CN110210021A - Read understanding method and device - Google Patents
Read understanding method and device Download PDFInfo
- Publication number
- CN110210021A CN110210021A CN201910429805.0A CN201910429805A CN110210021A CN 110210021 A CN110210021 A CN 110210021A CN 201910429805 A CN201910429805 A CN 201910429805A CN 110210021 A CN110210021 A CN 110210021A
- Authority
- CN
- China
- Prior art keywords
- answer
- text
- sample
- reading
- training
- 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
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F40/00—Handling natural language data
- G06F40/20—Natural language analysis
- G06F40/205—Parsing
-
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Artificial Intelligence (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Data Mining & Analysis (AREA)
- General Health & Medical Sciences (AREA)
- Computational Linguistics (AREA)
- Audiology, Speech & Language Pathology (AREA)
- Health & Medical Sciences (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Evolutionary Computation (AREA)
- Evolutionary Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Computational Biology (AREA)
- Life Sciences & Earth Sciences (AREA)
- Electrically Operated Instructional Devices (AREA)
- Machine Translation (AREA)
Abstract
The present invention proposes a kind of reading understanding method and device, wherein this method comprises: by obtaining preset target problem and text to be read;Understand that model carries out understanding analysis to the text to be read according to preset reading, generates the answer type probability, answer text and corresponding confidence level of the target problem;Target answer corresponding with the target problem is determined according to the answer type probability, the answer text and corresponding confidence level.To understand that model better understands out the long answer text or short answer text of high quality to target problem and text to be read based on preset reading, all different practical situations can show good, it is only good to a kind of understanding effect of the answer of answer type to be no longer limited to existing special purpose model, and improve the understanding effect of the answer to different answer types compared to existing universal model.
Description
Technical field
The present invention relates to artificial intelligence field more particularly to a kind of reading understanding methods and device.
Background technique
Artificial intelligence (Artificial Intelligence, AI) is research, develops for simulating, extending and extending people
Intelligence theory, method, a new technological sciences of technology and application system.Artificial intelligence is one of computer science
Branch, it attempts to understand the essence of intelligence, and produces a kind of new intelligence that can be made a response in such a way that human intelligence is similar
Energy machine, the research in the field include robot, language identification, image recognition, natural language processing, question answering system and expert system
System etc..
Currently, will be to be answered the problem of and relevant reading material, which are input to the reading that training finishes, understands that model carries out people
The intellectual reading of work understands more and more extensive.Existing reading understands that model point mainly has special purpose model and universal model:
Special purpose model is to be obtained using the reading material training of same answer type, but special purpose model is limited in that
Can only a kind of understanding effect to answer type it is relatively good.The long answer obtained for example, by using the sample of the long answer type of magnanimity
Model, the short answer model obtained using the sample training of the short answer type of magnanimity, due to long answer model and short answer mould
Type has different model parameter and training data, and long answer model is relatively good to the understanding effect of long answer, and to short answer
Understanding effect it is bad, short answer model can only be relatively good to the understanding effect of short answer, and not to the understanding effect of long answer
It is good.
Universal model is obtained despite the sample joint training using different answer types, but the structure ginseng of universal model
There is no the answers for different answer types to be respectively set for number, but shares a set of model parameter, causes to long answer
Understanding effect be less than long answer model, short answer model, the reading effect of model are less than to the understanding effect of short answer
It is bad.
Therefore, how preferably to carry out reading and understand as technical problem urgently to be resolved.
Summary of the invention
The present invention is directed to solve at least some of the technical problems in related technologies.
For this purpose, the first purpose of this invention is to propose a kind of reading understanding method.
Second object of the present invention is to propose that a kind of reading understands device.
Third object of the present invention is to propose a kind of computer equipment.
Fourth object of the present invention is to propose a kind of computer readable storage medium.
In order to achieve the above object, first aspect present invention embodiment proposes a kind of reading understanding method, comprising:
Obtain preset target problem and text to be read;
Understand that model carries out understanding analysis to the text to be read according to preset reading, generates the target problem
Answer type probability, answer text and corresponding confidence level;
According to the determination of the answer type probability, the answer text and corresponding confidence level and the target problem pair
The target answer answered.
In one possible implementation, it is described according to the answer text type probability, the answer text and
Corresponding confidence level determines that target answer corresponding with the target problem includes:
It obtains the first product of the confidence level of long answer type probability and long answer text, short answer type probability and short answers
Second product of the confidence level of case text;Judge whether first product is greater than second product;
If so, the long answer text is determined as target answer corresponding with the target problem;
If it is not, the short answer text is then determined as target answer corresponding with the target problem.
In one possible implementation, before the preset target problem of the acquisition and text to be read, also
Include:
Obtain first sample set, wherein the first sample that the first sample is concentrated includes training problem and reading material
Sample, starting position and end position of the correct option in reading material sample of long answer type, for characterizing described
The answer type of one sample is the labeled data of long answer;
Obtain the second sample set, wherein the second sample in second sample set includes training problem and reading material
Sample, starting position and end position of the correct option in reading material sample of short answer type, for characterizing described
The answer type of two samples is the labeled data of short answer;
Model is understood based on the initial reading of the first sample set and second sample set training, is obtained described default
Reading understand model.
In one possible implementation, the initial reading understands that model includes at least encoder, the first prediction
Layer, the second prediction interval, classifier, the reading initial based on the first sample set and second sample set training understand
Model obtains the preset reading and understands that model includes:
By the first sample and second sample stated in the second sample set point that the first sample is concentrated
It is not input in the encoder and is encoded;
Using first sample training first prediction interval after each coding, and using described after each coding
After the first sample set and each coding after trained second prediction interval of second sample set and each coding of use
The second sample set training classifier understands model to obtain the preset reading;
Wherein, the first prediction interval that training finishes can be predicted to need the problem of answering corresponding long answer text and its be set
Reliability, the second prediction interval that training finishes can be predicted to need the problem of answering corresponding short answer text and its confidence level, be instructed
The classifier that white silk finishes, which can differentiate, needs the problem of answering corresponding answer type probability.
In one possible implementation, after the preset target problem of the acquisition and text to be read, also
Include:
The target problem and the text to be read are spliced, wherein in splicing, in the target
The separator that characterization problems are added before problem, point of addition characterization paragraph before the paragraph of the text to be read
Every symbol.
Reading understanding method provided in an embodiment of the present invention, by obtaining preset target problem and text to be read;
Understand that model carries out understanding analysis to the text to be read according to preset reading, generates the answer class of the target problem
Type probability, answer text and corresponding confidence level;According to the answer type probability, the answer text and corresponding set
Reliability determines target answer corresponding with the target problem.To based on it is preset reading understand model to target problem and to
Long answer text or short answer text that text better understands out high quality are read, all different practical situations can show good
It is good, it is only good to a kind of understanding effect of the answer of answer type and logical compared to existing to be no longer limited to existing special purpose model
The understanding effect of the answer to different answer types is improved with model.
In order to achieve the above object, second aspect of the present invention embodiment, which proposes a kind of read, understands device, comprising:
Module is obtained, for obtaining preset target problem and text to be read;
Generation module, it is raw for understanding that model carries out understanding analysis to the text to be read according to preset reading
At the answer type probability, answer text and corresponding confidence level of the target problem;
Determining module, for according to the answer type probability, the answer text and corresponding confidence level determine with
The corresponding target answer of the target problem.
In one possible implementation, the determining module is specifically used for:
It obtains the first product of the confidence level of long answer type probability and long answer text, short answer type probability and short answers
Second product of the confidence level of case text;Judge whether first product is greater than second product;
If so, the long answer text is determined as target answer corresponding with the target problem;
If it is not, the short answer text is then determined as target answer corresponding with the target problem.
In one possible implementation, described device further include: training module;
The acquisition module, is also used to obtain first sample set, wherein the first sample that the first sample is concentrated includes
Starting position and stop bits of the correct option in reading material sample of training problem and reading material sample, long answer type
It sets, the labeled data that the answer type for characterizing the first sample is long answer;
The acquisition module, is also used to obtain the second sample set, wherein the second sample in second sample set includes
Starting position and stop bits of the correct option in reading material sample of training problem and reading material sample, short answer type
It sets, the labeled data that the answer type for characterizing second sample is short answer;
The training module, for being understood based on the initial reading of the first sample set and second sample set training
Model obtains the preset reading and understands model.
In one possible implementation, the initial reading understands that model includes at least encoder, the first prediction
Layer, the second prediction interval, classifier, the training module are specifically used for:
By the first sample and second sample stated in the second sample set point that the first sample is concentrated
It is not input in the encoder and is encoded;
Using first sample training first prediction interval after each coding, and using described after each coding
After the first sample set and each coding after trained second prediction interval of second sample set and each coding of use
The second sample set training classifier understands model to obtain the preset reading;
Wherein, the first prediction interval that training finishes can be predicted to need the problem of answering corresponding long answer text and its be set
Reliability, the second prediction interval that training finishes can be predicted to need the problem of answering corresponding short answer text and its confidence level, be instructed
The classifier that white silk finishes, which can differentiate, needs the problem of answering corresponding answer type probability.
In one possible implementation, described device further include: splicing module;
The splicing, for asking the target after the preset target problem of the acquisition and text to be read
Topic and the text to be read are spliced, wherein in splicing, addition characterization is asked before the target problem
The separator of topic, the separator of addition characterization paragraph before the paragraph of the text to be read.
Reading provided in an embodiment of the present invention understands device, by obtaining preset target problem and text to be read;
Understand that model carries out understanding analysis to the text to be read according to preset reading, generates the answer class of the target problem
Type probability, answer text and corresponding confidence level;According to the answer type probability, the answer text and corresponding set
Reliability determines target answer corresponding with the target problem.To based on it is preset reading understand model to target problem and to
Long answer text or short answer text that text better understands out high quality are read, all different practical situations can show good
It is good, it is only good to a kind of understanding effect of the answer of answer type and logical compared to existing to be no longer limited to existing special purpose model
The understanding effect of the answer to different answer types is improved with model.
In order to achieve the above object, third aspect present invention embodiment proposes a kind of computer equipment, including memory, processing
On a memory and the computer program that can run on a processor, when processor execution described program, is realized for device and storage
It is as described above to read understanding method.
To achieve the goals above, fourth aspect present invention embodiment proposes a kind of computer readable storage medium, when
When instruction in the storage medium is executed by processor, reading understanding method as described above is realized.
The additional aspect of the present invention and advantage will be set forth in part in the description, and will partially become from the following description
Obviously, or practice through the invention is recognized.
Detailed description of the invention
Above-mentioned and/or additional aspect and advantage of the invention will become from the following description of the accompanying drawings of embodiments
Obviously and it is readily appreciated that, in which:
Fig. 1 is a kind of flow diagram for reading understanding method provided in an embodiment of the present invention;
Fig. 2 is another flow diagram for reading understanding method provided in an embodiment of the present invention;
Fig. 3 is the model structure that exemplary existing reading understands model;
Fig. 4 is the model structure that illustrative preset reading provided in an embodiment of the present invention understands model;
Fig. 5 is the structural schematic diagram that a kind of reading provided in an embodiment of the present invention understands device;
Fig. 6 is the structural schematic diagram that another reading provided in an embodiment of the present invention understands device;
Fig. 7 is a kind of structural schematic diagram of computer equipment provided in an embodiment of the present invention.
Specific embodiment
The embodiment of the present invention is described below in detail, examples of the embodiments are shown in the accompanying drawings, wherein from beginning to end
Same or similar label indicates same or similar element or element with the same or similar functions.Below with reference to attached
The embodiment of figure description is exemplary, it is intended to is used to explain the present invention, and is not considered as limiting the invention.
Below with reference to the accompanying drawings the reading understanding method and device of the embodiment of the present invention are described.
Fig. 1 is a kind of flow diagram for reading understanding method provided in an embodiment of the present invention.Present embodiments provide one
Kind reads understanding method, and executing subject is to read to understand device, which is made of hardware and/or software.Read reason
Solution device be specifically as follows the software installed on hardware device, such as terminal device, background server etc. or hardware device or
Application program etc..
As shown in Figure 1, the reading understanding method, comprising the following steps:
S101, preset target problem and text to be read are obtained.
In practical applications, the answer of problem may be several words, phrase etc., it is also possible to very long sentence or section
It falls.Short answer text is referred to as to several words, the such answer of phrase, the answer of very long sentence or paragraph is claimed
Be long answer text, certainly, how short answer text and long answer text define depending on practical situation.
For example, reading material be " five danger one gold medals refer to that being collectively referred to as several protection treatment of labourer is given by employing unit,
Including endowment insurance, medical insurance, unemployment insurance, work-related injury insurance and birth insurance and public accumalation fund for housing construction ".
Problem 1: " five one gold medals of danger are protection treatment ", the answer text of problem 1 is "Yes", and the answer of problem 1 is compared
Briefly, the answer type of problem 1 is short answer.
Problem 2: " which five one gold medals of danger include ", the answer text of problem 2 is that " endowment insurance, medical insurance, unemployment are protected
Danger, work-related injury insurance and birth insurance and public accumalation fund for housing construction ", the answer of problem 2 is long, and the answer type of problem 2 is that length is answered
Case.
In the present embodiment, preset target problem is set according to the practical situation of text to be read.
For example, text to be read is that " five one gold medals of danger refer to that several protection treatment of labourer are given by employing unit
It is collectively referred to as, including endowment insurance, medical insurance, unemployment insurance, work-related injury insurance and birth insurance and public accumalation fund for housing construction." preset mesh
Mark problem is " five danger one gold medals be protection treatment ", " which five one gold medals of danger include " etc..
In another example text to be read is the article for introducing the earth, this article is related to earth composition, tellurian object
Kind etc., preset target problem is " which composition of the earth has ", " having which species on the earth " etc..
S102, understand that model carries out understanding analysis to the text to be read according to preset reading, generate the mesh
Answer type probability, answer text and the corresponding confidence level of mark problem.
In the present embodiment, preset reading is constructed using the training data of magnanimity in advance and understands model.This is preset to read
Reading understands that the universal performance of model is good, can be in the corresponding long answer of target problem, can be based on target problem and to be read
Text well understood that out the long answer of target problem;Again mesh can be based in the corresponding short answer of target problem
Mark problem and text to be read well understood that out the short answer of target problem.
Specifically, preset reading understands that model carries out understanding analysis to problem to be read based on target problem, exports
Understand that result includes: the answer type probability, answer text and corresponding confidence level of target problem, but it is not limited to this.
Wherein, the answer type probability of target problem includes long answer type probability, short answer type probability.
Wherein, the answer text of target problem includes long answer text, short answer text.
For example, reading understands mould after target problem and text input to be read are understood model to preset reading
The understanding result of type output are as follows: long answer text AlongAnd its confidence level Slong, short answer text AshortAnd its confidence level Sshort、
Long answer type probability Plong, short answer type probability Pshort。
Further, in order to which can make that model quickly determines input is section in problem or text to be read
It falls, before target problem and text input to be read are understood that model is understood to preset reading, to the target
Problem and the text to be read are spliced, wherein in splicing, characterization is added before the target problem
The separator of problem, the separator of characterization problems are, for example, [CLS];If reading material sample is made of one or more paragraphs,
The separator of addition characterization paragraph, the separator for characterizing paragraph are, for example, before the paragraph of the text to be read
[SEP]。
S103, it is determined and the target according to the answer type probability, the answer text and corresponding confidence level
The corresponding target answer of problem.
In the present embodiment, get after preset reading understands the understanding result of model, it can be there are many implementation root
According to the target answer for understanding that result determines target problem.Such as it can be according to long answer type probability and short answer type probability
Size determines the target answer of target problem from long answer text and short answer text, specifically, by the big answer text of probability
Originally it is determined as the target answer of target problem.In another example can be according to the confidence level of long answer text and the confidence of short answer text
The size of degree determines the target answer of target problem from long answer text and short answer text, specifically, confidence level is big
Answer text is determined as the target answer of target problem, and but it is not limited to this.
As a kind of possible implementation, in order to preferably carry out decision, comprehensive answer type probability and answer text
Confidence level determine the target answer of target problem, the specific implementation of step S103 are as follows:
S1031, the first product of the confidence level of the long answer type probability of acquisition and long answer text, short answer type probability
With the second product of the confidence level of short answer text.S1032, judge whether first product is greater than second product.
S1033, if so, the long answer text is determined as target answer corresponding with the target problem;
S1034, if it is not, the short answer text is then determined as target answer corresponding with the target problem.
For ease of description and understand, to understand result for long answer text AlongAnd its confidence level Slong, short answer text
AshortAnd its confidence level Sshort, long answer type probability Plong, short answer type probability PshortFor.
If Plong*SlongGreater than Pshort*Sshort, by long answer text AlongIt is determined as target corresponding with target problem to answer
Case;If Plong*SlongLess than Pshort*Sshort, by short answer text AshortIt is determined as target answer corresponding with target problem.
Reading understanding method provided in an embodiment of the present invention, by obtaining preset target problem and text to be read;
Understand that model carries out understanding analysis to the text to be read according to preset reading, generates the answer class of the target problem
Type probability, answer text and corresponding confidence level;According to the answer type probability, the answer text and corresponding set
Reliability determines target answer corresponding with the target problem.To based on it is preset reading understand model to target problem and to
Long answer text or short answer text that text better understands out high quality are read, different practical situations can be showed good
It is good, it is only good to a kind of understanding effect of the answer of answer type and logical compared to existing to be no longer limited to existing special purpose model
The understanding effect of the answer to different answer types is improved with model.
Fig. 2 is another flow diagram for reading understanding method provided in an embodiment of the present invention.The present embodiment is to default
Reading understand that the training stage of model is illustrated.In conjunction with reference Fig. 2, on the basis of embodiment shown in Fig. 1, in step
Before S101, the reading understanding method is further comprising the steps of:
S104, first sample set is obtained.
Wherein, the first sample that the first sample is concentrated includes training problem and reading material sample, long answer type
Starting position and end position of the correct option in reading material sample, the answer type for characterizing the first sample
For the labeled data of long answer.
S105, the second sample set is obtained.
Wherein, the second sample in second sample set includes training problem and reading material sample, short answer type
Starting position and end position of the correct option in reading material sample, the answer type for characterizing second sample
For the labeled data of short answer.
S106, model is understood based on the initial reading of the first sample set and second sample set training, obtains institute
It states preset reading and understands model.
As a kind of possible implementation, if initial reading understands that model includes including at least encoder, first in advance
Survey layer, the second prediction interval, classifier, the implementation of step S106 the following steps are included:
S1061, by the first sample that the first sample is concentrated and described second stated in the second sample set
Sample is separately input to be encoded in the coding layer.
It, will be in first sample or the second sample before being encoded to first sample or the second sample in the present embodiment
Training problem and reading material sample are spliced, and when splicing, the separator of characterization problems, table are added before training problem
The separator of sign problem is, for example, [CLS];If reading material sample is made of one or more paragraphs, add before each paragraph
The separator for adding characterization paragraph is, for example, [SEP].Encoder by identify separator can quickly determine input be problem also
It is the paragraph in reading material sample.
S1062, first prediction interval is trained using the first sample after each coding, and using after each coding
Second sample set training second prediction interval and using the first sample set and each volume after each coding
Second sample set training classifier after code understands model to obtain the preset reading.
In the present embodiment, the first prediction interval that training finishes can be predicted to need the problem of answering corresponding long answer text
And its confidence level, the second prediction interval that training finishes can be predicted to need the problem of answering corresponding short answer text and its confidence
Degree, the classifier that training finishes, which can differentiate, needs the problem of answering corresponding answer type probability.
Fig. 3 is the model structure that exemplary existing reading understands model.Understand model for BERT with existing reading
For model, BERT model includes an encoder, a prediction interval.The problem of to be answered and text to be read are spliced
Afterwards, it is input to encoder to be encoded, exports coding vector;Coding vector obtains corresponding answer by the prediction of prediction interval.
Fig. 4 is the model structure that illustrative preset reading provided in an embodiment of the present invention understands model.In Fig. 4
In, it devises an encoder and each first sample or each second sample is encoded, design first prediction interval (figure
Long answer prediction interval in 4) it can predict to need the long answer for the problem of answering and its confidence level, design second prediction interval
(the short answer prediction interval in Fig. 4) can be predicted to need the short answer for the problem of answering and its confidence level, devise an answer
Type sorter can judge to need the answer type probability for the problem of answering.Compared to shown in Fig. 3, existing reading understands mould
Type, the preset reading of the embodiment of the present invention have understood more than model a prediction interval and a classifier.
Training stage of the first prediction interval is illustrated at this:
Using in the first sample after coding training problem and reading material sample as input, by the first sample after coding
Starting position and end position of the correct option of long answer type in this in reading material sample are as desired output, instruction
Practice the first prediction interval, until the first prediction interval is restrained, the first prediction interval after convergence can be predicted to need the problem of answering corresponding
Long answer text and its confidence level.
It should be pointed out that the structure of the first prediction interval can understand the structure of the prediction interval in model with existing reading
It is identical, it can also be with designed, designed.
Training stage of the second prediction interval is illustrated at this:
Using in the second sample after coding training problem and reading material sample as input, by the second sample after coding
Starting position and end position of the correct option of short answer type in this in reading material sample are as desired output, instruction
Practice the second prediction interval, until the second prediction interval is restrained, the second prediction interval after convergence can be predicted to need the problem of answering corresponding
Short answer text and its confidence level.
It should be pointed out that the structure of the second prediction interval can understand the structure of the prediction interval in model with existing reading
It is identical, it can also be with designed, designed.
Training stage of classifier is illustrated at this:
By in the first sample after coding training problem, reading material sample, be used to characterize answering for the first sample
Case type be long answer labeled data as a training sample, and by the second sample after coding training problem,
Reading material sample, the answer type for characterizing second sample are the labeled data of short answer as a trained sample
This;
Using in training sample training problem, reading material sample as output, by training sample labeled data make
For desired output, training classifier, until classifier is restrained, the classifier differentiation after convergence needs the problem of answering corresponding to answer
Case type probability.
It may be noted that when, classifier be, for example, the classifier based on support vector machines, the classifier based on decision tree, but
It is not limited to this.
Reading understanding method provided in an embodiment of the present invention, compared with existing reading understands model, preset reading reason
Solving model not only includes the first prediction interval that can predict long answer in model structure, can predict the second prediction of short answer
Layer, further includes the classifier that can differentiate answer type, since the first prediction interval and the second prediction interval are to separate training, for
Different answer types corresponds to different model structure and parameter, therefore the preset reading understands model for different practical situations
Reading understand that scene can show well flexibility, better effect;Pass through joint training, model learning to preferably spy
Sign, universal performance is more preferable, be no longer limited to existing special purpose model only it is good to a kind of understanding effect of the answer of answer type, very
Solves the collision problem between long answer model and short answer model well, and existing general compared to common model parameter
Model improves the understanding effect of the answer to different answer types.
Fig. 5 is the structural schematic diagram that a kind of reading provided in an embodiment of the present invention understands device.Present embodiments provide one
Kind, which is read, understands device, which is the executing subject for reading understanding method, which is made of hardware and/or software.
As shown in figure 5, the reading understands that device includes: to obtain module 11, generation module 12, determining module 13.
Module 11 is obtained, for obtaining preset target problem and text to be read;
Generation module 12, for understanding that model carries out understanding analysis to the text to be read according to preset reading,
Generate the answer type probability, answer text and corresponding confidence level of the target problem;
Determining module 13, for being determined according to the answer type probability, the answer text and corresponding confidence level
Target answer corresponding with the target problem.
In one possible implementation, the determining module 13 is specifically used for:
It obtains the first product of the confidence level of long answer type probability and long answer text, short answer type probability and short answers
Second product of the confidence level of case text;Judge whether first product is greater than second product;
If so, the long answer text is determined as target answer corresponding with the target problem;
If it is not, the short answer text is then determined as target answer corresponding with the target problem.
In one possible implementation, described device further include: splicing module;
The splicing, for asking the target after the preset target problem of the acquisition and text to be read
Topic and the text to be read are spliced, wherein in splicing, addition characterization is asked before the target problem
The separator of topic, the separator of addition characterization paragraph before the paragraph of the text to be read.
It should be noted that the aforementioned reading for being also applied for the embodiment to the explanation for reading understanding method embodiment
Understand device, details are not described herein again.
Reading provided in an embodiment of the present invention understands device, by obtaining preset target problem and text to be read;
Understand that model carries out understanding analysis to the text to be read according to preset reading, generates the answer class of the target problem
Type probability, answer text and corresponding confidence level;According to the answer type probability, the answer text and corresponding set
Reliability determines target answer corresponding with the target problem.To based on it is preset reading understand model to target problem and to
Long answer text or short answer text that text better understands out high quality are read, all different practical situations can show good
It is good, it is only good to a kind of understanding effect of the answer of answer type and logical compared to existing to be no longer limited to existing special purpose model
The understanding effect of the answer to different answer types is improved with model.
Fig. 6 is the structural schematic diagram that another reading provided in an embodiment of the present invention understands device.In conjunction with reference Fig. 6,
On the basis of embodiment illustrated in fig. 5, described device further include: training module 14;
The acquisition module 11, is also used to obtain first sample set, wherein the first sample packet that the first sample is concentrated
Include starting position and end of the correct option in reading material sample of training problem and reading material sample, long answer type
Position, the labeled data that the answer type for characterizing the first sample is long answer;
The acquisition module 11, is also used to obtain the second sample set, wherein the second sample packet in second sample set
Include starting position and end of the correct option in reading material sample of training problem and reading material sample, short answer type
Position, the labeled data that the answer type for characterizing second sample is short answer;
The training module 14, for the reading reason initial based on the first sample set and second sample set training
Model is solved, the preset reading is obtained and understands model.
In one possible implementation, the initial reading understands that model includes at least encoder, the first prediction
Layer, the second prediction interval, classifier, the training module 14 are specifically used for:
By the first sample and second sample stated in the second sample set point that the first sample is concentrated
It is not input in the encoder and is encoded;
Using first sample training first prediction interval after each coding, and using described after each coding
After the first sample set and each coding after trained second prediction interval of second sample set and each coding of use
The second sample set training classifier understands model to obtain the preset reading;
Wherein, the first prediction interval that training finishes can be predicted to need the problem of answering corresponding long answer text and its be set
Reliability, the second prediction interval that training finishes can be predicted to need the problem of answering corresponding short answer text and its confidence level, be instructed
The classifier that white silk finishes, which can differentiate, needs the problem of answering corresponding answer type probability.
It should be noted that the aforementioned reading for being also applied for the embodiment to the explanation for reading understanding method embodiment
Understand device, details are not described herein again.
Reading provided in an embodiment of the present invention understands device, compared with existing reading understands model, preset reading reason
Solving model not only includes the first prediction interval that can predict long answer in model structure, can predict the second prediction of short answer
Layer, further includes the classifier that can differentiate answer type, since the first prediction interval and the second prediction interval are to separate training, for
Different answer types corresponds to different model structure and parameter, therefore the preset reading understands model for different practical situations
Reading understand that scene can show well flexibility, better effect, pass through joint training, model learning is to preferably special
Sign, universal performance is more preferable, be no longer limited to existing special purpose model only it is good to a kind of understanding effect of the answer of answer type, very
Solves the collision problem between long answer model and short answer model well, and existing general compared to common model parameter
Model improves the understanding effect of the answer to different answer types.
Fig. 7 is a kind of structural schematic diagram of computer equipment provided in an embodiment of the present invention.The computer equipment includes:
Memory 1001, processor 1002 and it is stored in the calculating that can be run on memory 1001 and on processor 1002
Machine program.
Processor 1002 realizes the reading understanding method provided in above-described embodiment when executing described program.
Further, computer equipment further include:
Communication interface 1003, for the communication between memory 1001 and processor 1002.
Memory 1001, for storing the computer program that can be run on processor 1002.
Memory 1001 may include high speed RAM memory, it is also possible to further include nonvolatile memory (non-
Volatile memory), a for example, at least magnetic disk storage.
Processor 1002 realizes reading understanding method described in above-described embodiment when for executing described program.
If memory 1001, processor 1002 and the independent realization of communication interface 1003, communication interface 1003, memory
1001 and processor 1002 can be connected with each other by bus and complete mutual communication.The bus can be industrial standard
Architecture (Industry Standard Architecture, referred to as ISA) bus, external equipment interconnection
(Peripheral Component, referred to as PCI) bus or extended industry-standard architecture (Extended Industry
Standard Architecture, referred to as EISA) bus etc..The bus can be divided into address bus, data/address bus, control
Bus processed etc..Only to be indicated with a thick line in Fig. 7, it is not intended that an only bus or a type of convenient for indicating
Bus.
Optionally, in specific implementation, if memory 1001, processor 1002 and communication interface 1003, are integrated in one
It is realized on block chip, then memory 1001, processor 1002 and communication interface 1003 can be completed mutual by internal interface
Communication.
Processor 1002 may be a central processing unit (Central Processing Unit, referred to as CPU), or
Person is specific integrated circuit (Application Specific Integrated Circuit, referred to as ASIC) or quilt
It is configured to implement one or more integrated circuits of the embodiment of the present invention.
The present embodiment also provides a kind of computer readable storage medium, is stored thereon with computer program, which is characterized in that
Reading understanding method as described above is realized when the program is executed by processor.
In the description of this specification, reference term " one embodiment ", " some embodiments ", " example ", " specifically show
The description of example " or " some examples " etc. means specific features, structure, material or spy described in conjunction with this embodiment or example
Point is included at least one embodiment or example of the invention.In the present specification, schematic expression of the above terms are not
It must be directed to identical embodiment or example.Moreover, particular features, structures, materials, or characteristics described can be in office
It can be combined in any suitable manner in one or more embodiment or examples.In addition, without conflicting with each other, the skill of this field
Art personnel can tie the feature of different embodiments or examples described in this specification and different embodiments or examples
It closes and combines.
In addition, term " first ", " second " are used for descriptive purposes only and cannot be understood as indicating or suggesting relative importance
Or implicitly indicate the quantity of indicated technical characteristic.Define " first " as a result, the feature of " second " can be expressed or
Implicitly include at least one this feature.In the description of the present invention, the meaning of " plurality " is at least two, such as two, three
It is a etc., unless otherwise specifically defined.
Any process described otherwise above or method description are construed as in flow chart or herein, and expression includes
It is one or more for realizing custom logic function or process the step of executable instruction code module, segment or portion
Point, and the range of the preferred embodiment of the present invention includes other realization, wherein can not press shown or discussed suitable
Sequence, including according to related function by it is basic simultaneously in the way of or in the opposite order, Lai Zhihang function, this should be of the invention
Embodiment person of ordinary skill in the field understood.
Expression or logic and/or step described otherwise above herein in flow charts, for example, being considered use
In the order list for the executable instruction for realizing logic function, may be embodied in any computer-readable medium, for
Instruction execution system, device or equipment (such as computer based system, including the system of processor or other can be held from instruction
The instruction fetch of row system, device or equipment and the system executed instruction) it uses, or combine these instruction execution systems, device or set
It is standby and use.For the purpose of this specification, " computer-readable medium ", which can be, any may include, stores, communicates, propagates or pass
Defeated program is for instruction execution system, device or equipment or the dress used in conjunction with these instruction execution systems, device or equipment
It sets.The more specific example (non-exhaustive list) of computer-readable medium include the following: there is the electricity of one or more wirings
Interconnecting piece (electronic device), portable computer diskette box (magnetic device), random access memory (RAM), read-only memory
(ROM), erasable edit read-only storage (EPROM or flash memory), fiber device and portable optic disk is read-only deposits
Reservoir (CDROM).In addition, computer-readable medium can even is that the paper that can print described program on it or other are suitable
Medium, because can then be edited, be interpreted or when necessary with it for example by carrying out optical scanner to paper or other media
His suitable method is handled electronically to obtain described program, is then stored in computer storage.
It should be appreciated that each section of the invention can be realized with hardware, software, firmware or their combination.Above-mentioned
In embodiment, software that multiple steps or method can be executed in memory and by suitable instruction execution system with storage
Or firmware is realized.Such as, if realized with hardware in another embodiment, following skill well known in the art can be used
Any one of art or their combination are realized: have for data-signal is realized the logic gates of logic function from
Logic circuit is dissipated, the specific integrated circuit with suitable combinational logic gate circuit, programmable gate array (PGA), scene can compile
Journey gate array (FPGA) etc..
Those skilled in the art are understood that realize all or part of step that above-described embodiment method carries
It suddenly is that relevant hardware can be instructed to complete by program, the program can store in a kind of computer-readable storage medium
In matter, which when being executed, includes the steps that one or a combination set of embodiment of the method.
It, can also be in addition, each functional unit in each embodiment of the present invention can integrate in a processing module
It is that each unit physically exists alone, can also be integrated in two or more units in a module.Above-mentioned integrated mould
Block both can take the form of hardware realization, can also be realized in the form of software function module.The integrated module is such as
Fruit is realized and when sold or used as an independent product in the form of software function module, also can store in a computer
In read/write memory medium.
Storage medium mentioned above can be read-only memory, disk or CD etc..Although having been shown and retouching above
The embodiment of the present invention is stated, it is to be understood that above-described embodiment is exemplary, and should not be understood as to limit of the invention
System, those skilled in the art can be changed above-described embodiment, modify, replace and become within the scope of the invention
Type.
Claims (12)
1. a kind of reading understanding method characterized by comprising
Obtain preset target problem and text to be read;
Understand that model carries out understanding analysis to the text to be read according to preset reading, generates answering for the target problem
Case type probability, answer text and corresponding confidence level;
It is corresponding with the target problem according to the determination of the answer type probability, the answer text and corresponding confidence level
Target answer.
2. the method according to claim 1, wherein it is described according to the answer text type probability, described answer
Case text and corresponding confidence level determine that target answer corresponding with the target problem includes:
Obtain the first product, short answer type probability and the short answer text of the confidence level of long answer type probability and long answer text
Second product of this confidence level;
Judge whether first product is greater than second product;
If so, the long answer text is determined as target answer corresponding with the target problem;
If it is not, the short answer text is then determined as target answer corresponding with the target problem.
3. the method according to claim 1, wherein in the preset target problem of the acquisition and text to be read
Before this, further includes:
Obtain first sample set, wherein the first sample that the first sample is concentrated include training problem and reading material sample,
Starting position and end position of the correct option of long answer type in reading material sample, for characterizing the first sample
Answer type be long answer labeled data;
Obtain the second sample set, wherein the second sample in second sample set include training problem and reading material sample,
Starting position and end position of the correct option of short answer type in reading material sample, for characterizing second sample
Answer type be short answer labeled data;
Model is understood based on the first sample set and the initial reading of second sample set training, obtains described preset read
Reading understands model.
4. according to the method described in claim 3, it is characterized in that, the initial reading understands that model includes at least coding
Device, the first prediction interval, the second prediction interval, classifier, it is described first based on the first sample set and second sample set training
The reading of beginning understands model, obtains the preset reading and understands that model includes:
The first sample that the first sample is concentrated and second sample difference stated in the second sample set is defeated
Enter into the encoder and is encoded;
Using first sample training first prediction interval after each coding, and using described second after each coding
It is described after the first sample set and each coding after trained second prediction interval of sample set and each coding of use
The second sample set training classifier understands model to obtain the preset reading;
Wherein, the first prediction interval that training finishes can be predicted to need the problem of answering corresponding long answer text and its confidence
Degree, the second prediction interval that training finishes can be predicted to need the problem of answering corresponding short answer text and its confidence level, be trained
The classifier finished, which can differentiate, needs the problem of answering corresponding answer type probability.
5. method according to any one of claims 1 to 4, which is characterized in that it is described obtain preset target problem and
After text to be read, further includes:
The target problem and the text to be read are spliced, wherein in splicing, in the target problem
Before add characterization problems separator, before the paragraph of the text to be read addition characterization paragraph separation
Symbol.
6. a kind of reading understands device characterized by comprising
Module is obtained, for obtaining preset target problem and text to be read;
Generation module generates institute for understanding that model carries out understanding analysis to the text to be read according to preset reading
State the answer type probability, answer text and corresponding confidence level of target problem;
Determining module, for according to the answer type probability, the answer text and corresponding confidence level it is determining with it is described
The corresponding target answer of target problem.
7. device according to claim 6, which is characterized in that the determining module is specifically used for:
Obtain the first product, short answer type probability and the short answer text of the confidence level of long answer type probability and long answer text
Second product of this confidence level;Judge whether first product is greater than second product;
If so, the long answer text is determined as target answer corresponding with the target problem;
If it is not, the short answer text is then determined as target answer corresponding with the target problem.
8. device according to claim 6, which is characterized in that further include: training module;
The acquisition module, is also used to obtain first sample set, wherein the first sample that the first sample is concentrated includes training
Problem and reading material sample, long answer type starting position and end position of the correct option in reading material sample,
Answer type for characterizing the first sample is the labeled data of long answer;
The acquisition module, is also used to obtain the second sample set, wherein the second sample in second sample set includes training
Problem and reading material sample, short answer type starting position and end position of the correct option in reading material sample,
Answer type for characterizing second sample is the labeled data of short answer;
The training module, for understanding mould based on the initial reading of the first sample set and second sample set training
Type obtains the preset reading and understands model.
9. device according to claim 8, which is characterized in that the initial reading understands that model includes at least coding
Device, the first prediction interval, the second prediction interval, classifier, the training module are specifically used for:
The first sample that the first sample is concentrated and second sample difference stated in the second sample set is defeated
Enter into the encoder and is encoded;
Using first sample training first prediction interval after each coding, and using described second after each coding
It is described after the first sample set and each coding after trained second prediction interval of sample set and each coding of use
The second sample set training classifier understands model to obtain the preset reading;
Wherein, the first prediction interval that training finishes can be predicted to need the problem of answering corresponding long answer text and its confidence
Degree, the second prediction interval that training finishes can be predicted to need the problem of answering corresponding short answer text and its confidence level, be trained
The classifier finished, which can differentiate, needs the problem of answering corresponding answer type probability.
10. according to the described in any item devices of claim 6 to 8, which is characterized in that further include: splicing module;
The splicing, for after the preset target problem of the acquisition and text to be read, to the target problem and
The text to be read is spliced, wherein in splicing, characterization problems are added before the target problem
Separator, the separator of addition characterization paragraph before the paragraph of the text to be read.
11. a kind of computer equipment characterized by comprising
Memory, processor and storage are on a memory and the computer program that can run on a processor, which is characterized in that institute
It states when processor executes described program and realizes such as reading understanding method as claimed in any one of claims 1 to 5.
12. a kind of computer readable storage medium, is stored thereon with computer program, which is characterized in that the program is by processor
Such as reading understanding method as claimed in any one of claims 1 to 5 is realized when execution.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910429805.0A CN110210021B (en) | 2019-05-22 | 2019-05-22 | Reading understanding method and device |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910429805.0A CN110210021B (en) | 2019-05-22 | 2019-05-22 | Reading understanding method and device |
Publications (2)
Publication Number | Publication Date |
---|---|
CN110210021A true CN110210021A (en) | 2019-09-06 |
CN110210021B CN110210021B (en) | 2021-05-28 |
Family
ID=67788119
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201910429805.0A Active CN110210021B (en) | 2019-05-22 | 2019-05-22 | Reading understanding method and device |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN110210021B (en) |
Cited By (12)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110955755A (en) * | 2019-11-29 | 2020-04-03 | 支付宝(杭州)信息技术有限公司 | Method and system for determining target standard information |
CN111428467A (en) * | 2020-02-19 | 2020-07-17 | 平安科技(深圳)有限公司 | Method, device, equipment and storage medium for generating reading comprehension question topic |
CN111858878A (en) * | 2020-06-18 | 2020-10-30 | 达而观信息科技(上海)有限公司 | Method, system and storage medium for automatically extracting answer from natural language text |
CN112215004A (en) * | 2020-09-04 | 2021-01-12 | 中国电子科技集团公司第二十八研究所 | Application method in extraction of text entities of military equipment based on transfer learning |
CN112507090A (en) * | 2020-11-30 | 2021-03-16 | 北京百度网讯科技有限公司 | Method, apparatus, device and storage medium for outputting information |
CN112509690A (en) * | 2020-11-30 | 2021-03-16 | 北京百度网讯科技有限公司 | Method, apparatus, device and storage medium for controlling quality |
WO2021135910A1 (en) * | 2020-06-24 | 2021-07-08 | 平安科技(深圳)有限公司 | Machine reading comprehension-based information extraction method and related device |
CN113111232A (en) * | 2020-02-13 | 2021-07-13 | 北京明亿科技有限公司 | Regular expression-based alarm receiving and processing text address extraction method and device |
CN113657089A (en) * | 2021-08-20 | 2021-11-16 | 西安电子科技大学 | English reading understanding auxiliary question setting method and system |
CN113792121A (en) * | 2021-04-08 | 2021-12-14 | 北京金山数字娱乐科技有限公司 | Reading understanding model training method and device and reading understanding method and device |
CN113935305A (en) * | 2021-08-10 | 2022-01-14 | 上海明略人工智能(集团)有限公司 | Multi-stage and multi-task text reading understanding method and device |
CN114490990A (en) * | 2022-01-28 | 2022-05-13 | 北京百度网讯科技有限公司 | Method, device and equipment for determining text to be marked and storage medium |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104471568A (en) * | 2012-07-02 | 2015-03-25 | 微软公司 | Learning-based processing of natural language questions |
US20180260382A1 (en) * | 2017-03-09 | 2018-09-13 | International Business Machines Corporation | Domain-specific method for distinguishing type-denoting domain terms from entity-denoting domain terms |
CN108846063A (en) * | 2018-06-04 | 2018-11-20 | 北京百度网讯科技有限公司 | Determine the method, apparatus, equipment and computer-readable medium of problem answers |
CN109344234A (en) * | 2018-09-06 | 2019-02-15 | 和美(深圳)信息技术股份有限公司 | Machine reads understanding method, device, computer equipment and storage medium |
-
2019
- 2019-05-22 CN CN201910429805.0A patent/CN110210021B/en active Active
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104471568A (en) * | 2012-07-02 | 2015-03-25 | 微软公司 | Learning-based processing of natural language questions |
US20180260382A1 (en) * | 2017-03-09 | 2018-09-13 | International Business Machines Corporation | Domain-specific method for distinguishing type-denoting domain terms from entity-denoting domain terms |
CN108846063A (en) * | 2018-06-04 | 2018-11-20 | 北京百度网讯科技有限公司 | Determine the method, apparatus, equipment and computer-readable medium of problem answers |
CN109344234A (en) * | 2018-09-06 | 2019-02-15 | 和美(深圳)信息技术股份有限公司 | Machine reads understanding method, device, computer equipment and storage medium |
Cited By (19)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110955755A (en) * | 2019-11-29 | 2020-04-03 | 支付宝(杭州)信息技术有限公司 | Method and system for determining target standard information |
CN113111232A (en) * | 2020-02-13 | 2021-07-13 | 北京明亿科技有限公司 | Regular expression-based alarm receiving and processing text address extraction method and device |
CN111428467A (en) * | 2020-02-19 | 2020-07-17 | 平安科技(深圳)有限公司 | Method, device, equipment and storage medium for generating reading comprehension question topic |
CN111428467B (en) * | 2020-02-19 | 2024-05-07 | 平安科技(深圳)有限公司 | Method, device, equipment and storage medium for generating problem questions for reading and understanding |
CN111858878A (en) * | 2020-06-18 | 2020-10-30 | 达而观信息科技(上海)有限公司 | Method, system and storage medium for automatically extracting answer from natural language text |
CN111858878B (en) * | 2020-06-18 | 2023-12-22 | 达观数据有限公司 | Method, system and storage medium for automatically extracting answer from natural language text |
WO2021135910A1 (en) * | 2020-06-24 | 2021-07-08 | 平安科技(深圳)有限公司 | Machine reading comprehension-based information extraction method and related device |
CN112215004B (en) * | 2020-09-04 | 2023-05-02 | 中国电子科技集团公司第二十八研究所 | Application method of transfer learning-based text entity extraction in military equipment |
CN112215004A (en) * | 2020-09-04 | 2021-01-12 | 中国电子科技集团公司第二十八研究所 | Application method in extraction of text entities of military equipment based on transfer learning |
CN112507090A (en) * | 2020-11-30 | 2021-03-16 | 北京百度网讯科技有限公司 | Method, apparatus, device and storage medium for outputting information |
CN112509690B (en) * | 2020-11-30 | 2023-08-04 | 北京百度网讯科技有限公司 | Method, apparatus, device and storage medium for controlling quality |
CN112507090B (en) * | 2020-11-30 | 2024-05-03 | 北京百度网讯科技有限公司 | Method, apparatus, device and storage medium for outputting information |
CN112509690A (en) * | 2020-11-30 | 2021-03-16 | 北京百度网讯科技有限公司 | Method, apparatus, device and storage medium for controlling quality |
CN113792121A (en) * | 2021-04-08 | 2021-12-14 | 北京金山数字娱乐科技有限公司 | Reading understanding model training method and device and reading understanding method and device |
CN113792121B (en) * | 2021-04-08 | 2023-09-22 | 北京金山数字娱乐科技有限公司 | Training method and device of reading and understanding model, reading and understanding method and device |
CN113935305A (en) * | 2021-08-10 | 2022-01-14 | 上海明略人工智能(集团)有限公司 | Multi-stage and multi-task text reading understanding method and device |
CN113657089A (en) * | 2021-08-20 | 2021-11-16 | 西安电子科技大学 | English reading understanding auxiliary question setting method and system |
CN114490990A (en) * | 2022-01-28 | 2022-05-13 | 北京百度网讯科技有限公司 | Method, device and equipment for determining text to be marked and storage medium |
CN114490990B (en) * | 2022-01-28 | 2023-05-23 | 北京百度网讯科技有限公司 | Method, device, equipment and storage medium for determining text to be annotated |
Also Published As
Publication number | Publication date |
---|---|
CN110210021B (en) | 2021-05-28 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN110210021A (en) | Read understanding method and device | |
CN110188350A (en) | Text coherence calculation method and device | |
CN106844341A (en) | News in brief extracting method and device based on artificial intelligence | |
CN107844417A (en) | Method for generating test case and device | |
CN110188362A (en) | Text handling method and device | |
CN109543022A (en) | Text error correction method and device | |
CN114943789A (en) | Image processing method, model training method and related device | |
CN112183747A (en) | Neural network training method, neural network compression method and related equipment | |
CN109599095A (en) | A kind of mask method of voice data, device, equipment and computer storage medium | |
CN109670491A (en) | Identify method, apparatus, equipment and the storage medium of facial image | |
JP2023509405A (en) | Translation method, device, electronic device and computer program | |
CN107122492A (en) | Lyric generation method and device based on picture content | |
US20180018392A1 (en) | Topic identification based on functional summarization | |
CN109840280A (en) | A kind of file classification method, device and computer readable storage medium | |
CN109800292A (en) | The determination method, device and equipment of question and answer matching degree | |
CN109461438A (en) | A kind of audio recognition method, device, equipment and storage medium | |
CN110187760A (en) | Intelligent interactive method and device | |
CN107193941A (en) | Story generation method and device based on picture content | |
JPWO2011018867A1 (en) | Information classification apparatus, information classification method, and program | |
CN110232117A (en) | Sentence fluency detection method, device and terminal | |
CN107748801A (en) | News recommends method, apparatus, terminal device and computer-readable recording medium | |
CN111680132A (en) | Noise filtering and automatic classifying method for internet text information | |
CN110188327A (en) | Text goes colloquial style method and device | |
CN111126566A (en) | Abnormal furniture layout data detection method based on GAN model | |
CN113610080B (en) | Cross-modal perception-based sensitive image identification method, device, equipment and medium |
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 | ||
GR01 | Patent grant |