first preset identification model to generate a plurality of fine-grained entity labels.
Optionally, the second identifier unit 30212 is specifically configured to:
calling a first preset recognition model to recognize the fine-grained entity labels according to the coarse granularity to obtain a plurality of narrow-sense entity characteristics and a plurality of limited entity characteristics, wherein the narrow-sense entity characteristics comprise symptoms, diseases, parts, medicine, examination and treatment, and the limited entity characteristics comprise time, frequency, degree, negative words, description and numerical values; combining the plurality of narrow-sense entity features and the plurality of limited entity features according to preset rules to generate a plurality of generalized entity features, wherein the generalized entity features comprise generalized symptoms, generalized inspection, generalized treatment and generalized drugs; a plurality of generalized entity features is determined as a plurality of coarse-grained entity labels.
Optionally, the entity linking module 303 is specifically configured to:
searching a plurality of standard medical terms corresponding to a plurality of coarse-grained entity labels in a preset medical entity synonym table, wherein each coarse-grained entity label corresponds to one standard medical term, and the coarse-grained entity labels and the standard medical terms are synonyms; fusing the plurality of coarse-grained entity labels to obtain a plurality of fused coarse-grained entity labels; and carrying out entity linking operation on the fused coarse-grained entity labels and the standard medical terms to generate linked entity labels.
Optionally, the map query module 306 is specifically configured to:
inquiring in a preset medical knowledge map based on the knowledge map inquiry statement to obtain a knowledge map inquiry result, wherein the knowledge map inquiry result comprises the relation of a target entity, the attribute of the target entity and a plurality of entities; and generating a corresponding target language operation according to the relation of the target entity and the attribute of the target entity, and sending the target language operation to a terminal.
According to the embodiment of the invention, the deep learning model fusing the multi-dimensional characteristics is adopted to separately perform entity identification and relationship extraction, and meanwhile, the coarse-grained entity identification is adopted to optimize the fine-grained entity identification result, so that the interference of error transmission and redundant information in the entity extraction process is reduced, the accuracy of the entity identification result is improved, and the accuracy of the intention identification result in the medical field is further improved. And this scheme can be applied to in the wisdom medical treatment field to promote the construction in wisdom city.
Fig. 3 to 4 describe the medical field intention identifying apparatus in the embodiment of the present invention in detail from the perspective of the modular functional entity, and the medical field intention identifying device in the embodiment of the present invention is described in detail from the perspective of hardware processing.
Fig. 5 is a schematic structural diagram of a medical field intention recognition apparatus 500 according to an embodiment of the present invention, where the medical field intention recognition apparatus 500 may have relatively large differences due to different configurations or performances, and may include one or more processors (CPUs) 510 (e.g., one or more processors) and a memory 520, and one or more storage media 530 (e.g., one or more mass storage devices) storing applications 533 or data 532. Memory 520 and storage media 530 may be, among other things, transient or persistent storage. The program stored on the storage medium 530 may include one or more modules (not shown), each of which may include a series of instruction operations for the medical field intention recognition device 500. Still further, the processor 510 may be configured to communicate with the storage medium 530 to execute a series of instruction operations in the storage medium 530 on the medical field intention identifying apparatus 500.
The medical field intent recognition device 500 may also include one or more power supplies 540, one or more wired or wireless network interfaces 550, one or more input-output interfaces 560, and/or one or more operating systems 531, such as Windows Server, Mac OS X, Unix, Linux, FreeBSD, etc. Those skilled in the art will appreciate that the medical field intent recognition device configuration shown in fig. 5 does not constitute a limitation of the medical field intent recognition device and may include more or fewer components than shown, or some components may be combined, or a different arrangement of components.
The present invention also provides a computer readable storage medium, which may be a non-volatile computer readable storage medium, which may also be a volatile computer readable storage medium, having stored therein instructions, which, when run on a computer, cause the computer to perform the steps of the medical field intent identification method.
It is clear to those skilled in the art that, for convenience and brevity of description, the specific working processes of the above-described systems, apparatuses and units may refer to the corresponding processes in the foregoing method embodiments, and are not described herein again.
The integrated unit, if implemented in the form of a software functional unit and sold or used as a stand-alone product, may be stored in a computer readable storage medium. Based on such understanding, the technical solution of the present invention may be embodied in the form of a software product, which is stored in a storage medium and includes instructions for causing a computer device (which may be a personal computer, a server, or a network device) to execute all or part of the steps of the method according to the embodiments of the present invention. And the aforementioned storage medium includes: various media capable of storing program codes, such as a usb disk, a removable hard disk, a read-only memory (ROM), a Random Access Memory (RAM), a magnetic disk, or an optical disk.
The above-mentioned embodiments are only used for illustrating the technical solutions of the present invention, and not for limiting the same; although the present invention has been described in detail with reference to the foregoing embodiments, it will be understood by those of ordinary skill in the art that: the technical solutions described in the foregoing embodiments may still be modified, or some technical features may be equivalently replaced; and such modifications or substitutions do not depart from the spirit and scope of the corresponding technical solutions of the embodiments of the present invention.

Claims (10)

1. A medical field intention recognition method, comprising:
acquiring an initial question sentence from a terminal, wherein the initial question sentence is a question sentence input by a target user in a medical intelligent question-answering system;
calling a preset identification model to perform entity identification on the initial question statement to obtain an entity identification result, wherein the entity identification result comprises a plurality of coarse-grained entity labels and a plurality of entity relations;
entity linkage is carried out on the plurality of coarse-grained entity labels according to a preset medical entity synonym table, and linked entity labels are obtained;
performing intention recognition on the initial question sentence according to a preset intention recognition model, the entity recognition result and the linked entity label to obtain a candidate medical intention;
generating a knowledge graph query statement according to the candidate medical intention;
and inquiring the medical knowledge map in a preset medical knowledge map based on the knowledge map inquiry statement to obtain a knowledge map inquiry result, generating a corresponding target language according to the knowledge map inquiry result and sending the target language to the terminal.
2. The method for recognizing medical field intention according to claim 1, wherein the calling a preset recognition model to perform entity recognition on the initial question statement to obtain an entity recognition result, wherein the entity recognition result includes a plurality of coarse-grained entity labels and a plurality of entity relationships, and includes:
calling a first preset identification model to perform entity identification on the initial question statement to obtain a plurality of coarse-grained entity labels;
calling a second preset identification model to extract the relation of the initial question statement to obtain a plurality of entity relations;
and generating an entity identification result according to the coarse-grained entity labels and the entity relations.
3. The medical field intention recognition method of claim 2, wherein the invoking a first preset recognition model to perform entity recognition on the initial question statement to obtain a plurality of coarse-grained entity labels comprises:
calling a first preset identification model to perform entity identification on the initial question statement according to the fine granularity to obtain a plurality of fine-granularity entity labels;
and calling a first preset identification model to perform entity identification on the fine-grained entity labels according to the coarse granularity to obtain a plurality of coarse-grained entity labels.
4. The medical field intention recognition method of claim 3, wherein the invoking of the first preset recognition model for entity recognition of the initial question statement according to a fine granularity obtains a plurality of fine-grained entity labels comprises:
extracting a plurality of characteristic dimension vectors for the initial problem according to fine granularity, wherein the characteristic dimension vectors comprise word vectors, word label vectors, word position vectors and part-of-speech feature vectors;
inputting the characteristic dimension vectors into a BilSTM layer of a first preset identification model to obtain a plurality of intermediate vectors output by the BilSTM layer;
and inputting the plurality of intermediate vectors into a CRF layer of a first preset identification model to generate a plurality of fine-grained entity labels.
5. The medical field intention recognition method of claim 3, wherein the invoking a first preset recognition model to perform entity recognition on the fine-grained entity labels according to a coarse granularity to obtain a plurality of coarse-grained entity labels comprises:
calling a first preset recognition model to recognize the fine-grained entity labels according to the coarse granularity to obtain a plurality of narrow-sense entity characteristics and a plurality of limited entity characteristics, wherein the narrow-sense entity characteristics comprise symptoms, diseases, parts, medicine, examination and treatment, and the limited entity characteristics comprise time, frequency, degree, negative words, description and numerical values;
combining the plurality of narrow-sense entity features and the plurality of limited entity features according to preset rules to generate a plurality of generalized entity features, wherein the generalized entity features comprise generalized symptoms, generalized inspection, generalized treatment and generalized drugs;
a plurality of generalized entity features is determined as a plurality of coarse-grained entity labels.
6. The method of claim 1, wherein the entity linking the coarse-grained entity labels according to a preset medical entity synonym table to obtain linked entity labels comprises:
searching a plurality of standard medical terms corresponding to a plurality of coarse-grained entity labels in a preset medical entity synonym table, wherein each coarse-grained entity label corresponds to one standard medical term, and the coarse-grained entity labels and the standard medical terms are synonyms;
fusing the plurality of coarse-grained entity labels to obtain a plurality of fused coarse-grained entity labels;
and carrying out entity linking operation on the fused coarse-grained entity labels and the standard medical terms to generate linked entity labels.
7. The medical field intention recognition method according to any one of claims 1 to 6, wherein the performing a knowledge graph query on a preset medical knowledge graph based on the knowledge graph query statement to obtain a knowledge graph query result, generating a corresponding target utterance according to the knowledge graph query result, and sending the target utterance to the terminal includes:
inquiring in a preset medical knowledge map based on the knowledge map inquiry statement to obtain a knowledge map inquiry result, wherein the knowledge map inquiry result comprises the relation of a target entity, the attribute of the target entity and a plurality of entities;
and generating a corresponding target language operation according to the relation of the target entity and the attribute of the target entity, and sending the target language operation to a terminal.
8. A medical field intention recognition apparatus, comprising:
the system comprises a statement acquisition module, a question generation module and a question generation module, wherein the statement acquisition module is used for acquiring an initial question statement from a terminal, and the initial question statement is a question statement input by a target user in a medical intelligent question-answering system;
the entity recognition module is used for calling a preset recognition model to perform entity recognition on the initial question statement to obtain an entity recognition result, and the entity recognition result comprises a plurality of coarse-grained entity labels and a plurality of entity relations;
the entity linking module is used for carrying out entity linking on the plurality of coarse-grained entity labels according to a preset medical entity synonym table to obtain linked entity labels;
the intention identification module is used for carrying out intention identification on the initial question sentence according to a preset intention identification model, the entity identification result and the linked entity label to obtain a candidate medical intention;
the statement generation module is used for generating a knowledge graph query statement according to the candidate medical intention;
and the map query module is used for querying a knowledge map in a preset medical knowledge map based on the knowledge map query statement to obtain a knowledge map query result, generating a corresponding target language according to the knowledge map query result and sending the target language to the terminal.
9. A medical field intention recognition device characterized by comprising: a memory having instructions stored therein and at least one processor, the memory and the at least one processor interconnected by a line;
the at least one processor invokes the instructions in the memory to cause the medical field intent recognition device to perform the medical field intent recognition method of any of claims 1-7.
10. A computer-readable storage medium, characterized in that the computer-readable storage medium stores instructions that, when executed by a processor, implement the medical field intent recognition method of any of claims 1-7.
CN202010884353.8A 2020-08-28 2020-08-28 Medical field intention recognition method, device, equipment and storage medium Pending CN112035635A (en)

Priority Applications (2)

Application Number Priority Date Filing Date Title
CN202010884353.8A CN112035635A (en) 2020-08-28 2020-08-28 Medical field intention recognition method, device, equipment and storage medium
PCT/CN2021/084659 WO2022041730A1 (en) 2020-08-28 2021-03-31 Medical field intention recognition method, apparatus and device, and storage medium

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202010884353.8A CN112035635A (en) 2020-08-28 2020-08-28 Medical field intention recognition method, device, equipment and storage medium

Publications (1)

Publication Number Publication Date
CN112035635A true CN112035635A (en) 2020-12-04

Family

ID=73586135

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202010884353.8A Pending CN112035635A (en) 2020-08-28 2020-08-28 Medical field intention recognition method, device, equipment and storage medium

Country Status (2)

Country Link
CN (1) CN112035635A (en)
WO (1) WO2022041730A1 (en)

Cited By (24)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112232059A (en) * 2020-12-14 2021-01-15 北京声智科技有限公司 Text error correction method and device, computer equipment and storage medium
CN112466463A (en) * 2020-12-10 2021-03-09 求臻医学科技（北京）有限公司 Intelligent answering system based on tumor accurate diagnosis and treatment knowledge graph
CN112925918A (en) * 2021-02-26 2021-06-08 华南理工大学 Question-answer matching system based on disease field knowledge graph
CN112966122A (en) * 2021-03-03 2021-06-15 平安科技（深圳）有限公司 Corpus intention identification method and device, storage medium and computer equipment
CN113157893A (en) * 2021-05-25 2021-07-23 网易（杭州）网络有限公司 Method, medium, apparatus, and computing device for intent recognition in multiple rounds of conversations
CN113282761A (en) * 2021-05-27 2021-08-20 平安科技（深圳）有限公司 Department information pushing method, device, equipment and storage medium
CN113327691A (en) * 2021-06-01 2021-08-31 平安科技（深圳）有限公司 Query method and device based on language model, computer equipment and storage medium
CN113345430A (en) * 2021-06-25 2021-09-03 上海适享文化传播有限公司 Multi-field query method based on voice fixed condition
CN113408274A (en) * 2021-07-13 2021-09-17 北京百度网讯科技有限公司 Method for training language model and label setting method
CN113468307A (en) * 2021-06-30 2021-10-01 网易（杭州）网络有限公司 Text processing method and device, electronic equipment and storage medium
CN113535919A (en) * 2021-07-16 2021-10-22 北京元年科技股份有限公司 Data query method and device, computer equipment and storage medium
CN113657102A (en) * 2021-08-17 2021-11-16 北京百度网讯科技有限公司 Information extraction method, information extraction device, information extraction apparatus, storage medium, and program
CN113688233A (en) * 2021-07-30 2021-11-23 达观数据(苏州)有限公司 Text understanding method for semantic search of knowledge graph
CN113707303A (en) * 2021-08-30 2021-11-26 康键信息技术（深圳）有限公司 Method, device, equipment and medium for solving medical problems based on knowledge graph
CN113793668A (en) * 2021-09-17 2021-12-14 平安科技（深圳）有限公司 Symptom standardization method and device based on artificial intelligence, electronic equipment and medium
WO2022041730A1 (en) * 2020-08-28 2022-03-03 康键信息技术（深圳）有限公司 Medical field intention recognition method, apparatus and device, and storage medium
CN114300128A (en) * 2021-12-31 2022-04-08 北京欧应信息技术有限公司 Medical concept linking system and storage medium for assisting intelligent diagnosis of disease
CN114464312A (en) * 2022-01-04 2022-05-10 北京欧应信息技术有限公司 System for assisting disease reasoning and storage medium
CN114722163A (en) * 2022-06-10 2022-07-08 科大讯飞股份有限公司 Data query method and device, electronic equipment and storage medium
CN114996412A (en) * 2022-08-02 2022-09-02 医智生命科技(天津)有限公司 Medical question-answering method and device, electronic equipment and storage medium
CN115062628A (en) * 2022-06-15 2022-09-16 北京信息科技大学 Automatic simulation method for doctor-patient communication conversation based on knowledge graph
CN116092493A (en) * 2023-04-07 2023-05-09 广州小鹏汽车科技有限公司 Voice interaction method, server and computer readable storage medium
CN116992861A (en) * 2023-09-25 2023-11-03 四川健康久远科技有限公司 Intelligent medical service processing method and system based on data processing
CN117235241A (en) * 2023-11-15 2023-12-15 安徽省立医院(中国科学技术大学附属第一医院) Man-machine interaction method oriented to hypertension inquiry follow-up scene

Families Citing this family (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112182235A (en) * 2020-08-29 2021-01-05 深圳呗佬智能有限公司 Method and device for constructing knowledge graph, computer equipment and storage medium
CN114911915A (en) * 2022-05-27 2022-08-16 重庆长安汽车股份有限公司 Knowledge graph-based question and answer searching method, system, equipment and medium
CN115630174B (en) * 2022-12-21 2023-07-21 上海金仕达软件科技股份有限公司 Multisource bulletin document processing method and device, storage medium and electronic equipment
CN116364296B (en) * 2023-02-17 2023-12-26 中国人民解放军总医院 Standard inspection project name confirmation method, device, equipment, medium and product
CN116108146B (en) * 2023-04-13 2023-06-27 天津数域智通科技有限公司 Information extraction method based on knowledge graph construction
CN116150406B (en) * 2023-04-23 2023-07-21 湖南星汉数智科技有限公司 Context sparse entity linking method, device, computer equipment and storage medium
CN116186359B (en) * 2023-05-04 2023-09-01 安徽宝信信息科技有限公司 Integrated management method, system and storage medium for multi-source heterogeneous data of universities
CN117056493A (en) * 2023-09-07 2023-11-14 四川大学 Large language model medical question-answering system based on medical record knowledge graph

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110110335A (en) * 2019-05-09 2019-08-09 南京大学 A kind of name entity recognition method based on Overlay model
CN110597970A (en) * 2019-08-19 2019-12-20 华东理工大学 Multi-granularity medical entity joint identification method and device
CN111522910A (en) * 2020-04-14 2020-08-11 浙江大学 Intelligent semantic retrieval method based on cultural relic knowledge graph

Family Cites Families (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US10248689B2 (en) * 2015-10-13 2019-04-02 International Business Machines Corporation Supplementing candidate answers
CN108959627B (en) * 2018-07-23 2021-12-17 北京光年无限科技有限公司 Question-answer interaction method and system based on intelligent robot
CN110659366A (en) * 2019-09-24 2020-01-07 Oppo广东移动通信有限公司 Semantic analysis method and device, electronic equipment and storage medium
CN112035635A (en) * 2020-08-28 2020-12-04 康键信息技术（深圳）有限公司 Medical field intention recognition method, device, equipment and storage medium

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110110335A (en) * 2019-05-09 2019-08-09 南京大学 A kind of name entity recognition method based on Overlay model
CN110597970A (en) * 2019-08-19 2019-12-20 华东理工大学 Multi-granularity medical entity joint identification method and device
CN111522910A (en) * 2020-04-14 2020-08-11 浙江大学 Intelligent semantic retrieval method based on cultural relic knowledge graph

Cited By (34)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2022041730A1 (en) * 2020-08-28 2022-03-03 康键信息技术（深圳）有限公司 Medical field intention recognition method, apparatus and device, and storage medium
CN112466463B (en) * 2020-12-10 2023-08-18 求臻医学科技(浙江)有限公司 Intelligent answering system based on tumor accurate diagnosis and treatment knowledge graph
CN112466463A (en) * 2020-12-10 2021-03-09 求臻医学科技（北京）有限公司 Intelligent answering system based on tumor accurate diagnosis and treatment knowledge graph
CN112232059A (en) * 2020-12-14 2021-01-15 北京声智科技有限公司 Text error correction method and device, computer equipment and storage medium
CN112925918A (en) * 2021-02-26 2021-06-08 华南理工大学 Question-answer matching system based on disease field knowledge graph
CN112966122A (en) * 2021-03-03 2021-06-15 平安科技（深圳）有限公司 Corpus intention identification method and device, storage medium and computer equipment
CN112966122B (en) * 2021-03-03 2024-05-10 平安科技（深圳）有限公司 Corpus intention recognition method and device, storage medium and computer equipment
CN113157893A (en) * 2021-05-25 2021-07-23 网易（杭州）网络有限公司 Method, medium, apparatus, and computing device for intent recognition in multiple rounds of conversations
CN113157893B (en) * 2021-05-25 2023-12-15 网易（杭州）网络有限公司 Method, medium, apparatus and computing device for intent recognition in multiple rounds of conversations
CN113282761A (en) * 2021-05-27 2021-08-20 平安科技（深圳）有限公司 Department information pushing method, device, equipment and storage medium
CN113327691B (en) * 2021-06-01 2022-08-12 平安科技（深圳）有限公司 Query method and device based on language model, computer equipment and storage medium
CN113327691A (en) * 2021-06-01 2021-08-31 平安科技（深圳）有限公司 Query method and device based on language model, computer equipment and storage medium
CN113345430A (en) * 2021-06-25 2021-09-03 上海适享文化传播有限公司 Multi-field query method based on voice fixed condition
CN113345430B (en) * 2021-06-25 2024-05-10 上海适享文化传播有限公司 Multi-field query method based on voice fixed condition
CN113468307A (en) * 2021-06-30 2021-10-01 网易（杭州）网络有限公司 Text processing method and device, electronic equipment and storage medium
CN113468307B (en) * 2021-06-30 2023-06-30 网易（杭州）网络有限公司 Text processing method, device, electronic equipment and storage medium
CN113408274B (en) * 2021-07-13 2022-06-24 北京百度网讯科技有限公司 Method for training language model and label setting method
CN113408274A (en) * 2021-07-13 2021-09-17 北京百度网讯科技有限公司 Method for training language model and label setting method
CN113535919A (en) * 2021-07-16 2021-10-22 北京元年科技股份有限公司 Data query method and device, computer equipment and storage medium
CN113688233A (en) * 2021-07-30 2021-11-23 达观数据(苏州)有限公司 Text understanding method for semantic search of knowledge graph
CN113657102A (en) * 2021-08-17 2021-11-16 北京百度网讯科技有限公司 Information extraction method, information extraction device, information extraction apparatus, storage medium, and program
CN113707303A (en) * 2021-08-30 2021-11-26 康键信息技术（深圳）有限公司 Method, device, equipment and medium for solving medical problems based on knowledge graph
WO2023029512A1 (en) * 2021-08-30 2023-03-09 康键信息技术（深圳）有限公司 Knowledge graph-based medical question answering method and apparatus, device and medium
CN113793668A (en) * 2021-09-17 2021-12-14 平安科技（深圳）有限公司 Symptom standardization method and device based on artificial intelligence, electronic equipment and medium
CN114300128A (en) * 2021-12-31 2022-04-08 北京欧应信息技术有限公司 Medical concept linking system and storage medium for assisting intelligent diagnosis of disease
CN114464312A (en) * 2022-01-04 2022-05-10 北京欧应信息技术有限公司 System for assisting disease reasoning and storage medium
CN114722163A (en) * 2022-06-10 2022-07-08 科大讯飞股份有限公司 Data query method and device, electronic equipment and storage medium
CN115062628A (en) * 2022-06-15 2022-09-16 北京信息科技大学 Automatic simulation method for doctor-patient communication conversation based on knowledge graph
CN114996412A (en) * 2022-08-02 2022-09-02 医智生命科技(天津)有限公司 Medical question-answering method and device, electronic equipment and storage medium
CN116092493A (en) * 2023-04-07 2023-05-09 广州小鹏汽车科技有限公司 Voice interaction method, server and computer readable storage medium
CN116092493B (en) * 2023-04-07 2023-08-25 广州小鹏汽车科技有限公司 Voice interaction method, server and computer readable storage medium
CN116992861A (en) * 2023-09-25 2023-11-03 四川健康久远科技有限公司 Intelligent medical service processing method and system based on data processing
CN116992861B (en) * 2023-09-25 2023-12-08 四川健康久远科技有限公司 Intelligent medical service processing method and system based on data processing
CN117235241A (en) * 2023-11-15 2023-12-15 安徽省立医院(中国科学技术大学附属第一医院) Man-machine interaction method oriented to hypertension inquiry follow-up scene

Also Published As

Publication number Publication date
WO2022041730A1 (en) 2022-03-03

Similar Documents

Publication Publication Date Title
CN112035635A (en) Medical field intention recognition method, device, equipment and storage medium
CN112035637A (en) Medical field intention recognition method, device, equipment and storage medium
CN111708874B (en) Man-machine interaction question-answering method and system based on intelligent complex intention recognition
WO2021139232A1 (en) Medical knowledge graph-based triage method and apparatus, device, and storage medium
CN111538894B (en) Query feedback method and device, computer equipment and storage medium
CN107247868B (en) Artificial intelligence auxiliary inquiry system
CN112131393B (en) Medical knowledge graph question-answering system construction method based on BERT and similarity algorithm
CN111813957A (en) Medical diagnosis guiding method based on knowledge graph and readable storage medium
Alicante et al. Unsupervised entity and relation extraction from clinical records in Italian
CN110489566A (en) A kind of hospital guide's method of intelligence hospital guide's service robot
CN110675944A (en) Triage method and device, computer equipment and medium
CN109299239A (en) ES-based electronic medical record retrieval method
CN109545373A (en) A kind of automatic abstracting method of human body diseases symptom characteristic, system and equipment
CN106991284A (en) Intelligent child-rearing knowledge services method and system
Friedman et al. Natural language and text processing in biomedicine
CN113505243A (en) Intelligent question-answering method and device based on medical knowledge graph
US20090313243A1 (en) Method and apparatus for processing semantic data resources
CN111723570A (en) Medicine knowledge graph construction method and device and computer equipment
CN113569023A (en) Chinese medicine question-answering system and method based on knowledge graph
CN114676233A (en) Medical automatic question-answering method based on skeletal muscle knowledge graph
CN114676260A (en) Human body bone motion rehabilitation model construction method based on knowledge graph
CN115248842A (en) ICD intelligent coding system based on knowledge graph and retrieval engine
CN113064960A (en) Method for accurately searching cases similar to patient's condition
CN117171329A (en) Semantic analysis-based traditional Chinese medicine domain knowledge graph question-answering method
Montenegro et al. The hope model architecture: a novel approach to pregnancy information retrieval based on conversational agents

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
RJ01 Rejection of invention patent application after publication

Application publication date: 20201204

RJ01 Rejection of invention patent application after publication