CN112836514A - Nested entity recognition method and device, electronic equipment and storage medium - Google Patents
Nested entity recognition method and device, electronic equipment and storage medium Download PDFInfo
- Publication number
- CN112836514A CN112836514A CN202110343229.5A CN202110343229A CN112836514A CN 112836514 A CN112836514 A CN 112836514A CN 202110343229 A CN202110343229 A CN 202110343229A CN 112836514 A CN112836514 A CN 112836514A
- Authority
- CN
- China
- Prior art keywords
- entity recognition
- nested
- recognition result
- text
- entity
- 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.)
- Granted
Links
- 238000000034 method Methods 0.000 title claims abstract description 47
- 238000012549 training Methods 0.000 claims abstract description 16
- 238000012937 correction Methods 0.000 claims description 72
- 230000015654 memory Effects 0.000 claims description 11
- 238000004590 computer program Methods 0.000 claims description 6
- 238000004891 communication Methods 0.000 description 8
- 238000002372 labelling Methods 0.000 description 7
- 230000011218 segmentation Effects 0.000 description 7
- 238000010586 diagram Methods 0.000 description 6
- 230000006870 function Effects 0.000 description 4
- 230000008569 process Effects 0.000 description 4
- 238000003058 natural language processing Methods 0.000 description 3
- 230000009466 transformation Effects 0.000 description 3
- 230000009286 beneficial effect Effects 0.000 description 2
- 230000002457 bidirectional effect Effects 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 238000000605 extraction Methods 0.000 description 2
- 230000007787 long-term memory Effects 0.000 description 2
- 239000011159 matrix material Substances 0.000 description 2
- 230000007246 mechanism Effects 0.000 description 2
- 238000012015 optical character recognition Methods 0.000 description 2
- 230000003287 optical effect Effects 0.000 description 2
- 238000010276 construction Methods 0.000 description 1
- 239000000284 extract Substances 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000000717 retained effect Effects 0.000 description 1
- 238000010845 search algorithm Methods 0.000 description 1
- 230000006403 short-term memory Effects 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
Images
Classifications
-
- 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
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/22—Matching criteria, e.g. proximity measures
-
- 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
- G06F40/242—Dictionaries
-
- 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
- G06F40/30—Semantic analysis
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Artificial Intelligence (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Health & Medical Sciences (AREA)
- Computational Linguistics (AREA)
- Audiology, Speech & Language Pathology (AREA)
- Health & Medical Sciences (AREA)
- Data Mining & Analysis (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Evolutionary Computation (AREA)
- Evolutionary Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Computational Biology (AREA)
- Life Sciences & Earth Sciences (AREA)
- Character Discrimination (AREA)
Abstract
The embodiment of the invention provides a nested entity identification method, a nested entity identification device, electronic equipment and a storage medium, wherein the method comprises the following steps: determining a text to be recognized; inputting a text to be recognized into the nested entity recognition model to obtain a simple entity recognition result and a nested entity recognition result which are output by the nested entity recognition model; the nested entity recognition model is obtained by training based on a sample text, a sample simple entity recognition result and a sample nested entity recognition result which correspond to the sample text; the nested entity recognition model is used for determining a simple entity recognition result based on the text to be recognized and determining a nested entity recognition result based on the text to be recognized and the simple entity recognition result. According to the nested entity identification method, the nested entity identification device, the electronic equipment and the storage medium, identification of the nested entity and the simple entities in the nested entity can be realized only by one model, and the nested relation between the nested entity and the simple entities in the nested entity is shown.
Description
Technical Field
The invention relates to the technical field of natural language processing, in particular to a nested entity identification method, a nested entity identification device, electronic equipment and a storage medium.
Background
Entity identification is an important step in the natural language processing process, and is widely applied to tasks such as information extraction, information retrieval, information recommendation and the like. Because of the diversity of natural languages, nested entities exist in a large amount of text. Here, the nested entity refers to a case where a single entity is formed as a whole and a plurality of simple entities are included therein. Therefore, in order to correctly identify a nested entity, a nested relationship between the nested entity in the text and a simple entity inside the nested entity needs to be identified.
However, in the prior art, entity identification is usually performed by using a sequence labeling model, and the existing sequence labeling model can only output one sequence labeling result, and cannot identify both nested entities and simple entities inside the nested entities.
Disclosure of Invention
The embodiment of the invention provides a nested entity identification method, a nested entity identification device, electronic equipment and a storage medium, which are used for solving the problem that the existing sequence labeling method cannot identify both nested entities and simple entities inside the nested entities.
In a first aspect, an embodiment of the present invention provides a method for identifying a nested entity, including:
determining a text to be recognized;
inputting the text to be recognized into a nested entity recognition model to obtain a simple entity recognition result and a nested entity recognition result which are output by the nested entity recognition model;
the nested entity recognition model is obtained by training based on a sample text, and a sample simple entity recognition result and a sample nested entity recognition result which correspond to the sample text;
the nested entity recognition model is used for determining the simple entity recognition result based on the text to be recognized and determining the nested entity recognition result based on the text to be recognized and the simple entity recognition result.
Optionally, the inputting the text to be recognized into a nested entity recognition model to obtain a simple entity recognition result and a nested entity recognition result output by the nested entity recognition model specifically includes:
inputting the first text vector of the text to be recognized to a simple entity recognition layer of the nested entity recognition model to obtain the simple entity recognition result output by the simple entity recognition layer;
inputting the simple entity recognition result to an attention layer of the nested entity recognition model to obtain a simple entity attention vector output by the attention layer;
and inputting the simple entity attention vector and the second text vector of the text to be recognized into a nested entity recognition layer of the nested entity recognition model to obtain the nested entity recognition result output by the nested entity recognition layer.
Optionally, the first text vector includes a word vector of each word in the text to be recognized, and a dictionary feature vector and/or a part-of-speech feature vector of each word in the text to be recognized.
Optionally, the dictionary feature vector of each word in the text to be recognized is obtained by matching the text to be recognized with a pre-constructed domain dictionary;
wherein the dictionary feature vector for any word represents the type of entity that the any word matches in the domain dictionary and the location of the any word in the matching entity.
Optionally, the inputting the simple entity attention vector and the text vector of the text to be recognized into a nested entity recognition layer of the nested entity recognition model to obtain the nested entity recognition result output by the nested entity recognition layer, and then further comprising:
and inputting the simple entity recognition result and/or the nested entity recognition result into a result correction layer of the nested entity recognition model to obtain a corrected simple entity recognition result and/or a nested entity recognition result output by the result correction layer.
Optionally, the inputting the simple entity recognition result or the nested entity recognition result into an outcome correction layer of the nested entity recognition model to obtain a corrected simple entity recognition result or a nested entity recognition result output by the outcome correction layer specifically includes:
and inputting the simple entity recognition result or the nested entity recognition result into the result correction layer, and performing result correction on the simple entity recognition result or the nested entity recognition result by the result correction layer based on an entity label rule to obtain a corrected simple entity recognition result or a corrected nested entity recognition result output by the result correction layer.
Optionally, the inputting the simple entity recognition result and the nested entity recognition result into an outcome correction layer of the nested entity recognition model to obtain a corrected simple entity recognition result and a nested entity recognition result output by the outcome correction layer specifically includes:
inputting the simple entity recognition result and the nested entity recognition result into the result correction layer, and when judging that the simple entity recognition result and the nested entity recognition result conflict, the result correction layer corrects the simple entity recognition result or the nested entity recognition result to obtain a corrected simple entity recognition result and a nested entity recognition result output by the result correction layer;
the condition that the simple entity recognition result conflicts with the nested entity recognition result comprises that at least one of repeated entities, entities with the same boundary but different entity types and entities with crossed boundaries exists between the simple entity recognition result and the nested entity recognition result.
In a second aspect, an embodiment of the present invention provides a nested entity identifying apparatus, including:
the text determining unit is used for determining a text to be recognized;
the nested recognition unit is used for inputting the text to be recognized into a nested entity recognition model to obtain a simple entity recognition result and a nested entity recognition result which are output by the nested entity recognition model;
the nested entity recognition model is obtained by training based on a sample text, and a sample simple entity recognition result and a sample nested entity recognition result which correspond to the sample text;
the nested entity recognition model is used for determining the simple entity recognition result based on the text to be recognized and determining the nested entity recognition result based on the text to be recognized and the simple entity recognition result.
In a third aspect, an embodiment of the present invention provides an electronic device, including a processor, a communication interface, a memory, and a bus, where the processor and the communication interface, the memory complete mutual communication through the bus, and the processor may call a logic command in the memory to perform the steps of the method provided in the first aspect.
In a fourth aspect, an embodiment of the present invention provides a non-transitory computer readable storage medium, on which a computer program is stored, which when executed by a processor, implements the steps of the method as provided in the first aspect.
According to the nested entity identification method, the nested entity identification device, the electronic equipment and the storage medium, the simple entity identification result is determined based on the text to be identified, the nested entity identification result is determined based on the text to be identified and the simple entity identification result, and the identification of the nested entity and the simple entity in the nested entity can be realized only by one model; and the obtained simple entity identification result and entity boundary information in the nested entity identification result show the nested relation between the nested entity and the simple entity inside the nested entity, and provide a better support function for the subsequent text parsing task.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described below, and it is obvious that the drawings in the following description are some embodiments of the present invention, and those skilled in the art can also obtain other drawings according to the drawings without creative efforts.
Fig. 1 is a schematic flowchart of a nested entity identification method according to an embodiment of the present invention;
fig. 2 is a schematic flowchart of a method for operating a nested entity recognition model according to an embodiment of the present invention;
FIG. 3 is a schematic structural diagram of a nested entity recognition model according to an embodiment of the present invention;
fig. 4 is a schematic structural diagram of a nested entity identifying apparatus according to an embodiment of the present invention;
fig. 5 is a schematic structural diagram of an electronic device according to an embodiment of the present invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are some, but not all, embodiments of the present invention. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
Entity identification is an important step in the natural language processing process, and is widely applied to tasks such as information extraction, information retrieval, information recommendation and the like. Due to the diversity of natural languages, nested entities may exist in the text to be recognized. Here, a nested entity refers to an entity in which several simple entities are nested, i.e., an entity that does not contain other entities inside. For example, for the text to be recognized in the building field, "house with height greater than 16m must be provided with elevator", wherein "house with height greater than 16 m" is a nested entity with type "Object", and the nested entity contains several simple entities inside, namely, entity "height" with type "Attribute", entity "greater than 16 m" with type "Attribute value", and entity "house" with type "Object". Therefore, when entity recognition is performed, the nested relation between the nested entity in the text and the simple entity inside the nested entity needs to be recognized.
Currently, entity identification is usually performed by using a sequence labeling method, for example, each word in a text to be identified is labeled by using a labeling method such as BIO or biees. In the biees notation, b (begin) represents the beginning character of an entity, i (intermediate) represents the middle character of an entity, e (end) represents the ending character of an entity, s (single) represents a single character, and o (other) represents others. However, the above method can only output a sequence marking result, and cannot identify the nested relation between the nested entity and the simple entity inside the nested entity, nor know whether the nested entity exists in the text to be identified.
Therefore, the embodiment of the invention provides a nested entity identification method. Fig. 1 is a schematic flowchart of a nested entity identification method according to an embodiment of the present invention, and as shown in fig. 1, the method includes:
Here, the text to be recognized is a text that needs to be recognized by the nested entity, and the text to be recognized may be an electronic text, or may be obtained by performing Character Recognition on an image of a paper text by applying a Character Recognition technology such as OCR (Optical Character Recognition), which is not specifically limited in this embodiment of the present invention.
the nested entity recognition model is obtained by training based on a sample text, a sample simple entity recognition result and a sample nested entity recognition result which correspond to the sample text;
the nested entity recognition model is used for determining a simple entity recognition result based on the text to be recognized and determining a nested entity recognition result based on the text to be recognized and the simple entity recognition result.
Specifically, after the text to be recognized is input into the nested entity recognition model, the nested entity recognition model extracts semantic information of the text to be recognized, and performs fine-grained simple entity recognition based on the semantic information of the text to be recognized to obtain a simple entity recognition result. The simple entity recognition result includes entity boundaries of all simple entities in the text to be recognized, and may also include entity types of all simple entities. Alternatively, the simple entity recognition result may be a sequence of entity tags corresponding to all simple entities contained in the text to be recognized. For example, for the text to be recognized "a house with a height greater than 16m must be provided with an elevator", the simple entity recognition result may be [ B-attr, E-attr, B-attrValue, I-attrValue, I-attrValue, E-attrValue, O, B-object, E-object, O, O, B-operator, E-operator, B-object, E-object ], where attr, attrValue, object, and operator are entity types, and B, I, E, O represents the start character, middle character, end character, and other characters of the entity.
After the simple entity recognition result is obtained, the nested entity recognition model carries out coarse-grained nested entity recognition on the basis of the text to be recognized and the simple entity recognition result to obtain a nested entity recognition result. The nested entity recognition result includes entity boundaries of the nested entities in the text to be recognized, and may also include entity types of the nested entities, and may also include entity boundaries and entity types of the remaining simple entities that cannot be aggregated in the text to be recognized. Alternatively, the nested entity recognition result may be an entity tag sequence corresponding to the nested entity and the simple entity contained in the text to be recognized. For example, for the text to be recognized that "a house with a height greater than 16m must be provided with an elevator", the nested entity recognition result may be [ B-object, I-object, I-object, I-object, I-object, I-object, E-object, O, O, B-operation, E-operation, B-object, E-object ]. When coarse-grained nested entity recognition is carried out, on the basis of the semantic information of the text to be recognized, the simple entity recognition result can bring more semantic information, such as the entity type and the entity boundary of each simple entity in the simple entity recognition result, and can help to determine the nested entities and the entity types and the entity boundaries of the simple entities, so that the accuracy of nested entity recognition is improved. It should be noted that when the simple entity identification layer and the nested entity identification layer mark the entity type of the text to be identified, the adopted entity type labels are the same, so that a set of entity type labels do not need to be specially designed for the nested entity identification layer.
After the simple entity recognition result and the nested entity recognition result are determined, entity boundary information in the simple entity recognition result and the nested entity recognition result can show the nested relation between the nested entity and the simple entity inside the nested entity. For example, the text to be recognized "a house having a height of more than 16m must be provided with an elevator", which corresponds to a simple entity recognition result of [ 2 ]B- attr，E-attr，B-attrValue，I-attrValue，I-attrValue，I-attrValue，E-atrValue，O，B- object，E-object，O，O，B-operate，E-operate，B-object，E-object]The result of the identification of the nested entity is [, ] [, ]B-object，I-object，I-object，I-object，I-object，I-object，I-object，I-object，I- object，E-object，O，O，B-operate，E-operate，B-object，E-object]. According to the simple entity recognition result and the entity boundary information of the underlined part in the nested entity recognition result, it can be seen that the nested entity 'house with the height larger than 16 m' has the nested relation with the simple entity 'height', 'house larger than 16 m' and 'house'.
Before step 120 is executed, the nested entity recognition model may be obtained through pre-training, and specifically, the nested entity recognition model may be obtained through training in the following manner: firstly, a large amount of sample texts are collected, and a sample simple entity recognition result and a sample nested entity recognition result corresponding to the sample texts are determined in a manual labeling mode. And training the initial model based on the sample text, and the sample simple entity recognition result and the sample nested entity recognition result corresponding to the sample text, so as to obtain the nested entity recognition model.
According to the method provided by the embodiment of the invention, the nested entity recognition model determines the simple entity recognition result based on the text to be recognized, and determines the nested entity recognition result based on the text to be recognized and the simple entity recognition result, and the recognition of the nested entity and the simple entity in the nested entity can be realized only by one model; and the obtained simple entity identification result and entity boundary information in the nested entity identification result show the nested relation between the nested entity and the simple entity inside the nested entity, and provide a better support function for the subsequent text parsing task.
Based on the foregoing embodiment, fig. 2 is a schematic flowchart of a method for operating a nested entity recognition model according to an embodiment of the present invention, and as shown in fig. 2, step 120 specifically includes:
and step 121, inputting the first text vector of the text to be recognized to a simple entity recognition layer of the nested entity recognition model to obtain a simple entity recognition result output by the simple entity recognition layer.
Specifically, a first text vector of the text to be recognized is used for representing semantic features of the text to be recognized. Alternatively, the first text vector of the text to be recognized may contain semantic information for each word in the text to be recognized. The simple entity recognition layer is used for recognizing all simple entities in the text to be recognized based on the first text vector of the text to be recognized, and obtaining a simple entity recognition result.
Alternatively, the simple entity recognition layer may be the structure of a bidirectional long and short term memory network BilSTM + conditional random field CRF. The Bi-LSTM can be used for coding a first text vector of a text to be recognized to obtain a context semantic vector of the text to be recognized. The CRF may determine the probability that each word in the text to be recognized corresponds to each entity tag based on the context semantic vector of the text to be recognized, and calculate a sequence of entity tags with the largest probability and using a dynamic programming method, such as the viterbi algorithm, and output the sequence as a simple entity recognition result.
And step 122, inputting the simple entity recognition result to an attention layer of the nested entity recognition model to obtain a simple entity attention vector output by the attention layer.
Specifically, the attention layer is used for performing self-attention transformation on the simple entity identification result based on the attention weight matrix to obtain a simple entity attention vector. Here, the self-attention transformation can dig out simple entities which are closely related in the simple entity recognition result and can be further aggregated to form new entities, so that the simple entities which can be aggregated to form nested entities in the simple entity recognition result are highlighted, the simple entities which cannot be further aggregated in the simple entity recognition result are weakened, and the accuracy of the nested entity recognition result is improved. Wherein, the attention weight matrix can be obtained by learning in the training process of the nested entity recognition model.
And step 123, inputting the simple entity attention vector and the second text vector of the text to be recognized into a nested entity recognition layer of the nested entity recognition model, and obtaining a nested entity recognition result output by the nested entity recognition layer.
Specifically, the second text vector of the text to be recognized is used for representing semantic features of the text to be recognized. It should be noted that the second text vector may be the same as the first text vector, i.e., the simple entity recognition layer and the nested entity recognition layer share one input, or the second text vector may be different from the first text vector. Alternatively, the second text vector may be a word vector for each word in the text to be recognized.
The nested entity recognition layer is used for carrying out coarse-grained nested entity recognition based on the simple entity attention vector and a second text vector of the text to be recognized. Alternatively, the structure of the nested entity recognition layer may be the same as that of the simple entity recognition layer, for example, the nested entity recognition layer may also be the structure of a bidirectional long-term memory network BiLSTM + conditional random field CRF.
The method provided by the embodiment of the invention is based on the self-attention mechanism, the simple entity recognition result is converted into the simple entity attention vector, and the nested entity recognition result is determined based on the simple entity attention vector and the second text vector of the text to be recognized, so that the accuracy of the nested entity recognition result is improved.
According to any of the above embodiments, in the method, the first text vector includes a word vector of each word in the text to be recognized, and a dictionary feature vector and/or a part-of-speech feature vector of each word in the text to be recognized.
Specifically, the first text vector of the text to be recognized includes a word vector of each word in the text to be recognized, and may include a dictionary feature vector and/or a part-of-speech feature vector of each word in addition to the word vector. Here, the Word vector of any Word may be determined based on a pre-trained Word vector model, such as a Word2vec model, a Bert model, etc.; the dictionary feature vector of any word is used for representing the semantic features of the corresponding entry of the word in the preset dictionary, and the part of speech feature vector of any word is used for representing the semantic features of the part of speech. The dictionary feature vector and the part-of-speech feature vector can be obtained by learning in the training process of the nested entity recognition model after random initialization.
The first text vector comprises a word vector of each word and also comprises a dictionary characteristic vector and/or a part-of-speech characteristic vector of each word, namely, on the basis of semantic information of each word, semantic information of a vocabulary entry corresponding to each word and/or part-of-speech information of each word are additionally introduced, so that the semantic information of the text to be recognized is enriched, and the accuracy of a simple entity recognition result is improved. For example, based on the dictionary feature vector of each word, it can be known which adjacent words in the text to be recognized form a specific word in the preset dictionary, and the adjacent words forming the specific word are more likely to form a simple entity without splitting the simple entity into two or more entities. For another example, based on the part of speech of each word, since a noun is more likely to be an entity, it is more likely to identify a word adjacent to the part of speech as a noun as a simple entity.
According to the method provided by the embodiment of the invention, the first text vector comprises the word vector of each word in the text to be recognized and the dictionary characteristic vector and/or part-of-speech characteristic vector of each word in the text to be recognized, so that the accuracy of the recognition result of the simple entity is improved.
Based on any one of the embodiments, in the method, the dictionary feature vector of each word in the text to be recognized is obtained by matching the text to be recognized with a pre-constructed domain dictionary; wherein the dictionary feature vector for any word represents the type of entity that the word matches in the domain dictionary and the location of the word in the matching entity.
Specifically, a domain dictionary corresponding to a domain needs to be constructed by collecting professional terms and core words of the domain as potential entities in advance based on texts in the domain related to the text to be recognized. The potential entities are terms which are possibly recognized as entities, and the domain dictionary comprises each potential entity and the corresponding entity type. For example, for the architectural domain, a domain dictionary can be constructed as shown in the following table:
| latent entity | Entity type |
| Living room (Hall) | object |
| Residential building | object |
| Area of building | attr |
| Fire resistance rating | attr |
| Is provided with | operate |
| Are connected with each other | operate |
| …… | …… |
And matching the text to be recognized with the domain dictionary to obtain the potential entity matched with the text to be recognized and the entity type of the potential entity. For example, for the text to be recognized, "residential building with four fire ratings", the potential entities and their entity types that the text to be recognized matches are: [ (four, attrValue), (fire rating, attr), (residential building, object) ]. Optionally, the matching algorithm of the text to be recognized and the domain dictionary may adopt a maximum backward matching algorithm or a maximum forward matching algorithm, which is not specifically limited in the embodiment of the present invention. Based on the potential entities matched with the text to be recognized and the entity types thereof, the dictionary feature vector of each word in the text to be recognized can be obtained, wherein the dictionary feature vector of any word can represent the entity type of the potential entity corresponding to the word and the position of the word in the potential entity. Based on the dictionary feature vector of each word, when the simple entity recognition is carried out, which adjacent words in the text to be recognized correspond to the same potential entity in the domain dictionary can be obtained, and the adjacent words corresponding to the same potential entity are preferentially recognized as a simple entity, so that the accuracy of the recognition result of the simple entity is improved.
The method provided by the embodiment of the invention matches the text to be recognized with the pre-constructed domain dictionary to obtain the dictionary feature vector of each word in the text to be recognized, and is beneficial to improving the accuracy of the simple entity recognition result.
Based on any embodiment, the text to be recognized can be subjected to word segmentation based on a word segmentation technology, such as jieba word segmentation, and then word segmentation correction is performed by combining a domain dictionary. For example, "living room (living room) should not be smaller than 10m in use area2", the result of the word segmentation is [ ' living room ', ' (', ' living room ', ') ', ' use ', ' area ', ' not ', ' should ', ' less than ', ' 10m2’]. In combination with potential entities such as "living room (living room)", "use area" in the domain dictionary, the word segmentation result can be corrected as follows: [ ' living room (living room) ', ' use area ', ' not ', ' should ', ' less ', ' 10m2’]. And then, performing part-of-speech analysis on the corrected word segmentation result to obtain the part-of-speech of each word in the text to be recognized, so as to determine the part-of-speech characteristic vector of each word in the text to be recognized. The part-of-speech feature vector of any word can represent the part of speech of the word and the position of the word in the word.
Based on any of the above embodiments, after step 123, the method further includes:
and inputting the simple entity recognition result and/or the nested entity recognition result into a result correction layer of the nested entity recognition model to obtain a corrected simple entity recognition result and/or a nested entity recognition result output by the result correction layer.
Specifically, in order to further improve the accuracy of the simple entity recognition result and/or the nested entity recognition result, the simple entity recognition result and/or the nested entity recognition result may be input to an outcome correction layer of the nested entity recognition model, so as to obtain a corrected simple entity recognition result and/or a nested entity recognition result output by the outcome correction layer.
Further, the simple entity recognition result or the nested entity recognition result can be only input into the result correction layer, so that the result correction layer can independently analyze the logic problem existing in the simple entity recognition result or the nested entity recognition result, and the result correction can be carried out on the simple entity recognition result or the nested entity recognition result; the simple entity recognition result and the nested entity recognition result can be simultaneously input into the result correction layer, so that the result correction layer can compare and analyze the conflict between the simple entity recognition result and the nested entity recognition result, and then the result correction can be carried out on the simple entity recognition result or the nested entity recognition result, and the corrected simple entity recognition result and the nested entity recognition result can be obtained.
The method provided by the embodiment of the invention corrects the simple entity recognition result and/or the nested entity recognition result based on the simple entity recognition result and/or the nested entity recognition result, and improves the accuracy of the simple entity recognition result and/or the nested entity recognition result.
Based on any of the above embodiments, inputting the simple entity recognition result or the nested entity recognition result to the result correction layer of the nested entity recognition model to obtain a corrected simple entity recognition result or a nested entity recognition result output by the result correction layer, specifically including:
and inputting the simple entity recognition result or the nested entity recognition result into the result correction layer, and performing result correction on the simple entity recognition result or the nested entity recognition result by the result correction layer based on the entity label rule to obtain a corrected simple entity recognition result or a corrected nested entity recognition result output by the result correction layer.
Specifically, the entity tag rule is a rule to be followed when the entity tag is used for sequence tagging, and the entity tag rule may be obtained in advance according to an actual application scenario. For example, toIn the biees tag, two tags B or two tags E should not appear in succession when sequence annotation is performed, and tags B and I should not appear at the end of an entity, e.g. "B-object, I-object,B-object，B-attr，E-attr”、“B-object，I-objectb-attr, E-attr "is not allowed, while tags E and I should not appear at the beginning of an entity, e.g." B-attr, E-attr,E- object，I-object，E-object”、“B-attr，E-attr，I-objecte-object "is also not allowed.
Therefore, whether the simple entity identification result or the nested entity identification result conforms to the entity label rule can be judged based on the entity label rule. If the simple entity identification result or the nested entity identification result does not meet the entity label rule, the result of the entity label rule needs to be corrected. Optionally, if the simple entity recognition result or the nested entity recognition result does not satisfy the entity tag rule, based on the probability that each word in the text to be recognized corresponds to each entity tag, an entity tag sequence which has the highest score and satisfies the entity tag rule is obtained by using an N-best search algorithm and is used as the corrected simple entity recognition result or the nested entity recognition result.
The method provided by the embodiment of the invention is used for correcting the result of the simple entity identification result or the nested entity identification result based on the entity label rule, so that the accuracy of the simple entity identification result or the nested entity identification result is improved.
Based on any of the above embodiments, inputting the simple entity recognition result and the nested entity recognition result to the result correction layer of the nested entity recognition model to obtain the corrected simple entity recognition result and the nested entity recognition result output by the result correction layer, specifically including:
inputting the simple entity recognition result and the nested entity recognition result into a result correction layer, and when the result correction layer judges that the simple entity recognition result and the nested entity recognition result conflict, performing result correction on the simple entity recognition result or the nested entity recognition result to obtain a corrected simple entity recognition result and a nested entity recognition result output by the result correction layer;
the condition that the simple entity recognition result conflicts with the nested entity recognition result comprises that at least one of repeated entities, entities with the same boundary but different entity types and entities with crossed boundaries exists between the simple entity recognition result and the nested entity recognition result.
Specifically, after the simple entity recognition result and the nested entity recognition result are input to the result correction layer, the result correction layer needs to compare the simple entity recognition result and the nested entity recognition result to determine whether a conflict exists between the simple entity recognition result and the nested entity recognition result. Here, the case where the simple entity recognition result and the nested entity recognition result conflict includes at least one of an entity having a duplicate entity, an entity having the same boundary but a different entity type, and an entity having a boundary crossing therebetween. For example, for a house with a text to be recognized "height greater than 16m, an elevator must be provided, and the use area of the living room should not be less than 10m2If the simple entity recognition result and the nested entity recognition result both comprise an entity elevator, the fact that a repeated entity exists between the simple entity recognition result and the nested entity recognition result is indicated; if the simple entity recognition result and the nested entity recognition result both comprise entity 'settings', but the type of the entity 'settings' in the simple entity recognition result is 'object', and the type of the entity 'settings' in the nested entity recognition result is 'operator', it indicates that an entity with the same boundary but different entity types exists between the simple entity recognition result and the nested entity recognition result; if the simple entity recognition result comprises an entity 'should' and the nested entity recognition result comprises an entity 'should' smaller, it indicates that the two entities have boundary crossing.
And if the result correction layer judges that the simple entity recognition result and the nested entity recognition result conflict, the result correction layer needs to correct the result, so that the corrected simple entity recognition result and the nested entity recognition result are obtained. It should be noted that, when performing result correction, a simple entity recognition result may be selected for correction, and a nested entity recognition result may also be selected for correction, which is not specifically limited in the embodiment of the present invention. Alternatively, if there is a conflict between the simple entity recognition result and the nested entity recognition result, only one of the two entities in which the conflict exists may be retained.
According to the method provided by the embodiment of the invention, when the conflict between the simple entity identification result and the nested entity identification result is judged and obtained, the result correction is carried out on the simple entity identification result or the nested entity identification result, so that the accuracy of the simple entity identification result and the nested entity identification result is improved.
Based on any embodiment, the construction method of the nested entity recognition model comprises the following steps:
firstly, a large number of sample texts are collected, and a sample simple entity recognition result and a sample nested entity recognition result corresponding to each sample text are determined.
The structure of the nested entity recognition model is then determined. Fig. 3 is a schematic structural diagram of a nested entity recognition model according to an embodiment of the present invention, and as shown in fig. 3, the nested entity recognition model includes a simple entity recognition layer, an attention layer, and a nested entity recognition layer. The simple entity identification layer and the nested entity identification layer have the same structure and are both the structure of BiLSTM + CRF.
The simple entity recognition layer is used for determining a simple entity recognition result based on a first text vector of a text to be recognized; the first text vector of the text to be recognized is formed by splicing a word vector, a dictionary characteristic vector and a part of speech characteristic vector of each word in the text to be recognized. And the attention layer is used for carrying out self-attention transformation on the simple entity recognition result to obtain a simple entity attention vector. The nested entity recognition layer is used for determining a nested entity recognition result based on the simple entity attention vector and a second text vector of the text to be recognized; and the second text vector of the text to be recognized is a word vector of each word in the text to be recognized. Optionally, the simple entity attention vector and the second text vector of the text to be recognized may be spliced and input to the nested entity recognition layer, so as to determine the nested entity recognition result.
And then training the parameters of the nested entity recognition model based on the sample text and the sample simple entity recognition result and the sample nested entity recognition result corresponding to the sample text.
Wherein, the loss function of the nested entity recognition model can be expressed as:
Loss＝α·Loss1+(1-α)·Loss2
wherein, Loss is the Loss of the nested entity recognition model, alpha belongs to (0,1) as a hyper-parameter, and Loss1And Loss2Loss of the simple entity recognition layer and the nested entity recognition layer respectively,
and
scores of the sample simple entity recognition result and the sample nested entity recognition result are respectively obtained,
the score for each possible entity tag sequence calculated for the simple entity identification layer,
a score is calculated for each possible entity tag sequence for the nested entity identification layer. On the basis, the training target of the nested entity recognition model is that the scores of the sample simple entity recognition result and the sample nested entity recognition result are the highest, and the ratio of the sample simple entity recognition result to the sum of the scores of all entity label sequences is larger and larger.
Based on any of the above embodiments, fig. 4 is a schematic structural diagram of a nested entity recognition apparatus provided in an embodiment of the present invention, and as shown in fig. 4, the apparatus includes a text determination unit 410 and a nested recognition unit 420.
The text determining unit 410 is configured to determine a text to be recognized;
the nested recognition unit 420 is configured to input a text to be recognized into the nested entity recognition model, and obtain a simple entity recognition result and a nested entity recognition result output by the nested entity recognition model;
the nested entity recognition model is obtained by training based on a sample text, a sample simple entity recognition result and a sample nested entity recognition result which correspond to the sample text;
the nested entity recognition model is used for determining a simple entity recognition result based on the text to be recognized and determining a nested entity recognition result based on the text to be recognized and the simple entity recognition result.
The device provided by the embodiment of the invention determines the identification result of the simple entity based on the text to be identified, determines the identification result of the nested entity based on the text to be identified and the identification result of the simple entity, and can realize the identification of the nested entity and the simple entity in the nested entity only by one model; and the obtained simple entity identification result and entity boundary information in the nested entity identification result show the nested relation between the nested entity and the simple entity inside the nested entity, and provide a better support function for the subsequent text parsing task.
Based on any of the above embodiments, the nested identifying unit 420 specifically includes:
the simple entity recognition unit is used for inputting a first text vector of a text to be recognized into a simple entity recognition layer of the nested entity recognition model to obtain a simple entity recognition result output by the simple entity recognition layer;
the attention unit is used for inputting the simple entity recognition result to an attention layer of the nested entity recognition model to obtain a simple entity attention vector output by the attention layer;
and the nested entity recognition unit is used for inputting the simple entity attention vector and the second text vector of the text to be recognized into a nested entity recognition layer of the nested entity recognition model to obtain a nested entity recognition result output by the nested entity recognition layer.
The device provided by the embodiment of the invention converts the simple entity recognition result into the simple entity attention vector based on the self-attention mechanism, and determines the nested entity recognition result based on the simple entity attention vector and the second text vector of the text to be recognized, thereby improving the accuracy of the nested entity recognition result.
According to any of the above embodiments, in the apparatus, the first text vector includes a word vector of each word in the text to be recognized, and a dictionary feature vector and/or a part-of-speech feature vector of each word in the text to be recognized.
According to the device provided by the embodiment of the invention, the first text vector comprises the word vector of each word in the text to be recognized and the dictionary characteristic vector and/or part-of-speech characteristic vector of each word in the text to be recognized, so that the accuracy of the recognition result of the simple entity is improved.
Based on any one of the embodiments, in the device, the dictionary feature vector of each word in the text to be recognized is obtained by matching the text to be recognized with a pre-constructed domain dictionary; wherein the dictionary feature vector for any word represents the type of entity that the word matches in the domain dictionary and the location of the word in the matching entity.
The device provided by the embodiment of the invention matches the text to be recognized with the pre-constructed domain dictionary to obtain the dictionary feature vector of each word in the text to be recognized, and is beneficial to improving the accuracy of the simple entity recognition result.
In any of the above embodiments, the apparatus further comprises an outcome correction unit.
The result correction unit is used for inputting the simple entity recognition result and/or the nested entity recognition result into a result correction layer of the nested entity recognition model to obtain a corrected simple entity recognition result and/or a nested entity recognition result output by the result correction layer.
The device provided by the embodiment of the invention corrects the simple entity recognition result and/or the nested entity recognition result based on the simple entity recognition result and/or the nested entity recognition result, and improves the accuracy of the simple entity recognition result and/or the nested entity recognition result.
Based on any of the embodiments above, the outcome correction unit is specifically configured to:
and inputting the simple entity recognition result or the nested entity recognition result into the result correction layer, and performing result correction on the simple entity recognition result or the nested entity recognition result by the result correction layer based on the entity label rule to obtain a corrected simple entity recognition result or a corrected nested entity recognition result output by the result correction layer.
The device provided by the embodiment of the invention corrects the result of the simple entity identification result or the nested entity identification result based on the entity label rule, and improves the accuracy of the simple entity identification result or the nested entity identification result.
Based on any of the embodiments above, the outcome correction unit is specifically configured to:
inputting the simple entity recognition result and the nested entity recognition result into a result correction layer, and when the result correction layer judges that the simple entity recognition result and the nested entity recognition result conflict, performing result correction on the simple entity recognition result or the nested entity recognition result to obtain a corrected simple entity recognition result and a nested entity recognition result output by the result correction layer;
the condition that the simple entity recognition result conflicts with the nested entity recognition result comprises that at least one of repeated entities, entities with the same boundary but different entity types and entities with crossed boundaries exists between the simple entity recognition result and the nested entity recognition result.
According to the device provided by the embodiment of the invention, when the conflict between the simple entity identification result and the nested entity identification result is judged and obtained, the result correction is carried out on the simple entity identification result or the nested entity identification result, so that the accuracy of the simple entity identification result and the nested entity identification result is improved.
Fig. 5 is a schematic structural diagram of an electronic device according to an embodiment of the present invention, and as shown in fig. 5, the electronic device may include: a processor (processor)510, a communication Interface (Communications Interface)520, a memory (memory)530 and a communication bus 540, wherein the processor 510, the communication Interface 520 and the memory 530 communicate with each other via the communication bus 540. Processor 510 may call logical commands in memory 530 to perform the following method: determining a text to be recognized; inputting the text to be recognized into a nested entity recognition model to obtain a simple entity recognition result and a nested entity recognition result which are output by the nested entity recognition model; the nested entity recognition model is obtained by training based on a sample text, and a sample simple entity recognition result and a sample nested entity recognition result which correspond to the sample text; the nested entity recognition model is used for determining the simple entity recognition result based on the text to be recognized and determining the nested entity recognition result based on the text to be recognized and the simple entity recognition result.
In addition, the logic commands in the memory 530 may be implemented in the form of software functional units and stored in a computer readable storage medium when the logic commands are sold or used as independent products. 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 a plurality of commands for enabling 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: a U-disk, a removable hard disk, a Read-Only Memory (ROM), a Random Access Memory (RAM), a magnetic disk or an optical disk, and other various media capable of storing program codes.
Embodiments of the present invention further provide a non-transitory computer-readable storage medium, on which a computer program is stored, where the computer program is implemented to perform the method provided in the foregoing embodiments when executed by a processor, and the method includes: determining a text to be recognized; inputting the text to be recognized into a nested entity recognition model to obtain a simple entity recognition result and a nested entity recognition result which are output by the nested entity recognition model; the nested entity recognition model is obtained by training based on a sample text, and a sample simple entity recognition result and a sample nested entity recognition result which correspond to the sample text; the nested entity recognition model is used for determining the simple entity recognition result based on the text to be recognized and determining the nested entity recognition result based on the text to be recognized and the simple entity recognition result.
The above-described embodiments of the apparatus are merely illustrative, and the units described as separate parts may or may not be physically separate, and parts displayed as units may or may not be physical units, may be located in one place, or may be distributed on a plurality of network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of the present embodiment. One of ordinary skill in the art can understand and implement it without inventive effort.
Through the above description of the embodiments, those skilled in the art will clearly understand that each embodiment can be implemented by software plus a necessary general hardware platform, and certainly can also be implemented by hardware. With this understanding in mind, the above technical solutions may be embodied in the form of a software product, which can be stored in a computer-readable storage medium, such as ROM/RAM, magnetic disk, optical disk, etc., and includes commands for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute the method according to the embodiments or some parts of the embodiments.
Finally, it should be noted that: the above examples are only intended to illustrate the technical solution of the present invention, but not to limit it; 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 method for identifying a nested entity, comprising:
determining a text to be recognized;
inputting the text to be recognized into a nested entity recognition model to obtain a simple entity recognition result and a nested entity recognition result which are output by the nested entity recognition model;
the nested entity recognition model is obtained by training based on a sample text, and a sample simple entity recognition result and a sample nested entity recognition result which correspond to the sample text;
the nested entity recognition model is used for determining the simple entity recognition result based on the text to be recognized and determining the nested entity recognition result based on the text to be recognized and the simple entity recognition result.
2. The method according to claim 1, wherein the step of inputting the text to be recognized into the nested entity recognition model to obtain the simple entity recognition result and the nested entity recognition result output by the nested entity recognition model specifically comprises:
inputting the first text vector of the text to be recognized to a simple entity recognition layer of the nested entity recognition model to obtain the simple entity recognition result output by the simple entity recognition layer;
inputting the simple entity recognition result to an attention layer of the nested entity recognition model to obtain a simple entity attention vector output by the attention layer;
and inputting the simple entity attention vector and the second text vector of the text to be recognized into a nested entity recognition layer of the nested entity recognition model to obtain the nested entity recognition result output by the nested entity recognition layer.
3. The nested entity recognition method of claim 2, wherein the first text vector comprises a word vector for each word in the text to be recognized, and a dictionary feature vector and/or part-of-speech feature vector for each word in the text to be recognized.
4. The nested entity recognition method of claim 3, wherein the dictionary feature vector of each word in the text to be recognized is obtained by matching the text to be recognized with a pre-constructed domain dictionary;
wherein the dictionary feature vector for any word represents the type of entity that the any word matches in the domain dictionary and the location of the any word in the matching entity.
5. The method according to any one of claims 2 to 4, wherein the inputting the simple entity attention vector and the text vector of the text to be recognized into a nested entity recognition layer of the nested entity recognition model to obtain the nested entity recognition result output by the nested entity recognition layer, and then further comprising:
and inputting the simple entity recognition result and/or the nested entity recognition result into a result correction layer of the nested entity recognition model to obtain a corrected simple entity recognition result and/or a nested entity recognition result output by the result correction layer.
6. The method according to claim 5, wherein the inputting the simple entity recognition result or the nested entity recognition result into an outcome correction layer of the nested entity recognition model to obtain a corrected simple entity recognition result or a nested entity recognition result output by the outcome correction layer specifically includes:
and inputting the simple entity recognition result or the nested entity recognition result into the result correction layer, and performing result correction on the simple entity recognition result or the nested entity recognition result by the result correction layer based on an entity label rule to obtain a corrected simple entity recognition result or a corrected nested entity recognition result output by the result correction layer.
7. The method according to claim 5, wherein the inputting the simple entity recognition result and the nested entity recognition result into an outcome correction layer of the nested entity recognition model to obtain a corrected simple entity recognition result and a nested entity recognition result output by the outcome correction layer specifically comprises:
inputting the simple entity recognition result and the nested entity recognition result into the result correction layer, and when judging that the simple entity recognition result and the nested entity recognition result conflict, the result correction layer corrects the simple entity recognition result or the nested entity recognition result to obtain a corrected simple entity recognition result and a nested entity recognition result output by the result correction layer;
the condition that the simple entity recognition result conflicts with the nested entity recognition result comprises that at least one of repeated entities, entities with the same boundary but different entity types and entities with crossed boundaries exists between the simple entity recognition result and the nested entity recognition result.
8. A nested entity recognition apparatus, comprising:
the text determining unit is used for determining a text to be recognized;
the nested recognition unit is used for inputting the text to be recognized into a nested entity recognition model to obtain a simple entity recognition result and a nested entity recognition result which are output by the nested entity recognition model;
the nested entity recognition model is obtained by training based on a sample text, and a sample simple entity recognition result and a sample nested entity recognition result which correspond to the sample text;
the nested entity recognition model is used for determining the simple entity recognition result based on the text to be recognized and determining the nested entity recognition result based on the text to be recognized and the simple entity recognition result.
9. An electronic device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, characterized in that the steps of the nested entity identification method of one of claims 1 to 7 are implemented when the program is executed by the processor.
10. A non-transitory computer readable storage medium, having stored thereon a computer program, characterized in that the computer program, when being executed by a processor, realizes the steps of the nested entity identification method of any one of claims 1 to 7.
Applications Claiming Priority (2)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010566433.9A CN111753545A (en) | 2020-06-19 | 2020-06-19 | Nested entity recognition method and device, electronic equipment and storage medium |
| CN2020105664339 | 2020-06-19 |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN112836514A true CN112836514A (en) | 2021-05-25 |
| CN112836514B CN112836514B (en) | 2024-07-02 |
Family
ID=72675518
Family Applications (2)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010566433.9A Withdrawn CN111753545A (en) | 2020-06-19 | 2020-06-19 | Nested entity recognition method and device, electronic equipment and storage medium |
| CN202110343229.5A Active CN112836514B (en) | 2020-06-19 | 2021-03-30 | Nested entity identification method, apparatus, electronic device and storage medium |
Family Applications Before (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010566433.9A Withdrawn CN111753545A (en) | 2020-06-19 | 2020-06-19 | Nested entity recognition method and device, electronic equipment and storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (2) | CN111753545A (en) |
Families Citing this family (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112257421B (en) * | 2020-12-21 | 2021-04-23 | 完美世界(北京)软件科技发展有限公司 | Nested entity data identification method and device and electronic equipment |
| CN112966511B (en) * | 2021-02-08 | 2024-03-15 | 广州探迹科技有限公司 | Entity word recognition method and device |
| CN113239659A (en) * | 2021-04-21 | 2021-08-10 | 上海快确信息科技有限公司 | Text number extraction device integrating rules |
| CN112988979B (en) * | 2021-04-29 | 2021-10-08 | 腾讯科技（深圳）有限公司 | Entity identification method, entity identification device, computer readable medium and electronic equipment |
| CN113392649B (en) * | 2021-07-08 | 2023-04-07 | 上海浦东发展银行股份有限公司 | Identification method, device, equipment and storage medium |
| CN114282538A (en) * | 2021-11-24 | 2022-04-05 | 重庆邮电大学 | Chinese text data word vector characterization method based on BIE position word list |
| CN114462391B (en) * | 2022-03-14 | 2024-05-14 | 和美（深圳）信息技术股份有限公司 | Nested entity identification method and system based on contrast learning |
| CN116843432B (en) * | 2023-05-10 | 2024-03-22 | 北京微聚智汇科技有限公司 | Anti-fraud method and device based on address text information |
| CN116522943B (en) * | 2023-05-11 | 2024-06-07 | 北京微聚智汇科技有限公司 | Address element extraction method and device, storage medium and computer equipment |
Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20180260474A1 (en) * | 2017-03-13 | 2018-09-13 | Arizona Board Of Regents On Behalf Of The University Of Arizona | Methods for extracting and assessing information from literature documents |
| CN109388807A (en) * | 2018-10-30 | 2019-02-26 | 中山大学 | The method, apparatus and storage medium of electronic health record name Entity recognition |
| CN110008469A (en) * | 2019-03-19 | 2019-07-12 | 桂林电子科技大学 | A kind of multi-level name entity recognition method |
| CN110597970A (en) * | 2019-08-19 | 2019-12-20 | 华东理工大学 | Multi-granularity medical entity joint identification method and device |
| CN110705302A (en) * | 2019-10-11 | 2020-01-17 | 掌阅科技股份有限公司 | Named entity recognition method, electronic device and computer storage medium |
| CN110866402A (en) * | 2019-11-18 | 2020-03-06 | 北京香侬慧语科技有限责任公司 | Named entity identification method and device, storage medium and electronic equipment |
| CN110956042A (en) * | 2019-12-16 | 2020-04-03 | 中国电子科技集团公司信息科学研究院 | Nested named entity recognition method and system, electronic device and readable medium |
| CN111104800A (en) * | 2019-12-24 | 2020-05-05 | 东软集团股份有限公司 | Entity identification method, device, equipment, storage medium and program product |
-
2020
- 2020-06-19 CN CN202010566433.9A patent/CN111753545A/en not_active Withdrawn
-
2021
- 2021-03-30 CN CN202110343229.5A patent/CN112836514B/en active Active
Patent Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20180260474A1 (en) * | 2017-03-13 | 2018-09-13 | Arizona Board Of Regents On Behalf Of The University Of Arizona | Methods for extracting and assessing information from literature documents |
| CN109388807A (en) * | 2018-10-30 | 2019-02-26 | 中山大学 | The method, apparatus and storage medium of electronic health record name Entity recognition |
| CN110008469A (en) * | 2019-03-19 | 2019-07-12 | 桂林电子科技大学 | A kind of multi-level name entity recognition method |
| CN110597970A (en) * | 2019-08-19 | 2019-12-20 | 华东理工大学 | Multi-granularity medical entity joint identification method and device |
| CN110705302A (en) * | 2019-10-11 | 2020-01-17 | 掌阅科技股份有限公司 | Named entity recognition method, electronic device and computer storage medium |
| CN110866402A (en) * | 2019-11-18 | 2020-03-06 | 北京香侬慧语科技有限责任公司 | Named entity identification method and device, storage medium and electronic equipment |
| CN110956042A (en) * | 2019-12-16 | 2020-04-03 | 中国电子科技集团公司信息科学研究院 | Nested named entity recognition method and system, electronic device and readable medium |
| CN111104800A (en) * | 2019-12-24 | 2020-05-05 | 东软集团股份有限公司 | Entity identification method, device, equipment, storage medium and program product |
Non-Patent Citations (5)
| Title |
|---|
| 付春元: ""汉语嵌套命名实体识别方法研究"", 《中国优秀硕士论文全文数据库》, pages 1 - 74 * |
| 张若彬;刘嘉勇;何祥;: "基于BLSTM-CRF模型的安全漏洞领域命名实体识别", 四川大学学报(自然科学版), no. 03 * |
| 杨健;黄瑞章;丁志远;陈艳平;秦永彬;: "基于边界识别与组合的裁判文书证据抽取方法研究", 中文信息学报, no. 03 * |
| 王海宁;周菊香;徐天伟;: "融合深度学习与规则的民族工艺品领域命名实体识别", 云南师范大学学报(自然科学版), no. 02 * |
| 胡俊锋;陈蓉;陈源;陈浩;于中华;: "一种松耦合的生物医学命名实体识别算法", 计算机应用, no. 11 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN112836514B (en) | 2024-07-02 |
| CN111753545A (en) | 2020-10-09 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN112836514A (en) | Nested entity recognition method and device, electronic equipment and storage medium | |
| CN110457675B (en) | Predictive model training method and device, storage medium and computer equipment | |
| CN110704576B (en) | Text-based entity relationship extraction method and device | |
| CN110147451B (en) | Dialogue command understanding method based on knowledge graph | |
| CN111985239A (en) | Entity identification method and device, electronic equipment and storage medium | |
| CN111414746B (en) | Method, device, equipment and storage medium for determining matching statement | |
| CN112992125B (en) | Voice recognition method and device, electronic equipment and readable storage medium | |
| CN112052324A (en) | Intelligent question answering method and device and computer equipment | |
| CN112487139A (en) | Text-based automatic question setting method and device and computer equipment | |
| CN110096572B (en) | Sample generation method, device and computer readable medium | |
| CN113268576B (en) | Deep learning-based department semantic information extraction method and device | |
| CN112016320A (en) | English punctuation adding method, system and equipment based on data enhancement | |
| CN112016271A (en) | Language style conversion model training method, text processing method and device | |
| CN111309893A (en) | Method and device for generating similar problems based on source problems | |
| CN115599901A (en) | Machine question-answering method, device, equipment and storage medium based on semantic prompt | |
| CN116304023A (en) | Method, system and storage medium for extracting bidding elements based on NLP technology | |
| CN112800184A (en) | Short text comment emotion analysis method based on Target-Aspect-Opinion joint extraction | |
| CN113408287A (en) | Entity identification method and device, electronic equipment and storage medium | |
| CN115017890A (en) | Text error correction method and device based on character pronunciation and character font similarity | |
| CN113553853B (en) | Named entity recognition method and device, computer equipment and storage medium | |
| CN112488111B (en) | Indication expression understanding method based on multi-level expression guide attention network | |
| CN113362815A (en) | Voice interaction method, system, electronic equipment and storage medium | |
| CN114003700A (en) | Method and system for processing session information, electronic device and storage medium | |
| CN107783958B (en) | Target statement identification method and device | |
| CN117828024A (en) | Plug-in retrieval method, device, storage medium and equipment |
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 | ||
| GR01 | Patent grant |