Data processing method and device, computer equipment and readable storage medium

Abstract

The invention relates to a data processing method and device, computer equipment, a computer readable storage medium and a computer program product. The method comprises the steps of obtaining to-be-processed text data, and preprocessing the to-be-processed text data to obtain a word identifier list; determining a relation matrix based on the matrix marking model and the word identification list; extracting each entity in the relation matrix according to a matrix marking model, and determining a target entity group corresponding to a target relation according to the target relation represented by the relation matrix and a relation threshold value in the matrix marking model; constructing a triple according to the target entity group and the target relationship; the triple is used for constructing the knowledge graph. By adopting the method, the data processing efficiency can be improved.

Description

Technical Field

[0001] The present application relates to the field of big data technology, and in particular to a data processing method, apparatus, computer equipment, computer-readable storage medium, and computer program product. Background Art

[0002] With the rapid development of Internet technology, users can search for the knowledge they need through the Internet. When users search for entity data, the search results contain the entity's knowledge graph, which contains the relationship between the entity and other entities.

[0003] The current data processing method predetermines the target relations that need to be extracted. The text data to be processed is obtained, and the text data to be processed includes each initial sentence. For each target relation, the staff extracts the target entity with the target relation in each initial sentence, and constructs a triple of target entity-target relation-target entity. Then, a knowledge graph is constructed based on each triple.

[0004] However, in current data processing methods, when there is a large amount of text data to be processed, manually extracting target entities with target relationships will consume a lot of time. Therefore, the efficiency of current data processing methods is low. Summary of the invention

[0005] Based on this, it is necessary to provide a data transmission method, apparatus, computer device, computer-readable storage medium and computer program product to address the above technical issues.

[0006] In a first aspect, the present application provides a data transmission method, comprising:

[0007] Acquire text data to be processed, and preprocess the text data to be processed to obtain a word identifier list;

[0008] Determining a relationship matrix based on the matrix labeling model and the word identification list;

[0009] Extracting each entity in the relationship matrix according to the matrix labeling model, and determining a target entity group corresponding to the target relationship according to the target relationship represented by the relationship matrix and a relationship threshold in the matrix labeling model;

[0010] A triple is constructed according to the target entity group and the target relationship; the triple is used to construct a knowledge graph.

[0011] In one embodiment, the preprocessing of the text data to be processed to obtain a word identifier list includes:

[0012] For each initial sentence in the text data to be processed, a head mark is added at the beginning of the initial sentence, and a tail mark is added at the end of the initial sentence to obtain a sentence;

[0013] The sentences are preprocessed according to a preset sentence length, and a word identification list is constructed based on each preprocessed sentence.

[0014] In one embodiment, determining a relationship matrix based on a matrix labeling model and the word identifier list includes:

[0015] Performing feature extraction on the word identification list according to a matrix tagging model to obtain a first word vector matrix and a second word vector matrix; the matrix tagging model includes a reference matrix of the target relationship;

[0016] Add the word vector of the head tag unit in the first word vector matrix to each word vector of the second word vector matrix to obtain a head vector matrix, and determine the first word vector matrix as a tail vector matrix;

[0017] The similarity is calculated for the reference matrix of the target relationship, the head vector matrix and the tail vector matrix to obtain a similarity matrix, and the similarity matrix is ​​normalized to obtain a relationship matrix.

[0018] In one embodiment, the entity includes a head entity and a tail entity, and extracting each entity in the relationship matrix according to the matrix labeling model includes:

[0019] Binarizing the relationship matrix according to the entity threshold in the matrix labeling model to obtain an entity binarization matrix;

[0020] Performing dimensionality reduction processing on the entity binary matrix to obtain a head entity vector and a tail entity vector;

[0021] The head entity and the tail entity corresponding to each sentence are extracted according to the head entity vector and the tail entity vector.

[0022] In one embodiment, each of the entities includes each head entity and each tail entity corresponding to each sentence, and determining the target entity group corresponding to the target relationship according to the target relationship represented by the relationship matrix and the relationship threshold in the matrix labeling model includes:

[0023] According to the relationship threshold in the matrix labeling model, the relationship matrix is ​​binarized to obtain a relationship binarization matrix;

[0024] For each of the sentences, combining the head entities and the tail entities corresponding to the sentence to obtain entity groups corresponding to the sentence;

[0025] According to the target relationship represented by the relationship binary matrix, each of the entity groups is screened to obtain a target entity group corresponding to the target relationship.

[0026] In one embodiment, the step of screening each of the entity groups according to the target relationship represented by the relationship binary matrix to obtain the target entity group corresponding to the target relationship includes:

[0027] Determining a relation binarization submatrix corresponding to each entity group in the relation binarization matrix;

[0028] Determine whether the value of the relationship binary submatrix meets the screening condition corresponding to the target relationship;

[0029] If the value of the relationship binarization submatrix meets the screening condition, the entity group is determined as the target entity group corresponding to the target relationship.

[0030] In a second aspect, the present application further provides a data processing device, comprising:

[0031] An acquisition module, used for acquiring text data to be processed, and preprocessing the text data to be processed to obtain a word identification list;

[0032] A determination module, configured to determine a relationship matrix based on a matrix labeling model and the word identification list;

[0033] An extraction module, used to extract each entity in the relationship matrix according to the matrix labeling model, and determine a target entity group corresponding to the target relationship according to the target relationship represented by the relationship matrix and the relationship threshold in the matrix labeling model;

[0034] A construction module is used to construct triples according to the target entity group and the target relationship; the triples are used to construct a knowledge graph.

[0035] In a third aspect, the present application further provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the following steps are implemented:

[0036] Acquire text data to be processed, and preprocess the text data to be processed to obtain a word identifier list;

[0037] Determining a relationship matrix based on the matrix labeling model and the word identification list;

[0038] Extracting each entity in the relationship matrix according to the matrix labeling model, and determining a target entity group corresponding to the target relationship according to the target relationship represented by the relationship matrix and a relationship threshold in the matrix labeling model;

[0039] A triple is constructed according to the target entity group and the target relationship; the triple is used to construct a knowledge graph.

[0040] In a fourth aspect, the present application further provides a computer-readable storage medium having a computer program stored thereon, wherein when the computer program is executed by a processor, the following steps are implemented:

[0041] Acquire text data to be processed, and preprocess the text data to be processed to obtain a word identifier list;

[0042] Determining a relationship matrix based on the matrix labeling model and the word identification list;

[0043] Extracting each entity in the relationship matrix according to the matrix labeling model, and determining a target entity group corresponding to the target relationship according to the target relationship represented by the relationship matrix and a relationship threshold in the matrix labeling model;

[0044] A triple is constructed according to the target entity group and the target relationship; the triple is used to construct a knowledge graph.

[0045] In a fifth aspect, the present application further provides a computer program product, including a computer program, which implements the following steps when executed by a processor:

[0046] Acquire text data to be processed, and preprocess the text data to be processed to obtain a word identifier list;

[0047] Determining a relationship matrix based on the matrix labeling model and the word identification list;

[0048] Extracting each entity in the relationship matrix according to the matrix labeling model, and determining a target entity group corresponding to the target relationship according to the target relationship represented by the relationship matrix and a relationship threshold in the matrix labeling model;

[0049] A triple is constructed according to the target entity group and the target relationship; the triple is used to construct a knowledge graph.

[0050] The data processing method, device, computer equipment, computer-readable storage medium and computer program product obtain the text data to be processed, and pre-process the text data to be processed to obtain a word identification list; determine the relationship matrix based on the matrix tag model and the word identification list; extract each entity in the relationship matrix according to the matrix tag model, and determine the target entity group corresponding to the target relationship according to the target relationship represented by the relationship matrix and the relationship threshold in the matrix tag model; construct triples according to the target entity group and the target relationship; the triples are used to construct a knowledge graph. Using this method, the relationship matrix is ​​determined through the matrix tag model and the text data to be processed, and the entities in the relationship matrix are extracted according to the matrix tag model, and each entity that may have a target relationship is obtained. Then, the target entity group with the target relationship represented by the relationship matrix and the relationship threshold are determined, and the automatic extraction of the target entity group is realized. Furthermore, each triple is constructed based on the target entity group and the target relationship, and the automatic construction of the triple is realized, which improves the efficiency of the data processing method. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] In order to more clearly illustrate the technical solutions in the embodiments of the present application or related technologies, the drawings required for use in the embodiments of the present application or related technical descriptions will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without paying creative work.

[0052] Figure 1 is a flow chart of a data processing method in one embodiment;

[0053] Figure 2 A schematic diagram of a process for constructing a word identification list in one embodiment;

[0054] Figure 3 A schematic diagram of a process for determining a relationship matrix in one embodiment;

[0055] Figure 4 A schematic diagram of a process for determining a matrix marking model in an exemplary embodiment;

[0056] Figure 5 A schematic diagram of a process of extracting a head entity and a tail entity in one embodiment;

[0057] Figure 6 A schematic diagram of a process for determining a target entity group in one embodiment;

[0058] Figure 7 A schematic diagram of a process for screening a target entity group in one embodiment;

[0059] Figure 8 A flow chart of a method for extracting entity relationships for an exemplary embodiment;

[0060] Fig. 9 is a flow chart of data processing of an exemplary embodiment;

[0061] Fig.10 is a structural block diagram of a data processing device in one embodiment;

[0062] Fig.11 FIG. 4 is a diagram showing the internal structure of a computer device in one embodiment. DETAILED DESCRIPTION

[0063] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0064] In one embodiment, Figure 1 As shown, a data processing method is provided. The embodiment of the present application takes the method applied to a computer device as an example for explanation. The embodiment of the present application does not limit the execution device for executing the data processing method, and includes the following steps:

[0065] Step 102, obtaining text data to be processed, and preprocessing the text data to be processed to obtain a word identification list.

[0066] In implementation, the computer device obtains the text data to be processed from the database, wherein the text sentences to be processed include each initial sentence, and the computer device pre-processes the length and format of each initial sentence to obtain a word identification list.

[0067] Specifically, the computer device crawls the initial text data to be processed from the web page and divides the initial text data into initial sentences. Then, the computer device stores the initial sentences in a database. The computer device obtains the initial sentences from the database and constructs the text data to be processed for each initial sentence. Then, the computer device adds a mark to each initial sentence to obtain a sentence, and pre-processes the sentence to obtain each processed sentence. Then, the computer device constructs a word identification list according to each processed sentence.

[0068] Optionally, the computer device may obtain the initial text data from, but is not limited to, journals, papers, and web pages. The embodiment of the present application does not limit the source of the initial text data.

[0069] Step 104, determining a relationship matrix based on the matrix labeling model and the word identifier list.

[0070] The matrix labeling model is obtained by training the language representation model based on the sample sentence set, which includes each sample sentence and a sample binary matrix corresponding to each sample matrix.

[0071] In implementation, the computer device extracts features from the word identification list based on the matrix tag model to obtain a head vector matrix and a tail vector matrix. Then, the computer device determines a relationship matrix based on the reference matrix, the head vector matrix and the tail vector matrix in the matrix tag model.

[0072] Specifically, a language representation model is pre-set in the computer device. The target relationship is the relationship between entities that need to be extracted. The computer device trains the language representation model according to the sample sentence set and the target relationship until the trained language representation model meets the preset training stop condition, and the computer device determines the trained language representation model as a matrix labeling model. The computer device performs feature extraction on the word identification list based on the matrix labeling model to obtain a head vector matrix and a tail vector matrix. Then, the computer device performs matrix operations on the head vector matrix and the tail vector matrix based on the reference matrix of the target relationship in the matrix labeling model to obtain a relationship matrix. The relationship matrix represents the degree of fit between each sentence in the word identification list and the target relationship.

[0073] Optionally, the target relationship may be, but is not limited to, one, and the embodiment of the present application does not limit the number of target relationships.

[0074] Step 106 , extracting entities in the relationship matrix according to the matrix labeling model, and determining a target entity group corresponding to the target relationship according to the target relationship represented by the relationship matrix and the relationship threshold in the matrix labeling model.

[0075] The matrix labeling model includes entity threshold and relationship threshold.

[0076] In implementation, the computer device extracts each entity in the relationship matrix according to the entity threshold in the matrix labeling model. Then, the computer device combines each entity to obtain each entity group. The computer device selects the target relationship entity group corresponding to the target relationship in each entity group based on the target relationship represented by the relationship matrix and the relationship threshold in the matrix labeling model.

[0077] Specifically, the computer device performs binarization processing on the relationship matrix according to the entity threshold in the matrix labeling model to obtain the entity binarization matrix. Then, the computer device extracts each entity in the entity binarization matrix. The computer device performs binarization processing on the relationship matrix according to the relationship threshold to obtain the relationship binarization matrix. Then, the computer device combines each entity to obtain each entity group. The computer device screens each entity group according to the target relationship represented by the relationship binarization matrix to obtain the target entity group corresponding to the target relationship.

[0078] Step 108: construct triples according to the target entity group and the target relationship.

[0079] Among them, triples are used to construct knowledge graphs.

[0080] In implementation, the computer device combines the target entity group and the target relationship to obtain triples, and constructs a knowledge graph based on each triple.

[0081] In an exemplary embodiment, if there is a target relationship and multiple target entity groups, the computer device constructs each target entity group and the target relationship into a triple. The computer device determines the knowledge triples in each triple where the knowledge entity exists. The knowledge entity is the entity for which the knowledge graph needs to be constructed. The computer device connects each entity in each knowledge triple according to the target relationship in each knowledge triple to obtain the knowledge graph of the knowledge entity.

[0082] In the above data processing method, the relationship matrix is ​​determined by the matrix labeling model and the text data to be processed, and the entities in the relationship matrix are extracted according to the matrix labeling model, so as to obtain the entities that may have the target relationship. Then, the target entity group with the target relationship is determined by the target relationship and the relationship threshold represented by the relationship matrix, and the automatic extraction of the target entity group is realized. Furthermore, each triple is constructed based on the target entity group and the target relationship, and the automatic construction of the triple is realized, thereby improving the efficiency of the data processing method.

[0083] In an exemplary embodiment, Figure 2 As shown, the specific process of preprocessing the text data to be processed in step 102 to obtain the word identification list includes steps 202 to 204. Among them:

[0084] Step 202: for each initial sentence in the text data to be processed, a head mark is added at the beginning of the initial sentence, and a tail mark is added at the end of the initial sentence to obtain a sentence.

[0085] In implementation, the computer device adds a head mark at the beginning of each initial sentence in the text data to be processed. At the same time, the computer device adds a tail mark at the end of the initial sentence to obtain a sentence. By adding the head mark and the tail mark, each matrix can be segmented, and the matrix mark model can clarify each sentence.

[0086] Optionally, the sentence start mark can be but not limited to the [cls] mark, and the sentence end mark can be but not limited to the [sep] mark. The sentence start mark and sentence end mark are set according to the processing requirements of the word identifier list. The embodiment of the present application does not limit the sentence start mark and sentence end mark.

[0087] Step 204 , preprocessing the sentences according to a preset sentence length, and constructing a word identification list based on each preprocessed sentence.

[0088] In implementation, the sentence length is preset in the computer device. The computer device determines for each sentence whether the length of the sentence is the sentence length. If the length of the sentence is the sentence length, the computer device does not preprocess the sentence. If the length of the sentence is less than the sentence length, the computer device adds a placeholder to the sentence according to the preset sentence length. If the length of the sentence is greater than the sentence length, the computer device trims the sentence according to the preset sentence length. The computer device combines the preprocessed sentences through the model dictionary to obtain a word identifier list.

[0089] In an exemplary embodiment, the sentence length is 7 characters. The text data to be processed includes three sentences, namely [cls] Small-headed father is the father of big-headed son [sep], [cls] Hee hee is a word [sep], and [cls] Jin Yong is the author [sep]. The computer device trims the sentence [cls] Small-headed father is the father of big-headed son [sep] according to the sentence length of 7 characters to obtain [cls] Small-headed father is big-headed son [sep]. Since the length of the sentence [cls] Hee hee is a word [sep] is 7 characters, that is, the sentence length, the computer device does not pre-process the sentence. The computer device adds two placeholders after the sentence [cls] Jin Yong is the author [sep] to obtain the pre-processed sentence: [cls] Jin Yong is the author [sep] XX. Among them, XX represents a placeholder. The computer device combines the three sentences [cls]Small-headed father is a big-headed son[sep], [cls]Hee hee is a word[sep] and [cls]Jin Yong is the author[sep]XX to obtain a list of word identifiers.

[0090] Optionally, the sentence length is determined according to processing requirements, and the embodiment of the present application does not limit the sentence length.

[0091] In this embodiment, sentences are obtained by adding sentence start markers and sentence end markers to each initial sentence, and the sentences are preprocessed according to the sentence length to obtain a word marker list, which facilitates the subsequent matrix labeling model to process the word marker list.

[0092] In an exemplary embodiment, Figure 3 As shown, the specific processing process of step 204 includes steps 302 to 306. Among them:

[0093] Step 302: extract features from the word identifier list according to the matrix labeling model to obtain a first word vector matrix and a second word vector matrix.

[0094] The matrix labeling model includes a reference matrix of the target relationship and a multi-layer Transformer encoder.

[0095] During implementation, the computer device inputs the word identifier list into the matrix labeling model, performs feature extraction on the word identifier list through the matrix labeling model, and obtains a first word vector matrix and a second word vector matrix.

[0096] Specifically, the computer device inputs the word identification list into the matrix identification model, performs feature extraction on the word identification list through the multi-layer Transformer encoder in the matrix identification model, and obtains the output of the last two layers of Transformer encoders in the matrix identification model. The computer device determines the output of the last layer of Transformer encoder in the matrix identification model as the first word vector matrix, and determines the output of the penultimate layer of Transformer encoder in the matrix identification model as the second word vector matrix. By determining the output of the last two layers of the matrix identification model as the word vector matrix, richer information can be obtained.

[0097] In an optional embodiment, a BERT (Bidirectional Encoder Representations from Transformers, a pre-trained natural language processing model) model is pre-set in the computer device. The computer device obtains each sample sentence and pre-processes each sample sentence to obtain a sample word identification list. At the same time, the computer device marks the sample sentence based on the 0-1 marking method for each sample sentence to obtain a sample binary matrix. Specifically, the computer device is pre-set with a target relationship to be extracted. The computer device determines each sample sentence as a row of the sample binary matrix, and determines each word in the sample sentence as a column in the sample binary matrix. The computer device determines whether there is an entity of the target relationship in each sample sentence. If there is an entity of the target relationship, the computer device marks the position of the entity with the target relationship as 1, and the positions of the remaining words without the target relationship as 0, thereby obtaining a sample binary matrix of the target relationship. The position (i, j) of the sample binary matrix corresponds to the word pair consisting of word i and word j. If word i is the word of the head entity of the target relation triplet and word j is the word of the tail entity of the same triplet, the position is marked as 1, otherwise it is marked as 0. The computer device inputs the sample word identification list into the BERT model, and fine-tunes the parameters of the BERT model based on the sample binary matrix and the weighted cross entropy loss function until the trained BERT model meets the preset training stop condition, and the computer device determines the trained BERT model as a matrix labeling model.

[0098] Optionally, the target relationship may not be limited to a father-son relationship, a master-apprentice relationship, etc. This application is private and does not impose any limitation on the target relationship.

[0099] Step 304, add the word vector of the head tag unit in the first word vector matrix to each word vector of the second word vector matrix to obtain a head vector matrix, and determine the first word vector matrix as a tail vector matrix.

[0100] The first word vector matrix includes each word vector, and each word vector includes a word vector of a head tag unit, a word vector of a tail tag unit, and other word vectors.

[0101] In implementation, the computer device adds the word vector of the head marker unit in the first word vector matrix to each word vector of the second word vector matrix to obtain the head vector matrix, so as to better learn the relationship between the head and the tail. The computer device determines the first word vector matrix as the tail vector matrix. The head entity word vector in the head vector matrix represents the possibility of the word as the head. The head is the first target entity in the target triple. The tail entity word vector in the tail vector matrix represents the possibility of the word as the tail. The tail is the second target entity in the target triple.

[0102] Step 306, similarity calculation is performed on the reference matrix, the head vector matrix and the tail vector matrix of the target relationship to obtain a similarity matrix, and the similarity matrix is ​​normalized to obtain a relationship matrix.

[0103] In the implementation, the computer device performs similarity calculation on the reference matrix, the head vector matrix and the tail vector matrix of the target relationship to obtain a similarity matrix representing the target relationship. Then, the computer device normalizes the values ​​of the similarity matrix into probability values ​​to obtain a relationship matrix representing the target relationship. The relationship matrix represents the possibility that each sentence in the word identifier list fits this relationship.

[0104] Specifically, the computer device performs a linear transformation on each head entity word vector in the head vector matrix based on the reference matrix to obtain the head entity vector after the linear transformation. At the same time, the computer device performs a linear transformation on each tail entity vector in the tail vector matrix based on the reference matrix to obtain the tail entity vector after the linear transformation. The computer device calculates the normalized inner product of each linearly transformed head entity vector and each linearly transformed tail entity vector to obtain each inner product result, and constructs a similarity matrix based on each inner product result. The inner product result represents the possibility that the head entity vector and the tail entity vector are the head and tail entity words (the first entity and the second entity) of the same triple phrase. Then, the computer device normalizes the values ​​of the similarity matrix to probability values ​​to obtain a relationship matrix representing the target relationship.

[0105] In an exemplary embodiment, the computer device includes a plurality of reference matrices of target relationships. The computer device performs similarity calculation on the reference matrix, the head vector matrix, and the tail vector matrix of each target relationship to obtain a similarity matrix representing the target relationship. Then, the computer device normalizes the value of each similarity matrix to a probability value to obtain a relationship matrix of the target relationship.

[0106] In an alternative embodiment, Figure 4 FIG. 4 is a flow chart of determining a matrix marking model in an exemplary embodiment. Figure 4 As shown, the process of determining the matrix marking model includes:

[0107] Step 401, obtaining each sample sentence.

[0108] Step 402: Label each sample sentence based on the 0-1 labeling method to obtain a sample binary matrix of the target relationship.

[0109] Step 403, preprocess each sample sentence to obtain a sample word identification list x={x1,x2,x3,…,xn}, where x represents the sample word identification list, n represents the number of words in the sample word identification list, and x1 represents the word id after preprocessing.

[0110] Step 404: input the sample word identification list into the BERT model, perform feature extraction on the sample word identification list based on the BERT model, and obtain sample output.

[0111] Step 405, calculate the loss value of the sample binary matrix and the sample output based on the weighted cross entropy loss function, and fine-tune the relationship matrix of the BERT model according to the calculated loss value, until the trained BERT model meets the preset training stop condition, and determine the trained BERT model as a matrix labeling model.

[0112] In this embodiment, the features of the word identifier list are extracted through the matrix labeling model to obtain the head vector matrix and the tail vector matrix, and the relationship matrix is ​​determined based on the head vector matrix, the tail vector matrix and the reference matrix of the target relationship. The adaptability of each sentence in the word vector matrix to the target relationship is clarified, which facilitates the subsequent extraction of target entity groups that are adapted to the target relationship.

[0113] In an exemplary embodiment, the entity includes a head entity and a tail entity, such as Figure 5 As shown, the specific processing process of extracting each entity in the relationship matrix according to the matrix labeling model in step 206 includes steps 502 to 506. Among them:

[0114] Step 502: Binarize the relationship matrix according to the entity threshold in the matrix labeling model to obtain an entity binarization matrix.

[0115] The matrix labeling model includes entity thresholds. The relationship matrix includes feature values ​​corresponding to each sentence in the word identification list, and the entity binarization matrix after binarization includes binarized entity feature values ​​corresponding to each sentence.

[0116] In implementation, the computer device determines whether each value in the relationship matrix is ​​greater than or equal to the entity threshold. If the value is greater than or equal to the entity threshold, the computer device updates the value to 1. If the value is less than the entity threshold, the computer device updates the value to 0. The computer device constructs an entity binary matrix based on each updated value.

[0117] Step 504 , performing dimensionality reduction processing on the entity binary matrix to obtain a head entity vector and a tail entity vector.

[0118] In implementation, the computer device performs addition and dimensionality reduction processing on the entity binary matrix in both row and column directions to obtain a head entity vector and a tail entity vector.

[0119] Specifically, the computer device performs addition and dimension reduction processing on the entity binary matrix in the row direction to obtain the head vector entity. At the same time, the computer device performs addition and dimension reduction processing on the entity binary matrix in the column direction to obtain the tail entity vector.

[0120] In an exemplary embodiment, if there are multiple entity binarization matrices corresponding to the target relationship, the computer device performs summing and dimensionality reduction processing on each entity binarization matrix in the row direction to obtain a head vector entity. At the same time, the computer device performs summing and dimensionality reduction processing on the entity binarization matrix in the column direction to obtain a tail entity vector.

[0121] Step 506: extract the head entity and tail entity corresponding to each sentence according to the head entity vector and the tail entity vector.

[0122] In implementation, the head entity vector includes the vector value of each sentence. The tail entity vector also includes the vector value of each sentence. The computer device determines the words whose vector values ​​are continuously greater than 1 in the sentence as the head entity for the vector value of each sentence in the head entity vector. The computer device determines the words whose vector values ​​are continuously greater than 1 in the sentence as the tail entity for the vector value of each sentence in the tail entity vector.

[0123] In an exemplary embodiment, the vector value of a sentence in the head entity vector is [101101110], and the computer device determines the word corresponding to the first "1" in the vector value as the head entity. Then, the computer device combines the words corresponding to the 3rd "1" and the 4th "1" in the head entity vector into a head entity. The computer device combines the words corresponding to the 6th "1", the 7th "1" and the 8th "1" in the head entity vector into a head entity. The vector value of a sentence in the tail entity vector is [01110001]. The computer device combines the words corresponding to the 2nd "1", the 3rd "1" and the 4th "1" in the tail entity vector into a tail entity. The computer device determines the word corresponding to the last "1" in the tail entity vector as a tail entity.

[0124] In this embodiment, entities of each sentence in the relationship matrix are extracted according to the matrix labeling model, and entities that may have target relationships are obtained. In addition, entities and target relationships are extracted from the relationship matrix at the same time, which reduces the propagation error of relationship extraction compared to extracting target relationships and entities separately. In addition, the simultaneous extraction of target relationships and entities in the present application upgrades the sequence labeling problem of traditional triples to the matrix labeling problem of relationships, which can effectively solve the problem of overlapping triples and improve the accuracy of data processing methods.

[0125] In an exemplary embodiment, each entity includes each head entity and each tail entity corresponding to each sentence, such as Figure 6 As shown, the specific processing process of determining the target entity group corresponding to the target relationship according to the target relationship represented by the relationship matrix and the relationship threshold in the matrix labeling model in step 206 includes steps 602 to 606. Among them:

[0126] Step 602: Binarize the relationship matrix according to the relationship threshold in the matrix labeling model to obtain a relationship binary matrix.

[0127] The matrix labeling model includes a relationship threshold, and the values ​​of the relationship threshold and the entity threshold are different.

[0128] In implementation, the computer device determines whether each value in the relationship matrix is ​​greater than or equal to the relationship threshold. If the value is greater than or equal to the relationship threshold, the computer device updates the value to 1. If the value is less than the relationship threshold, the computer device updates the value to 0. The computer device constructs a relationship binary matrix based on each updated value.

[0129] Step 604: for each sentence, combine the head entities and the tail entities corresponding to the sentence to obtain the entity groups corresponding to the sentence.

[0130] The entities corresponding to each sentence include head entities and tail entities.

[0131] During implementation, the computer device combines each head entity corresponding to each sentence with each tail entity corresponding to the sentence into entity groups.

[0132] In an exemplary embodiment, there are two sentences in the word identification list. The first sentence has the head entities of Big Head Son and Jin Yong. The tail entities are Small Head Dad and Author. The second sentence has the head entity of Hee Hee and the tail entity of Word. For the first sentence, the computer device combines Big Head Son with the tail entities Small Head Dad and Author corresponding to the sentence to obtain two entity groups of Big Head Son and Small Head Dad and Big Head Son and Author. Then, the computer device combines Jin Yong with the tail entities Small Head Dad and Author corresponding to the sentence to obtain two entity groups of Jin Yong and Small Head Dad and Jin Yong and Author. For the second sentence, the computer device combines the head entity Hee Hee and the tail entity Word corresponding to the second sentence to obtain the entity group of Hee Hee and Word.

[0133] Step 606 , screening each entity group according to the target relationship represented by the relationship binary matrix to obtain a target entity group corresponding to the target relationship.

[0134] In implementation, the computer device determines a binary sub-matrix corresponding to each entity group in the binary relationship matrix, and screens each entity group according to the value of the binary sub-matrix to obtain an entity group corresponding to the target relationship.

[0135] Specifically, the computer device determines the relationship binarization submatrix corresponding to each entity group in the relationship binarization matrix. Then, the computer device determines whether the relationship binarization submatrix represents the target relationship. If the relationship binarization submatrix represents the target relationship, the computer device determines the entity group as the target entity group corresponding to the target relationship.

[0136] In this embodiment, entity pairs and their relations are jointly extracted from unstructured text, thereby reducing the propagation error of relation extraction. In addition, by extracting target relations and entities simultaneously, the traditional triple sequence labeling problem is upgraded to a relation matrix labeling problem, which can effectively solve the problem of overlapping triples and improve the accuracy of the data processing method.

[0137] In an exemplary embodiment, Figure 7 As shown, the specific processing process of step 606 includes steps 702 to 706. Among them:

[0138] Step 702: Determine a relation binarization sub-matrix corresponding to each entity group in the relation binarization matrix.

[0139] In implementation, the computer device determines the relationship binarization sub-matrix corresponding to the entity group in the relationship binarization matrix according to the position of each entity group in the head entity vector and the tail entity vector.

[0140] Specifically, for each entity group, the computer device determines the range of the relationship binary submatrix rows in the relationship binary matrix according to the position of the head entity vector of the entity group. Then, the computer device determines the range of the relationship binary submatrix columns in the relationship binary matrix according to the position of the tail entity vector of the entity group.

[0141] In an exemplary embodiment, the entity group is Jin Yong and the author. Jin Yong is the word corresponding to the first "1" in the head entity vector. The author is the word corresponding to the eighth "1" in the tail entity vector. The computer device determines in the relationship binary matrix that the eighth value in the first row is the relationship binary matrix corresponding to the entity group Jin Yong and the author.

[0142] Step 704: determine whether the value of the binary relation submatrix satisfies the screening condition corresponding to the target relation.

[0143] Among them, the screening condition corresponding to the target relationship is that the values ​​of the binary submatrix of the relationship are all 1.

[0144] In implementation, the computer device determines whether the values ​​of the relation binary submatrix are all 1. If the values ​​of the relation binary submatrix are all 1, the computer device determines that the values ​​of the relation binary submatrix meet the screening conditions corresponding to the target relation, and determines that the relation binary submatrix represents the target relation.

[0145] In an optional embodiment, if the values ​​of the relation binary submatrix are not all 1, the computer device determines that the values ​​of the relation binary submatrix do not satisfy the screening condition corresponding to the target relation, and determines that the relation binary submatrix does not represent the target relation.

[0146] Step 706: If the value of the binary relation submatrix satisfies the screening condition, the entity group is determined as the target entity group corresponding to the target relation.

[0147] In implementation, if the value of the binary relation submatrix meets the screening condition of the target relation, the computer device determines the entity group as the target entity group corresponding to the target relation, thereby constructing a triplet of the target relation according to the target entity group.

[0148] In this embodiment, the target entity group is automatically extracted by binarizing the relation submatrix corresponding to each entity group and judging whether the entity group is the target entity group according to the value of the binarized relation matrix. Furthermore, each triple is constructed based on the target entity group and the target relation, and the triple is automatically constructed, thereby improving the efficiency of the data processing method.

[0149] In an exemplary embodiment, Figure 8 FIG. 1 is a flow chart of a method for extracting entity relationships according to an exemplary embodiment. Figure 8 As shown, the method for extracting entity relationships includes:

[0150] Step 801, obtain a word identifier list x.

[0151] Step 802: Based on the matrix tag model with 768 coding units, the word tag list is subjected to feature extraction to obtain a first word vector matrix Etail and a second word vector matrix Ehead. Wherein, Ehead, Etail∈Rl*768. Rl is the BERT encoder in the matrix tag model.

[0152] Step 803, add the word vector of the head tag unit in the first word vector matrix (the word vector of [cls]) to each word vector of the second word vector matrix to obtain the head vector matrix Ehead', and determine the first word vector matrix as the tail vector matrix Etail.

[0153] Step 804: input the head vector matrix Ehead' and the tail vector matrix Etail into the relation calculation layer, and perform similarity calculation on the head vector matrix and the tail vector matrix through the reference matrix of the target relation to obtain the relation matrix.

[0154] Step 805: extracting entities in the relationship matrix according to the matrix labeling model.

[0155] Step 806: Determine the target entity group corresponding to the target relationship according to the target relationship represented by the relationship matrix and the relationship threshold in the matrix labeling model.

[0156] Step 807: construct triples according to the target entity group and the target relationship.

[0157] In an exemplary embodiment, Fig. 9 FIG. 1 is a flow chart of data processing of an exemplary embodiment. Fig. 9 As shown, the data processing method includes:

[0158] Step 901, obtaining text data to be processed.

[0159] Step 902: pre-process the text data to be processed to obtain a word identifier list x.

[0160] Step 903: Perform feature extraction on the word identifier list based on the matrix optimization model to obtain a first word vector matrix and a second word vector matrix.

[0161] Step 904, determining a head vector matrix and a tail vector matrix based on the first word vector matrix and the second word vector matrix.

[0162] Step 905 , determining a relation matrix based on a reference matrix, a head vector matrix, and a tail vector matrix of the target relation.

[0163] Step 906: Binarize the relationship matrix based on the entity threshold to obtain an entity binarization matrix.

[0164] Step 907, performing dimensionality reduction processing on the entity binary matrix to obtain a head entity vector and a tail entity vector.

[0165] Step 908: The words whose values ​​in the head entity vector are continuously greater than 1 are regarded as head entities, and the words whose values ​​in the tail entity vector are continuously greater than 1 are extracted as tail entities.

[0166] Step 909: Binarize the relationship matrix based on the relationship threshold to obtain a binary relationship matrix.

[0167] Step 910: Combine each head entity and each tail entity to obtain each entity group.

[0168] Step 911 : determining a relation binarization submatrix corresponding to each entity group in the relation binarization matrix, and determining the entity group as a target entity group when all values ​​of the relation binarization submatrix are 1.

[0169] Step 912: construct triples according to the target relationship and the target entity group.

[0170] It should be understood that, although the various steps in the flowcharts involved in the above-mentioned embodiments are displayed in sequence according to the indication of the arrows, these steps are not necessarily executed in sequence according to the order indicated by the arrows. Unless there is a clear explanation in this article, the execution of these steps does not have a strict order restriction, and these steps can be executed in other orders. Moreover, at least a part of the steps in the flowcharts involved in the above-mentioned embodiments can include multiple steps or multiple stages, and these steps or stages are not necessarily executed at the same time, but can be executed at different times, and the execution order of these steps or stages is not necessarily to be carried out in sequence, but can be executed in turn or alternately with other steps or at least a part of the steps or stages in other steps.

[0171] Based on the same inventive concept, the embodiment of the present application also provides a data processing device for implementing the data processing method involved above. The implementation scheme for solving the problem provided by the device is similar to the implementation scheme recorded in the above method, so the specific limitations in the one or more data processing device embodiments provided below can refer to the limitations on the data processing method above, and will not be repeated here.

[0172] In an exemplary embodiment, Fig.10 As shown, a data processing device 1000 is provided, comprising: an acquisition module 1001, a determination module 1002, an extraction module 1003 and a construction module 1004, wherein:

[0173] The acquisition module 1001 is used to acquire the text data to be processed and pre-process the text data to be processed to obtain a word identification list.

[0174] The determination module 1002 is used to determine the relationship matrix based on the matrix labeling model and the word identification list.

[0175] The extraction module 1003 is used to extract the entities in the relationship matrix according to the matrix labeling model, and determine the target entity group corresponding to the target relationship according to the target relationship represented by the relationship matrix and the relationship threshold in the matrix labeling model.

[0176] Construction module 1004 is used to construct triples according to the target entity group and the target relationship; the triples are used to construct the knowledge graph.

[0177] In an exemplary embodiment, the acquisition module 1001 includes a first acquisition submodule and a first processing submodule. The first processing submodule includes:

[0178] The first adding submodule is used for adding a head mark at the beginning of each initial sentence in the text data to be processed, and adding a tail mark at the end of the initial sentence to obtain a sentence.

[0179] The second processing submodule is used to pre-process the sentences according to a preset sentence length and construct a word identification list according to each pre-processed sentence.

[0180] In an exemplary embodiment, the determination module 1002 includes:

[0181] The first extraction submodule is used to perform feature extraction on the word identification list according to the matrix labeling model to obtain a first word vector matrix and a second word vector matrix; the matrix labeling model contains a reference matrix of the target relationship.

[0182] The second adding submodule is used to add the word vector of the head tag unit in the first word vector matrix to each word vector of the second word vector matrix to obtain the head vector matrix, and determine the first word vector matrix as the tail vector matrix.

[0183] The first calculation submodule is used to perform similarity calculation on the reference matrix, the head vector matrix and the tail vector matrix of the target relationship to obtain a similarity matrix, and normalize the similarity matrix to obtain a relationship matrix.

[0184] In an exemplary embodiment, the entity includes a head entity and a tail entity, and the extraction module 1003 includes a second extraction submodule and a first determination submodule. The second extraction submodule includes:

[0185] The third processing submodule is used to perform binarization processing on the relationship matrix according to the entity threshold in the matrix labeling model to obtain an entity binarization matrix.

[0186] The fourth processing submodule is used to perform dimensionality reduction processing on the entity binary matrix to obtain a head entity vector and a tail entity vector.

[0187] The second extraction submodule is used to extract the head entity and the tail entity corresponding to each sentence according to the head entity vector and the tail entity vector.

[0188] In an exemplary embodiment, each entity includes each head entity and each tail entity corresponding to each sentence, and the extraction module 1003 includes a second extraction submodule and a first determination submodule. The first determination submodule includes:

[0189] The fourth processing submodule is used to perform binarization processing on the relationship matrix according to the relationship threshold in the matrix labeling model to obtain a relationship binarization matrix.

[0190] The combination submodule is used to combine the head entities and the tail entities corresponding to each sentence to obtain the entity groups corresponding to the sentence.

[0191] The first screening submodule is used to screen each entity group according to the target relationship represented by the relationship binary matrix to obtain the target entity group corresponding to the target relationship.

[0192] In an exemplary embodiment, the first screening submodule includes:

[0193] The second determination submodule is used to determine the relationship binarization submatrix corresponding to each entity group in the relationship binarization matrix.

[0194] The first judgment submodule is used to judge whether the value of the relationship binary submatrix meets the screening condition corresponding to the target relationship.

[0195] The third determination submodule is used to determine the entity group as the target entity group corresponding to the target relationship if the value of the relationship binary submatrix meets the screening condition.

[0196] Each module in the above data processing device can be implemented in whole or in part by software, hardware or a combination thereof. Each module can be embedded in or independent of a processor in a computer device in the form of hardware, or can be stored in a memory in a computer device in the form of software, so that the processor can call and execute operations corresponding to each module.

[0197] In an exemplary embodiment, a computer device is provided. The computer device may be a terminal, and its internal structure diagram may be as shown in FIG. Fig.11As shown. The computer device includes a processor, a memory, an input / output interface, a communication interface, a display unit and an input device. The processor, the memory and the input / output interface are connected through a system bus, and the communication interface, the display unit and the input device are connected to the system bus through the input / output interface. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The input / output interface of the computer device is used to exchange information between the processor and an external device. The communication interface of the computer device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner can be implemented through WIFI, a mobile cellular network, near field communication (NFC) or other technologies. When the computer program is executed by the processor, a data processing method is implemented. The display unit of the computer device is used to form a visually visible picture, which can be a display screen, a projection device or a virtual reality imaging device. The display screen can be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device can be a touch layer covering the display screen, or a button, trackball or touchpad set on the computer device shell, or an external keyboard, touchpad or mouse.

[0198] Those skilled in the art will understand that Fig.11 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0199] In one embodiment, a computer device is further provided, including a memory and a processor, wherein a computer program is stored in the memory, and the processor implements the steps in the above method embodiments when executing the computer program.

[0200] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the steps in the above-mentioned method embodiments are implemented.

[0201] In one embodiment, a computer program product is provided, including a computer program, which implements the steps in the above method embodiments when executed by a processor.

[0202] A person of ordinary skill in the art can understand that all or part of the processes in the above-mentioned embodiment method can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to the memory, database or other medium used in the embodiments provided in the present application can include at least one of non-volatile memory and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. As an illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM). The database involved in each embodiment provided in this application may include at least one of a relational database and a non-relational database. Non-relational databases may include distributed databases based on blockchains, etc., but are not limited to this. The processor involved in each embodiment provided in this application may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, an artificial intelligence (AI) processor, etc., but are not limited to this.

[0203] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this application.

[0204] The above-described embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the present application. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the attached claims.

Claims

1. A data processing method, characterized in that: The method comprises: Acquire text data to be processed, and preprocess the text data to be processed to obtain a word identifier list; Determining a relationship matrix based on the matrix labeling model and the word identification list; Extracting each entity in the relationship matrix according to the matrix labeling model, and determining a target entity group corresponding to the target relationship according to the target relationship represented by the relationship matrix and a relationship threshold in the matrix labeling model; A triple is constructed according to the target entity group and the target relationship; the triple is used to construct a knowledge graph.

2. The method according to claim 1, characterized in that The preprocessing of the text data to be processed to obtain a word identification list includes: For each initial sentence in the text data to be processed, a head mark is added at the beginning of the initial sentence, and a tail mark is added at the end of the initial sentence to obtain a sentence; The sentences are preprocessed according to a preset sentence length, and a word identification list is constructed based on each preprocessed sentence.

3. The method according to claim 1, characterized in that The determining of the relationship matrix based on the matrix labeling model and the word identification list includes: Performing feature extraction on the word identification list according to a matrix tagging model to obtain a first word vector matrix and a second word vector matrix; the matrix tagging model includes a reference matrix of the target relationship; Add the word vector of the head tag unit in the first word vector matrix to each word vector of the second word vector matrix to obtain a head vector matrix, and determine the first word vector matrix as a tail vector matrix; The similarity is calculated for the reference matrix of the target relationship, the head vector matrix and the tail vector matrix to obtain a similarity matrix, and the similarity matrix is ​​normalized to obtain a relationship matrix.

4. The method according to claim 1, characterized in that: The entity includes a head entity and a tail entity, and the extracting each entity in the relationship matrix according to the matrix labeling model includes: According to the entity threshold in the matrix labeling model, the relationship matrix is ​​binarized to obtain an entity binarization matrix; Performing dimensionality reduction processing on the entity binary matrix to obtain a head entity vector and a tail entity vector; The head entity and the tail entity corresponding to each sentence are extracted according to the head entity vector and the tail entity vector.

5. The method according to claim 1, characterized in that Each of the entities includes each head entity and each tail entity corresponding to each sentence, and determining the target entity group corresponding to the target relationship according to the target relationship represented by the relationship matrix and the relationship threshold in the matrix labeling model includes: According to the relationship threshold in the matrix labeling model, the relationship matrix is ​​binarized to obtain a relationship binarization matrix; For each of the sentences, combining the head entities and the tail entities corresponding to the sentence to obtain entity groups corresponding to the sentence; According to the target relationship represented by the relationship binary matrix, each of the entity groups is screened to obtain a target entity group corresponding to the target relationship.

6. The method according to claim 5, characterized in that The step of screening each of the entity groups according to the target relationship represented by the relationship binary matrix to obtain a target entity group corresponding to the target relationship includes: Determining a relation binarization submatrix corresponding to each entity group in the relation binarization matrix; Determine whether the value of the relationship binary submatrix meets the screening condition corresponding to the target relationship; If the value of the relationship binary submatrix meets the screening condition, the entity group is determined as the target entity group corresponding to the target relationship.

7. A data processing device, characterized in that: The device comprises: An acquisition module, used for acquiring text data to be processed, and preprocessing the text data to be processed to obtain a word identification list; A determination module, configured to determine a relationship matrix based on a matrix labeling model and the word identification list; An extraction module, used to extract each entity in the relationship matrix according to the matrix labeling model, and determine a target entity group corresponding to the target relationship according to the target relationship represented by the relationship matrix and the relationship threshold in the matrix labeling model; A construction module is used to construct triples according to the target entity group and the target relationship; the triples are used to construct a knowledge graph.

8. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 6 are implemented.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.

10. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.

Images