CN112559713B - Text relevance judging method and device, model, electronic equipment and readable medium - Google Patents
Text relevance judging method and device, model, electronic equipment and readable medium Download PDFInfo
- Publication number
- CN112559713B CN112559713B CN202011548198.9A CN202011548198A CN112559713B CN 112559713 B CN112559713 B CN 112559713B CN 202011548198 A CN202011548198 A CN 202011548198A CN 112559713 B CN112559713 B CN 112559713B
- Authority
- CN
- China
- Prior art keywords
- text
- vector
- semantic
- dependency
- vectors
- 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.)
- Active
Links
- 238000000034 method Methods 0.000 title claims abstract description 72
- 239000013598 vector Substances 0.000 claims abstract description 296
- 230000004927 fusion Effects 0.000 claims abstract description 39
- 238000013528 artificial neural network Methods 0.000 claims abstract description 32
- 238000012545 processing Methods 0.000 claims description 50
- 238000011176 pooling Methods 0.000 claims description 13
- 238000004458 analytical method Methods 0.000 claims description 7
- 238000005096 rolling process Methods 0.000 claims description 4
- 238000013473 artificial intelligence Methods 0.000 abstract description 6
- 238000004590 computer program Methods 0.000 description 11
- 238000004891 communication Methods 0.000 description 10
- 238000005516 engineering process Methods 0.000 description 9
- 230000008569 process Effects 0.000 description 9
- 238000010586 diagram Methods 0.000 description 8
- 230000008859 change Effects 0.000 description 6
- 238000004422 calculation algorithm Methods 0.000 description 5
- 230000006835 compression Effects 0.000 description 4
- 238000007906 compression Methods 0.000 description 4
- 238000013527 convolutional neural network Methods 0.000 description 4
- 230000006870 function Effects 0.000 description 4
- 238000007499 fusion processing Methods 0.000 description 3
- 230000002452 interceptive effect Effects 0.000 description 3
- 238000012986 modification Methods 0.000 description 3
- 230000004048 modification Effects 0.000 description 3
- 230000003287 optical effect Effects 0.000 description 3
- 238000009412 basement excavation Methods 0.000 description 2
- 238000010276 construction Methods 0.000 description 2
- 235000013399 edible fruits Nutrition 0.000 description 2
- 230000003993 interaction Effects 0.000 description 2
- 238000010801 machine learning Methods 0.000 description 2
- 239000011159 matrix material Substances 0.000 description 2
- 230000000306 recurrent effect Effects 0.000 description 2
- 238000003491 array Methods 0.000 description 1
- 230000006399 behavior Effects 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 230000001413 cellular effect Effects 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 239000000835 fiber Substances 0.000 description 1
- 230000014509 gene expression Effects 0.000 description 1
- 239000004973 liquid crystal related substance Substances 0.000 description 1
- 238000003058 natural language processing Methods 0.000 description 1
- 239000013307 optical fiber Substances 0.000 description 1
- 230000000717 retained effect Effects 0.000 description 1
- 239000004065 semiconductor Substances 0.000 description 1
- 230000001953 sensory effect Effects 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 230000000007 visual effect Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
- G06F16/33—Querying
- G06F16/332—Query formulation
- G06F16/3329—Natural language query formulation or dialogue systems
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
- G06F16/33—Querying
- G06F16/3331—Query processing
- G06F16/334—Query execution
- G06F16/3344—Query execution using natural language analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
- G06F16/35—Clustering; Classification
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/25—Fusion techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F40/00—Handling natural language data
- G06F40/30—Semantic analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Engineering & Computer Science (AREA)
- Artificial Intelligence (AREA)
- Computational Linguistics (AREA)
- Mathematical Physics (AREA)
- Databases & Information Systems (AREA)
- Life Sciences & Earth Sciences (AREA)
- Evolutionary Computation (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Audiology, Speech & Language Pathology (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Human Computer Interaction (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Software Systems (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Machine Translation (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The disclosure provides a text relevance judging method and device, a model, electronic equipment and a readable medium, and relates to the technical field of computers and artificial intelligence. The specific implementation scheme is as follows: determining a semantic role vector and a dependency vector in the first text, and a semantic role vector and a dependency vector in the second text; respectively carrying out vector fusion on the semantic role vector and the dependency relationship vector in the first text and the semantic role vector and the dependency relationship vector in the second text, and carrying out vector fusion on the vector fusion results obtained respectively again to obtain fused vectors of the first text and the second text; classifying the fused vectors through a preset neural network, and determining the correlation between the first text and the second text according to the classification result. According to the scheme of the disclosure, the answer questions and questions provided in the question and answer service can be identified.
Description
Technical Field
The disclosure relates to the technical field of computers and artificial intelligence, in particular to a text relevance judging method and device, a model, electronic equipment and a readable medium.
Background
With the development of network technology, more and more users acquire information through a question-answer service provided by a network. For example, a user may ask questions on a platform that provides question-answering services according to his own needs, and other users on the platform may answer the questions.
The quality of answers provided by the question-answer service varies greatly due to the openness of the network. Some answers can help the questioner obtain information, and some answers cannot meet the requirement of the questioner, namely, the questioner can answer the questions, so that the contents of the questions and the answers provided in the questioning and answering service need to be identified.
Disclosure of Invention
A text relevance judging method and device, a model, electronic equipment and a readable medium are provided.
According to a first aspect, there is provided a correlation determination method, including: determining a semantic role vector and a dependency vector in the first text, and a semantic role vector and a dependency vector in the second text; respectively carrying out vector fusion on the semantic role vector and the dependency relationship vector in the first text and the semantic role vector and the dependency relationship vector in the second text, and carrying out vector fusion on the vector fusion results obtained respectively again to obtain the fused vectors of the first text and the second text; classifying the fused vectors through a preset neural network, and determining the correlation between the first text and the second text according to the classification result.
According to a second aspect, there is provided a correlation determination apparatus including: the vector determining module is used for determining semantic role vectors and dependency vectors in the first text and semantic role vectors and dependency vectors in the second text; the vector fusion module is used for carrying out vector fusion on the semantic role vector and the dependency relationship vector in the first text and the semantic role vector and the dependency relationship vector in the second text respectively, and carrying out vector fusion on the vector fusion results obtained respectively again to obtain fused vectors of the first text and the second text; the classification determining module is used for classifying the fused vectors through a preset neural network and determining the correlation between the first text and the second text according to the classification result.
According to a third aspect, a text relevance determining model is provided, which is configured to execute any one of the relevance determining methods according to the received first text and second text, so as to obtain a relevance determining result of the first text and the second text.
According to a fourth aspect, there is provided an electronic device comprising: at least one processor; and a memory communicatively coupled to the at least one processor; the memory stores instructions executable by the at least one processor to enable the at least one processor to perform any one of the correlation determination methods described above.
According to a fifth aspect, there is provided a non-transitory computer-readable storage medium storing computer instructions for causing a computer to execute any one of the above-described correlation determination methods.
According to a sixth aspect, there is provided a computer program product comprising a computer program which, when executed by a processor, implements any of the above-mentioned correlation determination methods.
It should be understood that the description in this section is not intended to identify key or critical features of the embodiments of the disclosure, nor is it intended to be used to limit the scope of the disclosure. Other features of the present disclosure will become apparent from the following specification.
Drawings
The drawings are for a better understanding of the present solution and are not to be construed as limiting the present disclosure. Wherein:
FIG. 1 is a schematic illustration of a scenario provided by an embodiment of the present disclosure;
FIG. 2 is a flowchart of a text relevance determination method provided by an embodiment of the present disclosure;
fig. 3 is a schematic diagram of a model of a text relevance determining method according to an embodiment of the present disclosure
FIG. 4 is a flow chart of a text relevance determination method provided by an exemplary embodiment of the present disclosure;
fig. 5 is a schematic structural diagram of a text relevance determining apparatus according to an embodiment of the present disclosure;
Fig. 6 is a block diagram of an electronic device for implementing a text relevance determining method of an embodiment of the present disclosure.
Detailed Description
Exemplary embodiments of the present disclosure are described below in conjunction with the accompanying drawings, which include various details of the embodiments of the present disclosure to facilitate understanding, and should be considered as merely exemplary. Accordingly, one of ordinary skill in the art will recognize that various changes and modifications of the embodiments described herein can be made without departing from the scope and spirit of the present disclosure. Also, descriptions of well-known functions and constructions are omitted in the following description for clarity and conciseness.
Embodiments of the disclosure and features of embodiments may be combined with each other without conflict.
As used herein, the term "and/or" includes any and all combinations of one or more of the associated listed items.
The terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of the disclosure. As used herein, the singular forms "a," "an," and "the" are intended to include the plural forms as well, unless the context clearly indicates otherwise. It will be further understood that the terms "comprises" and/or "comprising," when used in this specification, specify the presence of stated features, integers, steps, operations, elements, and/or components, but do not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and/or groups thereof.
In implementations of the present disclosure, the question-answering service may be, for example, a question-answering service provided through a question-answering service platform such as a community question-answering platform, a website question-answering platform, and an interactive question-answering platform. Taking a community question and answer platform as an example, the community question and answer platform allows users to present questions according to own needs and allows users to collaborate to edit and answer the questions, so that an interactive community with knowledge attributes is formed.
The mode of the interactive community provides a channel for users to acquire knowledge, but because of the openness of the community, the answer quality of the question-answer community is large, some answers can help the questioner acquire information, and some answers can not meet the requirements of the questioner, namely the questioner can answer the question. Thus, the quality difference of the answer content is a major problem to be solved in the community question-and-answer process.
Low quality answers can be identified, typically, in two ways, one based on a similarity algorithm and the other based on a vocabulary.
When the identification is performed based on a similarity algorithm, the question-answer content can be converted into corresponding feature vectors, the direct distance between the question-answer vectors is calculated through cosine similarity between the questions and the answers, and the semantic relevance between the questions and the answers is determined according to the distance between the vectors, so that answer content with lower semantic relevance is obtained. However, the application range of the recognition based on the similarity algorithm is limited, and the recognition method is only used for a question and answer content recognition scene with low correlation, and the accurate recognition cannot be performed for the situation with high correlation of question and answer semantics. By way of example, the question is "apple is bad? When the answer is "apple is a fruit", the questions and the answers are all describing the apple, the similarity calculated based on the similarity algorithm is extremely high, and the contents of the questions and answers cannot be successfully identified.
When recognition is performed based on the vocabulary, whether the answer is a question of a non-answer question is judged by detecting whether the answer hits a preset low-quality vocabulary, and the judging effect of the answer is greatly related to the magnitude of the low-quality vocabulary. Therefore, the recognition method based on the vocabulary is only suitable for low-quality contents of the hit preset feature words, the missed contents cannot be provided with judging capability, and meanwhile, the arrangement of the low-quality vocabulary also has great manual workload.
Fig. 1 is a schematic view of a scenario of an embodiment of the present disclosure. In the scenario shown in fig. 1, it comprises: a terminal 11, a question and answer service platform 12, a network 13 and a text processing device 14.
The terminal 11 may include, but is not limited to: personal computers, smart phones, tablet computers, personal digital assistants, servers, etc. The user may input the questioning contents to the questioning and answering service platform 12 through the terminal 11.
The question-answering service platform 12 may be implemented as a web platform or an application providing a question-answering service, and if implemented as an application providing a question-answering service, may be installed in the terminal 11. The question and answer service platform 12 may obtain answer contents of the question contents by searching in the network 13 according to the received question contents; alternatively, the answer content of the questioning content may be obtained by collecting other user edits and answers to the questioning content in the network 13.
The network 13 is used as a medium to provide communications links between various platforms and electronic devices. In particular, the network 13 may include various connection types, such as wired, wireless communication links, or fiber optic cables, etc.
As shown in fig. 1, in one embodiment, the question and answer service platform 12 may input the received question content and the obtained answer content to the text processing device 14; the text processing device 14 performs the correlation judgment method of the embodiment of the present disclosure based on the question contents and the answer contents received from the question-answer service platform 12 to determine whether the answer contents answer a question.
In another embodiment, after receiving the answer content provided by the question and answer service platform 12, the terminal 11 may also input the question content and the received answer content to the text processing device 14, where the text processing device 14 executes the relevance determining method according to the embodiment of the present disclosure according to the received question content and answer content from the user terminal 11, so as to determine whether the answer content is not answered.
It should be understood that the number of devices in fig. 1 is merely illustrative. According to the practical application needs, flexible adjustment can be performed. For example, the text processing device 14 may be one service device or may be a server cluster including a plurality of service devices. And the configuration can be flexibly carried out according to the requirements, and the content is not limited in this respect.
Fig. 2 is a flowchart of a text relevance determining method of an embodiment of the present disclosure.
In a first aspect, referring to fig. 1, an embodiment of the present disclosure provides a text relevance determining method, which may include the following steps.
S110, determining a semantic role vector and a dependency vector in the first text, and a semantic role vector and a dependency vector in the second text.
S120, respectively carrying out vector fusion on the semantic role vector and the dependency relationship vector in the first text and the semantic role vector and the dependency relationship vector in the second text, and carrying out vector fusion on the respectively obtained vector fusion results again to obtain the fused vectors of the first text and the second text.
S130, classifying the fused vectors through a preset neural network, and determining the correlation between the first text and the second text according to the classification result.
According to the text relevance judging method, the semantic role vector and the dependency relationship vector of the first text can be respectively subjected to vector fusion, the semantic role and the dependency relationship of the second text can be respectively subjected to vector fusion, the vector fusion results obtained respectively are subjected to vector fusion again to obtain the fused vectors between the first text and the second text, the fused vectors are classified by utilizing a preset neural network, and the relevance of the first text and the second text is determined according to the classification results.
According to the text relevance judging method disclosed by the embodiment of the disclosure, in the case that the first text and the second text are the question text and the answer text, the relevance of the questions and the answers can be judged, so that the non-answer questions can be identified.
In the method, compared with a low-quality recognition method based on a vocabulary, the embodiment of the disclosure is not limited to the preset vocabulary content, manual excavation and vocabulary arrangement are not needed, and the method can save labor cost and has wider application.
In the method, compared with a judgment method of semantic relativity, because semantic role vectors and dependency relation vectors of texts embody grammar structures of texts, the embodiment of the invention can acquire vectors of a first text and vectors of a second text through the angle of the grammar structures, and can remove the influence of similar word meanings on a similarity judgment result, so that the judgment result has stability; moreover, because the change space of the grammar structure is far smaller than the change space of the word meaning of the text, the text similarity judging method has wider universality; further, the feature space corresponding to the grammar structure is smaller, so that when the feature vector is processed through the neural network, the space occupied by the required module is smaller, the processing speed is high, and the recognition efficiency is high.
In some embodiments, the first text and the second text are text or strings having a length less than a predetermined length threshold.
In this embodiment, the shorter the text length, the fewer semantic roles in the text, the simpler the dependency relationship in the text, and the more accurate the determination result.
In some embodiments, the first text and the second text are question text and answer text.
In this embodiment, when the first text and the second text are the question text and the answer text, the semantic role vector and the dependency relationship vector of the question text can be fused by vectors based on the angle of the grammar structure, the semantic role and the dependency relationship of the answer text can be fused by vectors, the vector fusion result of the question text and the vector fusion result of the answer text can be fused again to obtain the fused vector, the fused vector is classified by using a preset neural network, the relevance between the question text and the answer text is determined according to the classification result, so that the question and the answer are matched completely based on the grammar structure, the relevance judgment result is not affected by the word sense, and the stability is provided; the grammar structure change space is far smaller than the change space of word meaning, so the grammar structure has wider universality in question-answer related judgment; the feature space corresponding to the grammar structures of the question text and the answer text is smaller, and when the feature vector is processed through the neural network, the space occupied by the required module is smaller, and the processing speed is faster.
In the embodiment of the present disclosure, the first text and the second text may include, but are not limited to, a question text and an answer text, and for any two texts needing to be subjected to relevance judgment, the text similarity judgment method of the embodiment of the present disclosure may be used to process the text, so that diversity and flexibility of processing data in text similarity judgment can be improved.
In some embodiments, prior to step S110, the method further comprises: s11, analyzing the semantic roles and the dependency relationships of the acquired first text and the acquired second text respectively to obtain the semantic roles and the dependency relationships of the first text and the semantic roles and the dependency relationships of the second text.
Illustratively, the semantic roles include one or more of subject, object, indirect object, predicate verb, mode, time, and non-semantic roles in text; the dependency relationship includes one or more of a master-predicate relationship, a parallel relationship, and a guest-move relationship.
In the embodiment, through analysis of semantic roles and dependency relationships, the semantic roles and dependency relationships of the first text and the semantic roles and dependency relationships of the second text are obtained, and a data basis is provided for subsequent text relevance judgment based on a grammar structure.
In some embodiments, step S110 may specifically include the following steps.
S21, respectively encoding the semantic roles and the dependency relationships in the first text and the semantic roles and the dependency relationships in the second text to obtain sparse feature codes corresponding to the semantic roles and the dependency relationships in the first text and sparse feature codes corresponding to the semantic roles and the dependency relationships in the second text.
S22, embedding the sparse feature codes corresponding to the semantic roles and the dependency relationships in the first text to obtain dense feature codes corresponding to the semantic roles and the dependency relationships in the second text.
S23, rolling and pooling processing is carried out on the corresponding dense feature codes, so that semantic role vectors and dependency vectors in the first text and semantic role vectors and dependency vectors in the second text are obtained.
In this embodiment, the neural network information compression processing step of S21-S23 is performed on the semantic roles and the dependencies in the first text and the semantic roles and the dependencies in the second text, respectively, to obtain the semantic role vector and the dependency vector in the first text and the semantic role vector and the dependency vector in the second text.
In step S21, the purpose of the encoding is to convert the semantic roles and dependencies in the first text into codes that the computer is able to recognize, and to convert the semantic roles and dependencies in the second text into codes that the computer is able to recognize. Illustratively, the encoding mode may include, but is not limited to, any one of one-hot (one-hot) encoding, binarization encoding, histogram encoding, and count encoding.
And obtaining sparse feature codes corresponding to the semantic roles and the dependencies in the first text and sparse feature codes corresponding to the semantic roles and the dependencies in the second text through feature codes.
In step S22, each sparse feature code generated in step S21 may be converted into a corresponding dense feature code through an Embedding (Embedding) process, thereby reducing the dimension of the feature space and the complexity of feature construction, and reducing the amount of computation.
In step S23, the orderly information mixing can be performed on each dense feature code generated in step S22 through convolution processing, and the redundant information is removed through pooling processing, so that the key information is retained. After convolution processing and pooling processing, semantic role vectors and dependency vectors in the first text and semantic role vectors and dependency vectors in the second text are obtained, and the original text expressions of the semantic roles and the dependency relations in the first text and the semantic roles and the dependency relations in the first text are converted into comprehensive feature vectors containing more information.
In some embodiments, step S120 may specifically include the following steps.
S31, carrying out vector addition on the semantic role vector and the dependency relationship vector in the first text to obtain a problem vector;
s32, carrying out vector addition on the semantic role vector and the dependency relationship vector in the second text to obtain an answer vector;
and S33, vector splicing is carried out on the question vector and the answer vector, and a fused vector of the first text and the second text is obtained.
In the step, vector addition is carried out on the semantic role vector and the dependency relationship vector in the same text to obtain a vector after text addition; and vector splicing is carried out on the added vectors of the two texts, so that a new vector fused in a splicing mode is obtained.
In some embodiments, the pre-set neural network includes a fully connected layer; step S130 may specifically include: s41, classifying the fused vectors through a full connection layer to obtain a classification result of the fused vectors; s42, determining that the first text is related to the second text when the classification result is a preset first value, and determining that the first text is not related to the second text when the classification result is a preset second value.
In this embodiment, the fused vector of the first text and the second text is input into a neural network for processing, where the neural network includes a full connection layer, and the full connection layer is in the neural network and can function as a classifier, so as to obtain a classification result of the fused vector, and determine the relevance and matching degree of the first text and the second text according to the value of the classification result.
In the text relativity judging method of the embodiment of the disclosure, whether the first text and the second text are related or not can be judged according to a grammar structure, the text can be disassembled according to the grammar structure through semantic roles such as a main predicate and the like and the dependency relationship among the main predicate and the second predicate, and information obtained after disassembly is input into a neural network for classification and identification; for example, when the above-mentioned problem is "apple is not delicious" and the answer is "apple is a fruit", the text relevance determining method according to the embodiment of the present disclosure can remove the influence of word meaning on the text relevance determining result, and obtain a more accurate determining result.
For a better understanding of the present disclosure, a text relevance determination process of an exemplary embodiment of the present disclosure is described below with reference to fig. 3 and 4.
Fig. 3 is a schematic diagram of a model of a text relevance determining method according to an embodiment of the present disclosure. As shown in fig. 3, the model structure of the embodiment of the present disclosure may include: an input module 310, a parse (parameter) module 320, an encoding module 330, an embedded processing module 340, a convolution and pooling module 350, a merge (Cancat) module 360, a fusion module 370, a full connection module 380, and an output module 390.
In fig. 3, the input module 310 is configured to receive a first text and a second text, where the first text is, for example, the question text "is blue in sky" and the second text is, for example, the answer text "not blue".
The parsing module 320 is configured to parse the question and the answer into semantic roles and dependency relationships, and parse a list of semantic roles (verbs, nouns, adjectives, etc.) and a list of dependency relationships (dynamic guest relationships, master predicate relationships, association structures, etc.) included in the question and the answer, respectively.
The encoding module 330 is configured to encode the semantic role list and the dependency list of the first text, and encode the semantic role list and the dependency list of the second text, so as to obtain a sparse feature code corresponding to the semantic role and the dependency in the first text and a sparse feature code corresponding to the semantic role and the dependency in the second text.
The embedding processing module 340 is configured to perform embedding processing on the sparse feature code corresponding to the semantic role and the dependency relationship in the first text and the sparse feature code corresponding to the semantic role and the dependency relationship in the second text, so as to obtain a dense feature code corresponding to the semantic role and the dependency relationship in the first text and a dense feature code corresponding to the semantic role and the dependency relationship in the second text.
Illustratively, when using one-hot for feature encoding, the sparse matrix formed by one-hot may be compressed into a dense matrix by an emmbedding operation.
The rolling and pooling module 350 is configured to generate a semantic role vector and a dependency vector of the first text and a semantic role vector and a dependency vector of the second text through one or more one-dimensional rolling and one-dimensional pooling compression.
The merging module 360 is configured to add the semantic role vector of the first text and the dependency relationship to obtain a vector of the first text, i.e. a problem vector; and the method is also used for carrying out vector addition on the semantic role vector and the dependency relationship of the second text to obtain a vector of the second text, namely an answer vector.
The fusion module 370 is configured to perform vector stitching on the vector of the first text and the vector of the second text, so as to obtain a fused vector of the first text and the second text, which may be, for example, recorded as a question & answer vector.
And the full connection module 380 is configured to process the fused vector of the first text and the second text to obtain a classification result of the fused vector.
And an output module 390, configured to output the classification result of the fused vector.
Through the model structure of the neural network in the embodiment of the disclosure, the input first text can be subjected to analysis of a grammar structure and the second text can be subjected to analysis of the grammar structure, respective analysis results are respectively subjected to coding and vectorization processing to obtain semantic role vectors and dependency vectors of the first text and semantic role vectors and dependency vectors of the second text, and the semantic roles and dependency relationships of the first text and the semantic roles and dependency relationships of the second text are subjected to vector fusion by a vector addition and vector splicing method, so that the fused vectors are classified by using a full connection layer, classification results of the fused vectors are obtained, and correlation of the first text and the second text is determined.
It should be understood that the neural network in the embodiments of the present disclosure may be a recurrent neural network (Recurrent Neural Network, RNN) or a convolutional neural network (Convolutional Neural Networks, CNN), or may be other neural networks other than RNN and CNN, and the user may select an appropriate neural network for processing according to actual needs, which is not specifically limited in the embodiments of the present disclosure.
Fig. 4 shows a flow diagram of a text relevance determining method according to an exemplary embodiment of the present disclosure. As shown in fig. 4, in the case where the first text and the second text are question text and answer text, the text relevance determining method may include the following steps.
S410, obtaining semantic roles and dependency relationships of the question text through grammar structure analysis, and answering the semantic roles and dependency relationships in the text.
S420, respectively performing neural network information compression processing on the semantic roles and the dependences of the question text and the semantic roles and the dependences of the answer text to obtain semantic role vectors of the question text, the dependences vectors of the question text, the semantic role vectors of the answer text and the dependences vectors of the answer text.
In this step, the neural network information compression processing in the embodiment of the disclosure may include, for example, feature encoding, embedding processing, convolution processing, and pooling processing, and feature encoding, embedding processing, convolution processing, and pooling processing are performed on the semantic roles and dependencies in the first text and the semantic roles and dependencies in the second text, respectively, to obtain a semantic role vector and a dependency vector in the first text and a semantic role vector and a dependency vector in the second text.
S430, carrying out semantic and dependency relationship fusion processing on the semantic role vector and the dependency relationship vector in the question text, and carrying out semantic and dependency relationship fusion processing on the semantic role vector and the dependency relationship vector in the answer text.
In this step, the semantic role vector of the first text is added to the dependency vector to obtain a question vector, and the semantic role vector of the second text is added to the dependency vector to obtain an answer vector.
S440, fusion processing is carried out on the question vector and the answer vector.
In this step, the question vector and the answer vector are spliced to obtain a spliced new vector, i.e., a fused vector of the question vector and the answer vector shown in fig. 4.
S450, processing the full connection layer in the neural network on the fused vector to obtain a classification result.
S460, judging whether the question text and the answer text are related according to the classification result.
As an example, the classification result is a preset classification value, for example, the classification value is 0 or 1, if the preset classification value is 0 to indicate that the questions are irrelevant, and the classification value is 1 to indicate that the questions are relevant, then it can be determined whether the questions are relevant to the answer texts, that is, whether the questions are not answered.
Fig. 5 is a schematic structural diagram of a text relevance determining apparatus according to an embodiment of the present disclosure.
In a second aspect, referring to fig. 5, an embodiment of the present disclosure provides a text relevance determining apparatus 500, which may include the following modules.
Vector determination module 510 is configured to determine a semantic role vector and a dependency vector in the first text and a semantic role vector and a dependency vector in the second text.
The vector fusion module 520 is configured to perform vector fusion on the semantic role vector and the dependency vector in the first text, and on the semantic role vector and the dependency vector in the second text, and perform vector fusion on the vector fusion results obtained respectively again, so as to obtain a fused vector of the first text and the second text.
The classification determining module 530 is configured to classify the fused vector through a preset neural network, and determine a correlation between the first text and the second text according to the classification result.
In some embodiments, the first text and the second text are question text and answer text.
In some embodiments, the first text and the second text are text or character strings having a length less than a predetermined threshold.
In some embodiments, the text relevance determining apparatus 500 further includes: the analysis module is used for respectively analyzing the semantic roles and the dependences of the acquired first text and the acquired second text before determining the semantic roles and the dependences of the first text and the semantic roles and the dependences of the second text, so as to obtain the semantic roles and the dependences of the first text and the semantic roles and the dependences of the second role.
In some embodiments, the vector determination module 510 includes: the encoding unit is used for respectively encoding the semantic roles and the dependences in the first text and the semantic roles and the dependences in the second text to respectively obtain sparse feature codes corresponding to the semantic roles and the dependences in the first text and sparse feature codes corresponding to the semantic roles and the dependences in the second text; the embedding processing unit is used for carrying out embedding processing on the sparse feature codes corresponding to the semantic roles and the dependency relationships respectively to obtain dense feature codes corresponding to the semantic roles and the dependency relationships in the first text and dense feature codes corresponding to the semantic roles and the dependency relationships in the second text; the convolution and pooling processing unit is used for carrying out convolution and pooling processing on the dense feature codes corresponding to the first text, and obtaining the semantic role vector and the dependency vector in the first text and the semantic role vector and the dependency vector in the second text.
In some embodiments, vector fusion module 520 includes: the first vector adding unit is used for carrying out vector addition on the semantic role vector and the dependency relationship vector in the first text to obtain a problem vector; the second vector adding unit is used for carrying out vector addition on the semantic role vector and the dependency relationship vector in the second text to obtain an answer vector; and the vector splicing unit is used for carrying out vector splicing on the question vector and the answer vector to obtain a fused vector of the first text and the second text.
In some embodiments, the pre-set neural network includes a fully connected layer; the classification determination module 530 includes: the vector classification unit is used for classifying the fused vectors through the full connection layer to obtain classification results of the fused vectors; the classification determining module 530 is further configured to determine that the first text is related to the second text if the classification result is a preset first value, and determine that the first text is not related to the second text if the classification result is a preset second value.
In some embodiments, the first text and the second text are question text and answer text.
In some embodiments, the first text and the second text are text or character strings having a length less than a predetermined threshold.
According to the text relevance judging device disclosed by the embodiment of the disclosure, the semantic role vector and the dependency relationship vector of the first text and the semantic role and the dependency relationship of the second text can be subjected to vector fusion, the fused vectors are classified by utilizing a preset neural network, and the relevance of the first text and the second text is determined according to the classification result.
The embodiment of the disclosure is not limited to the content of the preset vocabulary, manual excavation and vocabulary arrangement are not needed, the labor cost can be saved, and the method and the device have wider application; because the semantic role vector and the dependency relation vector of the text embody the grammar structure of the text, the embodiment of the disclosure can acquire the vectors of the first text and the second text through the angle of the grammar structure, and can remove the influence of the similar word meaning on the similarity judgment result, so the judgment result has stability; moreover, because the change space of the grammar structure is far smaller than the change space of the word meaning of the text, the text similarity judging method has wider universality; further, the feature space corresponding to the grammar structure is smaller, so that the space occupied by the required module is smaller and the processing speed is faster when the feature vector is processed through the neural network.
In a third aspect, embodiments of the present disclosure further provide a text relevance determining model, which is configured to perform any one of the text relevance determining methods described in connection with fig. 2 to fig. 4 above according to the received first text and second text.
In one embodiment, the model structure of the text relevance determining model may refer to the model structure described in connection with fig. 3, and the processing flow of the acquired first text and the acquired second text by the text relevance determining model may refer to the processing flow of the module structure shown in fig. 3 and the processing flow of the text relevance determining method described in connection with fig. 2-4.
It should be clear that, for convenience and brevity of description, detailed descriptions of known methods are omitted herein, and specific working processes of the systems, modules and units described above may refer to corresponding processes in the foregoing method embodiments, which are not repeated herein.
According to embodiments of the present disclosure, the present disclosure also provides an electronic device and a readable storage medium.
Fig. 6 illustrates a schematic block diagram of an example electronic device 600 that may be used to implement embodiments of the present disclosure. Electronic devices are intended to represent various forms of digital computers, such as laptops, desktops, workstations, personal digital assistants, servers, blade servers, mainframes, and other appropriate computers. The electronic device may also represent various forms of mobile devices, such as personal digital processing, cellular telephones, smartphones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions, are meant to be exemplary only, and are not meant to limit implementations of the disclosure described and/or claimed herein.
As shown in fig. 6, the apparatus 600 includes a computing unit 601 that can perform various appropriate actions and processes according to a computer program stored in a Read Only Memory (ROM) 602 or a computer program in a Random Access Memory (RAM) 603 from a storage unit 608. In the RAM603, various programs and data required for the operation of the device 600 may also be stored. The computing unit 601, ROM602, and RAM603 are connected to each other by a bus 604. An input/output (I/O) interface 605 is also connected to bus 604. Various components in the device 600 are connected to the I/O interface 605, including: an input unit 606 such as a keyboard, mouse, etc.; an output unit 607 such as various types of displays, speakers, and the like; a storage unit 608, such as a magnetic disk, optical disk, or the like; and a communication unit 609 such as a network card, modem, wireless communication transceiver, etc. The communication unit 609 allows the device 600 to exchange information/data with other devices via a computer network, such as the internet, and/or various telecommunication networks.
The computing unit 601 may be a variety of general and/or special purpose processing components having processing and computing capabilities. Some examples of computing unit 601 include, but are not limited to, a Central Processing Unit (CPU), a Graphics Processing Unit (GPU), various specialized Artificial Intelligence (AI) computing chips, various computing units running machine learning model algorithms, a Digital Signal Processor (DSP), and any suitable processor, controller, microcontroller, etc. The calculation unit 601 performs the respective methods and processes described above, for example, the text relevance determination method. For example, in some embodiments, the text relevance determining method may be implemented as a computer software program tangibly embodied on a machine-readable medium, such as the storage unit 608. In some embodiments, part or all of the computer program may be loaded and/or installed onto the device 600 via the ROM602 and/or the communication unit 609. When the computer program is loaded into the RAM603 and executed by the computing unit 601, one or more steps of the text relevance determining method described above may be performed. Alternatively, in other embodiments, the computing unit 601 may be configured to perform the text relevance determination method in any other suitable way (e.g. by means of firmware).
Various implementations of the systems and techniques described here above may be implemented in digital electronic circuitry, integrated circuit systems, field Programmable Gate Arrays (FPGAs), application Specific Integrated Circuits (ASICs), application Specific Standard Products (ASSPs), systems On Chip (SOCs), load programmable logic devices (CPLDs), computer hardware, firmware, software, and/or combinations thereof. These various embodiments may include: implemented in one or more computer programs, the one or more computer programs may be executed and/or interpreted on a programmable system including at least one programmable processor, which may be a special purpose or general-purpose programmable processor, that may receive data and instructions from, and transmit data and instructions to, a storage system, at least one input device, and at least one output device.
Program code for carrying out methods of the present disclosure may be written in any combination of one or more programming languages. These program code may be provided to a processor or controller of a general purpose computer, special purpose computer, or other programmable data processing apparatus such that the program code, when executed by the processor or controller, causes the functions/operations specified in the flowchart and/or block diagram to be implemented. The program code may execute entirely on the machine, partly on the machine, as a stand-alone software package, partly on the machine and partly on a remote machine or entirely on the remote machine or server.
In the context of this disclosure, a machine-readable medium may be a tangible medium that can contain, or store a program for use by or in connection with an instruction execution system, apparatus, or device. The machine-readable medium may be a machine-readable signal medium or a machine-readable storage medium. The machine-readable medium may include, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any suitable combination of the foregoing. More specific examples of a machine-readable storage medium would include an electrical connection based on one or more wires, a portable computer diskette, a hard disk, a Random Access Memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.
To provide for interaction with a user, the systems and techniques described here can be implemented on a computer having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to a user; and a keyboard and pointing device (e.g., a mouse or trackball) by which a user can provide input to the computer. Other kinds of devices may also be used to provide for interaction with a user; for example, feedback provided to the user may be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user may be received in any form, including acoustic input, speech input, or tactile input.
The systems and techniques described here can be implemented in a computing system that includes a background component (e.g., as a data server), or that includes a middleware component (e.g., an application server), or that includes a front-end component (e.g., a user computer having a graphical user interface or a web browser through which a user can interact with an implementation of the systems and techniques described here), or any combination of such background, middleware, or front-end components. The components of the system can be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include: local Area Networks (LANs), wide Area Networks (WANs), and the internet.
The computer system may include a client and a server. The client and server are typically remote from each other and typically interact through a communication network. The relationship of client and server arises by virtue of computer programs running on the respective computers and having a client-server relationship to each other.
According to an embodiment of the present disclosure, the present disclosure also provides a computer program product comprising a computer program which, when executed by a processor, implements any of the above methods of text relevance determination.
Artificial intelligence is the discipline of studying the process of making a computer simulate certain thinking and intelligent behavior (e.g., learning, reasoning, planning, etc.) of a person, both hardware-level and software-level techniques. The artificial intelligence hardware technology generally comprises technologies such as a sensor, a special artificial intelligence chip, cloud computing, distributed storage, big data processing and the like; the artificial intelligence software technology comprises a computer vision technology, a voice recognition technology, a natural language processing technology, a machine learning/deep learning technology, a big data processing technology, a knowledge graph technology and the like.
It should be appreciated that various forms of the flows shown above may be used to reorder, add, or delete steps. For example, the steps recited in the present application may be performed in parallel or sequentially or in a different order, provided that the desired results of the disclosed embodiments are achieved, and are not limited herein.
The above detailed description should not be taken as limiting the scope of the present disclosure. It will be apparent to those skilled in the art that various modifications, combinations, sub-combinations and alternatives are possible, depending on design requirements and other factors. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present disclosure are intended to be included within the scope of the present disclosure.
Claims (13)
1. A text relevance determining method, comprising:
respectively analyzing the semantic roles and the dependency relationships of the acquired first text and the acquired second text to obtain the semantic roles and the dependency relationships of the first text and the semantic roles and the dependency relationships of the second roles;
determining the semantic role vector and the dependency vector in the first text and the semantic role vector and the dependency vector in the second text comprises: encoding semantic roles and dependency relationships in the first text and semantic roles and dependency relationships in the second text respectively to obtain sparse feature codes corresponding to the semantic roles and the dependency relationships in the first text and sparse feature codes corresponding to the semantic roles and the dependency relationships in the second text; embedding the sparse feature codes respectively corresponding to the sparse feature codes to obtain dense feature codes corresponding to the semantic roles and the dependency relationships in the first text and dense feature codes corresponding to the semantic roles and the dependency relationships in the second text; rolling and pooling the corresponding dense feature codes to obtain semantic role vectors and dependency vectors in the first text and semantic role vectors and dependency vectors in the second text;
Respectively carrying out vector fusion on the semantic role vector and the dependency relationship vector in the first text and the semantic role vector and the dependency relationship vector in the second text, and carrying out vector fusion on the vector fusion results obtained respectively again to obtain the fused vectors of the first text and the second text;
classifying the fused vectors through a preset neural network, and determining the correlation between the first text and the second text according to the classification result.
2. The method according to claim 1, wherein the performing vector fusion on the semantic role vector and the dependency vector in the first text and the semantic role vector and the dependency vector in the second text respectively, and performing vector fusion on the vector fusion result obtained respectively again to obtain the fused vectors of the first text and the second text, includes:
vector addition is carried out on the semantic role vector and the dependency relationship vector in the first text, so that a problem vector is obtained;
vector addition is carried out on the semantic role vector and the dependency relationship vector in the second text, and an answer vector is obtained;
And vector splicing is carried out on the question vector and the answer vector, so that a fused vector of the first text and the second text is obtained.
3. The method of claim 1, wherein the pre-set neural network comprises a fully connected layer; the classifying, by a preset neural network, the fused vector, and determining the correlation between the first text and the second text according to the classification result, including:
classifying the fused vectors through the full connection layer to obtain classification results of the fused vectors;
and determining that the first text is related to the second text under the condition that the classification result is a preset first value, and determining that the first text is not related to the second text under the condition that the classification result is a preset second value.
4. A method according to any one of claims 1 to 3, wherein,
the first text and the second text are question text and answer text.
5. A method according to any one of claims 1 to 3, wherein,
the first text and the second text are texts or character strings with lengths smaller than a preset threshold value.
6. A correlation determination apparatus, comprising:
the analysis module is used for respectively analyzing the semantic roles and the dependences of the acquired first text and the acquired second text before determining the semantic roles and the dependences of the first text and the semantic roles and the dependences of the second text, so as to obtain the semantic roles and the dependences of the first text and the semantic roles and the dependences of the second role;
the vector determining module is used for determining semantic role vectors and dependency vectors in the first text and semantic role vectors and dependency vectors in the second text;
the vector fusion module is used for carrying out vector fusion on the semantic role vector and the dependency relationship vector in the first text and the semantic role vector and the dependency relationship vector in the second text respectively, and carrying out vector fusion on the vector fusion results obtained respectively again to obtain the fused vectors of the first text and the second text;
the classification determining module is used for classifying the fused vectors through a preset neural network and determining the correlation between the first text and the second text according to the classification result;
The vector determination module includes: the encoding unit is used for respectively encoding the semantic roles and the dependences in the first text and the semantic roles and the dependences in the second text to respectively obtain sparse feature codes corresponding to the semantic roles and the dependences in the first text and sparse feature codes corresponding to the semantic roles and the dependences in the second text; the embedding processing unit is used for carrying out embedding processing on the sparse feature codes corresponding to the semantic roles and the dependency relationships respectively to obtain dense feature codes corresponding to the semantic roles and the dependency relationships in the first text and dense feature codes corresponding to the semantic roles and the dependency relationships in the second text; the convolution and pooling processing unit is used for carrying out convolution and pooling processing on the dense feature codes corresponding to the first text, and obtaining the semantic role vector and the dependency vector in the first text and the semantic role vector and the dependency vector in the second text.
7. The apparatus of claim 6, wherein the vector fusion module comprises:
the first vector adding unit is used for carrying out vector addition on the semantic role vector and the dependency relationship vector in the first text to obtain a problem vector;
The second vector adding unit is used for carrying out vector addition on the semantic role vector and the dependency relationship vector in the second text to obtain an answer vector;
and the vector splicing unit is used for carrying out vector splicing on the question vector and the answer vector to obtain a fused vector of the first text and the second text.
8. The apparatus of claim 6, wherein the pre-set neural network comprises a fully connected layer; the classification determination module comprises:
the vector classification unit is used for classifying the fused vectors through the full connection layer to obtain classification results of the fused vectors;
the classification determining module is further configured to determine that the first text is related to the second text if the classification result is a preset first value, and determine that the first text is not related to the second text if the classification result is a preset second value.
9. The device according to any one of claims 6 to 8, wherein,
the first text and the second text are question text and answer text.
10. The device according to any one of claims 6 to 8, wherein,
The first text and the second text are texts or character strings with lengths smaller than a preset threshold value.
11. A text relevance judgment model is characterized in that,
the text relevance judgment model is used for executing the method of any one of claims 1-5 according to the received first text and second text to obtain a relevance judgment result of the first text and the second text.
12. An electronic device, comprising:
at least one processor; and
a memory communicatively coupled to the at least one processor; wherein,
the memory stores instructions executable by the at least one processor to enable the at least one processor to perform the method of any one of claims 1-5.
13. A non-transitory computer readable storage medium storing computer instructions for causing the computer to perform the method of any one of claims 1-5.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011548198.9A CN112559713B (en) | 2020-12-24 | 2020-12-24 | Text relevance judging method and device, model, electronic equipment and readable medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011548198.9A CN112559713B (en) | 2020-12-24 | 2020-12-24 | Text relevance judging method and device, model, electronic equipment and readable medium |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN112559713A CN112559713A (en) | 2021-03-26 |
| CN112559713B true CN112559713B (en) | 2023-12-01 |
Family
ID=75033129
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202011548198.9A Active CN112559713B (en) | 2020-12-24 | 2020-12-24 | Text relevance judging method and device, model, electronic equipment and readable medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112559713B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113268566B (en) * | 2021-05-28 | 2022-06-14 | 平安国际智慧城市科技股份有限公司 | Question and answer pair quality evaluation method, device, equipment and storage medium |
Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104298658A (en) * | 2014-10-29 | 2015-01-21 | 百度在线网络技术（北京）有限公司 | Method and device for acquiring search result |
| CN104462327A (en) * | 2014-12-02 | 2015-03-25 | 百度在线网络技术（北京）有限公司 | Computing method, search processing method, computing device and search processing device for sentence similarity |
| CN106202010A (en) * | 2016-07-12 | 2016-12-07 | 重庆兆光科技股份有限公司 | The method and apparatus building Law Text syntax tree based on deep neural network |
| CN107818081A (en) * | 2017-09-25 | 2018-03-20 | 沈阳航空航天大学 | Sentence similarity appraisal procedure based on deep semantic model and semantic character labeling |
| CN110969023A (en) * | 2018-09-29 | 2020-04-07 | 北京国双科技有限公司 | Text similarity determination method and device |
| CN111353306A (en) * | 2020-02-22 | 2020-06-30 | 杭州电子科技大学 | Entity relationship and dependency Tree-LSTM-based combined event extraction method |
| CN111930914A (en) * | 2020-08-14 | 2020-11-13 | 工银科技有限公司 | Question generation method and device, electronic equipment and computer-readable storage medium |
| CN112036189A (en) * | 2020-08-10 | 2020-12-04 | 中国人民大学 | Method and system for recognizing gold semantic |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109902166A (en) * | 2019-03-12 | 2019-06-18 | 北京百度网讯科技有限公司 | Vision Question-Answering Model, electronic equipment and storage medium |
-
2020
- 2020-12-24 CN CN202011548198.9A patent/CN112559713B/en active Active
Patent Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104298658A (en) * | 2014-10-29 | 2015-01-21 | 百度在线网络技术（北京）有限公司 | Method and device for acquiring search result |
| CN104462327A (en) * | 2014-12-02 | 2015-03-25 | 百度在线网络技术（北京）有限公司 | Computing method, search processing method, computing device and search processing device for sentence similarity |
| CN106202010A (en) * | 2016-07-12 | 2016-12-07 | 重庆兆光科技股份有限公司 | The method and apparatus building Law Text syntax tree based on deep neural network |
| CN107818081A (en) * | 2017-09-25 | 2018-03-20 | 沈阳航空航天大学 | Sentence similarity appraisal procedure based on deep semantic model and semantic character labeling |
| CN110969023A (en) * | 2018-09-29 | 2020-04-07 | 北京国双科技有限公司 | Text similarity determination method and device |
| CN111353306A (en) * | 2020-02-22 | 2020-06-30 | 杭州电子科技大学 | Entity relationship and dependency Tree-LSTM-based combined event extraction method |
| CN112036189A (en) * | 2020-08-10 | 2020-12-04 | 中国人民大学 | Method and system for recognizing gold semantic |
| CN111930914A (en) * | 2020-08-14 | 2020-11-13 | 工银科技有限公司 | Question generation method and device, electronic equipment and computer-readable storage medium |
Non-Patent Citations (1)
| Title |
|---|
| 融合Gate过滤机制与深度Bi-LSTM-CRF的汉语语义角色标注;张苗苗;刘明童;张玉洁;徐金安;陈钰枫;;情报工程(第02期);第46-54页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN112559713A (en) | 2021-03-26 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN113553412B (en) | Question-answering processing method, question-answering processing device, electronic equipment and storage medium | |
| CN113590776B (en) | Knowledge graph-based text processing method and device, electronic equipment and medium | |
| CN113360711B (en) | Model training and executing method, device, equipment and medium for video understanding task | |
| CN113139043B (en) | Question-answer sample generation method and device, electronic equipment and storage medium | |
| CN113836268A (en) | Document understanding method and device, electronic equipment and medium | |
| CN115062718A (en) | Language model training method and device, electronic equipment and storage medium | |
| CN116049597A (en) | Pre-training method and device for multi-task model of webpage and electronic equipment | |
| CN112560461A (en) | News clue generation method and device, electronic equipment and storage medium | |
| CN112528146B (en) | Content resource recommendation method and device, electronic equipment and storage medium | |
| CN114817478A (en) | Text-based question and answer method and device, computer equipment and storage medium | |
| CN112559713B (en) | Text relevance judging method and device, model, electronic equipment and readable medium | |
| CN114120166A (en) | Video question and answer method and device, electronic equipment and storage medium | |
| CN117333889A (en) | Training method and device for document detection model and electronic equipment | |
| CN117493595A (en) | Image searching method, device, equipment and medium based on large model | |
| CN117421403A (en) | Intelligent dialogue method and device and electronic equipment | |
| CN114647739B (en) | Entity chain finger method, device, electronic equipment and storage medium | |
| CN115048523B (en) | Text classification method, device, equipment and storage medium | |
| CN114818736B (en) | Text processing method, chain finger method and device for short text and storage medium | |
| CN115601620A (en) | Feature fusion method and device, electronic equipment and computer readable storage medium | |
| CN114118049B (en) | Information acquisition method, device, electronic equipment and storage medium | |
| CN114970666A (en) | Spoken language processing method and device, electronic equipment and storage medium | |
| CN115204139A (en) | Text matching processing method and device | |
| CN113886543A (en) | Method, apparatus, medium, and program product for generating an intent recognition model | |
| CN113761874A (en) | Event reality prediction method and device, electronic equipment and storage medium | |
| CN114942980B (en) | Method and device for determining text matching |
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 |