Data processing method, device and computer program product

Abstract

The invention relates to a data processing method and device and a computer program product. The method comprises the following steps: under the condition that a first query object comprising a target vocabulary is received, obtaining an associated vocabulary having an association relationship with the target vocabulary; adjusting the target vocabulary in the first query object based on the associated vocabulary to obtain a second query object; and obtaining a target query result based on the second query object. The query intention indicated by the first query object can be clearer, so that the accuracy of the obtained target query result and the correlation with the first query object are improved.

Description

Technical Field

[0001] The present disclosure relates to but is not limited to the field of information technology, and in particular to a data processing method, device, and computer program product. Background Art

[0002] Data query refers to querying and analyzing the data in the database based on the query object input by the user, and then obtaining the query result. For example, you can perform a query operation on the input query object and obtain the answer corresponding to the query object.

[0003] However, when there is incomprehensible content in the query object, the output answer may be inaccurate or query results irrelevant to the query object may be obtained. Summary of the invention

[0004] To overcome the problems existing in the related art, the present disclosure provides a data processing method, device and computer program product, which can more clearly define the query intent indicated by the first query object, thereby improving the accuracy of the obtained target query result and its relevance to the first query object.

[0005] According to a first aspect of an embodiment of the present disclosure, there is provided a data processing method, including:

[0006] When receiving a first query object including a target word, obtaining associated words having an associated relationship with the target word;

[0007] Adjusting the target vocabulary in the first query object based on the associated vocabulary to obtain a second query object;

[0008] Based on the second query object, a target query result is obtained.

[0009] According to a second aspect of an embodiment of the present disclosure, there is provided a data processing device, including:

[0010] A first acquisition module is configured to acquire associated words having an associated relationship with the target word when receiving a first query object including the target word;

[0011] an adjustment module, configured to adjust the target vocabulary in the first query object based on the associated vocabulary to obtain a second query object;

[0012] The generating module is configured to obtain a target query result based on the second query object.

[0013] According to a third aspect of an embodiment of the present disclosure, there is provided an electronic device, including:

[0014] processor;

[0015] Memory for storing computer programs or instructions;

[0016] The processor executes the computer program or instructions to implement the steps of any one of the methods in the first aspect above.

[0017] According to a fourth aspect of an embodiment of the present disclosure, a non-temporary computer-readable storage medium is provided, wherein the storage medium stores a computer program or instructions. When the computer program or instructions in the storage medium are executed by a processor, the steps of the method described in any one of the first aspects are implemented.

[0018] According to a fifth aspect of an embodiment of the present disclosure, a computer program product is provided, including a computer program or instructions, which, when executed by a processor, implement the steps of any one of the methods described in the first aspect.

[0019] The technical solution provided by the embodiments of the present disclosure may have the following beneficial effects:

[0020] In the case where there is an incomprehensible target word in the first query object, the present disclosure can obtain the associated words of the target word, rewrite the first query object based on the associated words to obtain the second query object, and then obtain the target query result based on the second query object. Compared with obtaining the target query result directly based on the first query object, by combining the associated words, the query intention indicated by the first query object can be more clearly defined, thereby improving the accuracy of the obtained target query result and the relevance to the first query object.

[0021] It is to be understood that the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the present disclosure. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present disclosure and, together with the description, serve to explain the principles of the present disclosure.

[0023] Figure 1 A data processing method according to an exemplary embodiment is shown in FIG. Figure 1 .

[0024] Figure 2 The figure is a schematic diagram of enterprise jargon according to an exemplary embodiment.

[0025] Figure 3 The figure is a schematic diagram showing the association relationship between target words and associated words according to an exemplary embodiment.

[0026] Figure 4A data processing method according to an exemplary embodiment is shown in FIG. Figure 2 .

[0027] Figure 5 A data processing method according to an exemplary embodiment is shown in FIG. Figure 3 .

[0028] Figure 6 It is a block diagram of a data processing device according to an exemplary embodiment.

[0029] Figure 7 The diagram is a schematic structural diagram of an electronic device according to an exemplary embodiment. DETAILED DESCRIPTION

[0030] Exemplary embodiments will be described in detail herein, examples of which are shown in the accompanying drawings. When the following description refers to the drawings, the same numbers in different drawings represent the same or similar elements unless otherwise indicated. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present disclosure. Instead, they are merely examples of devices and methods consistent with some aspects of the present disclosure as detailed in the appended claims.

[0031] Figure 1 A data processing method according to an exemplary embodiment is shown in FIG. Figure 1 .like Figure 1 As shown, the method mainly includes the following steps:

[0032] In step 101, when a first query object including a target word is received, an associated word having an associated relationship with the target word is obtained;

[0033] In step 102, the target words in the first query object are adjusted based on the associated words to obtain a second query object;

[0034] In step 103, a target query result is obtained based on the second query object.

[0035] The data processing method proposed in the present disclosure can be applied to electronic devices or servers. Here, the electronic device may include: terminal devices, for example, mobile terminals or fixed terminals. Among them, mobile terminals may include: mobile phones, tablet computers, laptop computers and other devices. Fixed terminals may include: desktop computers or smart TVs, etc. The data processing method in the embodiment of the present disclosure may be configured in a data processing device, which may be set in a server or in an electronic device, and the embodiment of the present disclosure does not limit this.

[0036] The execution subject of the embodiments of the present disclosure may be, for example, a central processing unit (CPU) in a server or electronic device in terms of hardware, or may be, for example, a related background service in a server or electronic device in terms of software, without limitation.

[0037] In some embodiments, the data processing method of the present disclosure can be applied to a system that needs to parse and respond to the intent of a query object input by a user. Exemplarily, the data processing method of the present disclosure can be applied to a question-and-answer system or a conversation system.

[0038] It should be noted that, in the case of using the question-answering system in the above example, the question-answering system can be a system with a retrieval-augmented generation (RAG) function, that is, it can be a query system including a RAG architecture. The RAG architecture is a language model architecture that combines retrieval technology and generation technology.

[0039] In some embodiments, the question-answering system may include a retrieval module and a generation module. When performing a question-answering task, the retrieval module may be used to retrieve text fragments related to the query statement (question) to be processed from a preset corpus (knowledge base), and then the generation module may be used to generate an answer based on the retrieved text fragments and context information of the query statement to be processed. The query statement to be processed may be a question input by a user, and the text fragment may also be called a document fragment.

[0040] Exemplarily, the question-answering system is a RAG system, which first uses a retrieval module to retrieve text fragments related to the question from a preset corpus (e.g., a large text library), and then uses a generation module to generate a final answer based on the retrieved text fragments and the context information of the question.

[0041] In some embodiments, the predetermined knowledge source can be preprocessed based on the question-answering system, and then the preprocessed data can be converted into vectors and stored as a data source that the question-answering system needs to refer to, that is, a knowledge base. After receiving the query statement to be processed, the question-answering system can determine the user's question intention based on the query statement to be processed, and perform word segmentation on the query statement to be processed, and then retrieve relevant text fragments from the knowledge base based on the result of the word segmentation, sort the retrieved text fragments, and the question-answering system obtains the final answer based on the sorting result and the text fragments.

[0042] Exemplarily, the RAG system can first prepare knowledge: filter and / or segment the predetermined knowledge sources (e.g., documents), and convert the results of the filtering and / or segmentation into vectors and store them as a data source for the big model to answer references, i.e., the knowledge base; then the sentences input by the user can be processed, the big model can determine the user's question intention, and perform word segmentation on the sentences input by the user, and then perform retrieval based on the results of the word segmentation, and sort the retrieved text fragments, and obtain the final answer based on the sorting results and the text fragments.

[0043] In some embodiments, text fragments with higher relevance, timeliness, authority, etc. in the knowledge base can be recalled preferentially, the text fragments can be sorted, and the sorting results and the corresponding text fragments can be input into the big model to control the big model to answer based on the text fragments (recall fragments), thereby improving the accuracy, explainability and timeliness of the answers output by the model.

[0044] In the disclosed embodiment, when the question-answering system receives a query object, it can identify whether the query object includes the target vocabulary, and determine the query object including the target vocabulary as the first query object. It should be noted that the query object can be an object input by the user for performing a query operation, the query object can include vocabulary, or the content included in the query object can be recognized and understood as including the target vocabulary, for example, the expression of a smiley face can be understood as happy or smiley face, and if the target vocabulary is happy, the expression of the smiley face can be understood as including the target vocabulary.

[0045] In some embodiments, the format and type of the query object are related to the entry point for the user to input the query object or the user's needs. The query object may include content in one format, or may be composed of a combination of content in different formats. Exemplarily, the query object may include at least one of the following: content in audio format; content in text format; content in image format. For example, the query object may include at least one of the following: audio, image, text, emoticon, or dynamic image.

[0046] In some other embodiments, the query object may also be a query statement. Taking the query object as an example, the query statement may include at least one of the following: a query statement in voice format, a query statement in text format, and a query statement in image format.

[0047] In some embodiments, the question-answering system of the present disclosure includes not only the RAG system, but also all modules with retrieval and query functions. That is to say, the technical solution of the present disclosure can be applied not only to the retrieval part in the RAG system, but also to all modules with retrieval and query functions. For example, it can be applied to other searches that broadly involve a given data source, such as searching for documents, searching for emails, etc.

[0048] Taking the application of the data processing method of the present disclosure to a question-answering system as an example, the target vocabulary may include vocabulary that the question-answering system cannot understand or can understand but cannot accurately conduct in-depth analysis. In some embodiments, a vocabulary library may be created in advance, and the target vocabulary may be stored in the pre-created vocabulary library. When the question-answering system receives a query object, the query object may be analyzed to determine each vocabulary in the query object, and each determined vocabulary may be compared with each target vocabulary in the vocabulary library, and the vocabulary in the query object that exists in the vocabulary library may be determined as the target vocabulary.

[0049] The target vocabulary may include proprietary vocabulary. Exemplarily, the target vocabulary may include non-standard vocabulary with directionality, corresponding to different scenarios and / or groups, respectively, and the intended information indicated by the target vocabulary in different scenarios and / or groups may be different. Exemplarily, the target vocabulary may include corporate jargon, which can be understood as an informal, industry-specific term or expression popular in a specific industry or enterprise, facilitating efficient communication among people in the industry. Exemplarily again, the target vocabulary may include common network terms.

[0050] Taking the correspondence between target vocabulary and group as an example, the target vocabulary may include words or terms with normative attributes within the group, which have "encrypted" characteristics relative to external members. Figure 2 is a schematic diagram of corporate jargon according to an exemplary embodiment. Figure 2 As shown, corporate jargon can be a set of conventional communication rules that are frequently used within the enterprise, i.e., corporate jargon 201 (knowledge structure) or industry terminology 202 (vertical vocabulary), which is relatively difficult for people outside the enterprise or the industry to which the enterprise belongs to understand. Among them, vertical vocabulary includes proprietary vocabulary used in specific fields or industries.

[0051] Taking Internet company A as an example, members of company A can use some of the terms in the Internet industry to communicate, such as "front end" and "granularity", among which "granularity" is not easy to understand for people outside the industry. Members of company A can also use abbreviations from the internal knowledge structure of the enterprise. These words correspond to the internal organizational structure, product names or business lines of the enterprise, such as "middle platform" and "KVP". Among these abbreviations, "KVP" is not easy to understand for people outside the enterprise.

[0052] In some embodiments, target words that the large model cannot understand or can understand but cannot accurately conduct in-depth analysis can be classified according to the target word's expression form. For example, in the RAG program, you can first perform a search and then execute the generation step. For special references in certain industries, such as "cola combination", the large model is likely to understand it, that is, it can answer that it represents a combination of two drugs, but in a specific question-and-answer scenario, this word cannot be directly used for retrieval, which prevents the large model from generating the correct answer.

[0053] For example, when the target vocabulary is an abbreviation, if the abbreviation is common in the industry (public domain knowledge), the big model is likely to understand it. If the abbreviation is used internally by the enterprise (private domain knowledge), the big model is likely to be unable to understand it. The classification of target vocabulary is shown in Table 1:

[0054] Table 1

[0055]

[0056] Based on Table 1, we can see that the target vocabulary can be divided into three categories, namely special reference, vocabulary meaning, and knowledge structure. Among them, the application scenario of special reference is the correspondence between words, and the difficulty coefficient is level 4; the application scenario of vocabulary meaning is to determine the representational meaning of vocabulary, and the difficulty coefficient is level 2; the application scenario of knowledge structure is the parallel relationship between two or more words, or the inclusion relationship between two or more words, and the difficulty coefficient is level 5.

[0057] In the disclosed embodiment, after determining the first query object including the target vocabulary, the associated vocabulary having an associated relationship with the target vocabulary can be obtained. It should be noted that the associated relationship can be used to indicate the degree of association between vocabulary. When the degree of association between two vocabulary is greater than or equal to a preset association threshold, it can be determined that the two vocabulary have an associated relationship. When it is determined that the two vocabulary have an associated relationship, the type of the association relationship between the two vocabulary, i.e., the association type, can also be determined based on the magnitude of the association between the two vocabulary. In other embodiments, when the degree of association between the two vocabulary is less than a preset association threshold, it can be determined that there is no association relationship between the two vocabulary. Among them, the degree of association between the vocabulary can be determined based on a semantic analysis model. For example, the degree of association between the vocabulary can be determined based on the semantic similarity between the vocabulary, which is not specifically limited here.

[0058] Exemplarily, the association relationship between the associated vocabulary and the target vocabulary may include at least one of a hierarchical relationship, a same-level relationship, and a semantically related relationship, wherein the hierarchical relationship may include a superior relationship and a subordinate relationship, and the same-level relationship may include a parallel relationship or a close relationship. Exemplarily, the associated vocabulary may be a superior vocabulary, a subordinate vocabulary, a same-level word, a synonym, a near-synonym, a semantically related word, etc. of the target vocabulary.

[0059] Figure 3 FIG. 1 is a schematic diagram showing the association relationship between target words and associated words according to an exemplary embodiment. Figure 3 As shown, the associated words of the target word AA may include: parallel words BB, hierarchical words XX and AAa, synonyms AB, and related words AC and BC.

[0060] In some embodiments, a vocabulary can be created in advance, and the target vocabulary, associated vocabulary, and the association relationship between the associated vocabulary and the target vocabulary can be stored in the vocabulary. In other embodiments, descriptive information for the target vocabulary can also be obtained, wherein the descriptive information includes information for explaining the vocabulary. Exemplarily, the target vocabulary, associated vocabulary, and the association relationship can be stored in different entries in the vocabulary based on a preset data structure, such as the target vocabulary, associated vocabulary, and the association relationship can be stored in different entries in the vocabulary in the form of entries. For example, in the vocabulary, the data structure of each storage entry can be expressed as: word-explanation of the word-synonym / nearbyword / same-level word / superior-subordinate word. In the disclosed embodiment, the target vocabulary, associated vocabulary, and the association relationship can be stored in the vocabulary according to the following data structure: target vocabulary-descriptive information-associated vocabulary.

[0061] In some embodiments, the data in the word library (target words, associated words, and associated relationships) can be obtained from a variety of data sources, and can be an industry word library from an external source, such as from an open source website, or can be a word-to-word relationship manually constructed from within an enterprise. Exemplarily, an industry word library can be constructed based on industry words obtained from the outside, for example, the target words, associated words, and associated relationships can be obtained from an open source website. Another exemplary method is to manually construct an associated relationship between the target words and associated words based on personnel within the enterprise.

[0062] In other embodiments, corresponding lexicons may be created for different business scenarios. When the first query object is received, the current business scenario may be determined, and the lexicon for the current business scenario may be determined from a preset lexicon based on the current business scenario. In some embodiments, during the testing phase, lexicons for each business scenario may be created based on data obtained in different business scenarios, and lexicons for each business scenario may be pre-stored. Exemplarily, a mapping relationship may be established between each lexicon and the corresponding business scenario, so that when the business scenario is determined, the corresponding lexicon may be determined based on the business scenario and the mapping relationship.

[0063] In other embodiments, when the first query object is received, the current business scenario can also be determined based on the historical query information of the user who initiated the first query object. Exemplarily, a user tag can be generated based on the user's historical query information, and a vocabulary of the business scenario can be determined based on the user tag, wherein the vocabulary of the business scenario can be pre-identified based on the user tag.

[0064] In other embodiments, when the first query object is received, the question-and-answer system can determine the industry label corresponding to the user, and determine the vocabulary of the business scenario based on the industry label. Exemplarily, the industry vocabulary can be pre-built, that is, the vocabulary of the business scenario is pre-identified based on the industry label. Exemplarily, the user named the knowledge base as School Information, and also set prompt words related to school ranking consultation in the RAG configuration process, and imported a file named XX School Ranking. In the construction process, it can be known that the user may be in the education industry, and the vocabulary of the education industry can be pre-stored for the user.

[0065] In other embodiments, selection items may also be displayed on the front-end page, and the user may manually select a business scenario (eg, industry) or manually import a vocabulary.

[0066] In some embodiments, upon receiving a first query object including a target vocabulary, associated vocabulary that has an association relationship with the target vocabulary is obtained, including: upon receiving the first query object, determining a current business scenario based on the first query object, and determining a vocabulary library for the current business scenario; determining vocabulary in the vocabulary library of the current business scenario that has an association relationship with the target vocabulary as associated vocabulary of the target vocabulary in the current business scenario.

[0067] In the embodiment of the present disclosure, when the first query object is received, a semantic analysis can be performed on the first query object to obtain a semantic analysis result, and the current business scenario can be determined based on the semantic analysis result. After the current business scenario is determined, the vocabulary of the current business scenario can be determined. For example, if the current business scenario is a scenario of the pharmaceutical industry, the vocabulary of the pharmaceutical scenario can be determined; if the current business scenario is a scenario of the Internet industry, the vocabulary of the Internet scenario can be determined.

[0068] After determining the vocabulary of the current business scenario, the target vocabulary can be compared with each vocabulary pre-stored in the vocabulary of the current business scenario to determine a preset vocabulary that matches the target vocabulary, determine the vocabulary associated with the preset vocabulary as the vocabulary associated with the target vocabulary, and determine the vocabulary associated with the target vocabulary as the associated vocabulary of the target vocabulary in the current business scenario. The preset vocabulary that matches the target vocabulary may include: a vocabulary whose similarity with the target vocabulary is less than a preset similarity threshold, for example, a preset vocabulary that is the same as the target vocabulary.

[0069] In the disclosed embodiment, after receiving the first query object, the current business scenario can be determined based on the first query object, and the vocabulary of the current business scenario can be queried based on the target vocabulary, and then the vocabulary in the current business scenario that has an association relationship with the target vocabulary can be determined as the associated vocabulary of the target vocabulary in the current business scenario.

[0070] For example, the business scenario (question and answer scenario) of X company is the consulting business of the education industry. Then there may be a series of proper nouns about college majors or special terms in the education industry. In the configuration of the question and answer system, the abbreviations and aliases of the proper words can be preset, and a vocabulary can be built based on the abbreviations and aliases of the proper words. In the use of the question and answer system, the vocabulary of the consulting scenario is queried based on the target vocabulary, and the associated vocabulary of the target vocabulary in the consulting scenario is determined, thereby improving the question and answer effect of the consulting scenario.

[0071] By matching the word library with the business scenario, after determining the business scenario, you can directly search the word library of the business scenario based on the target vocabulary. Compared with storing all data in a database, it can improve the efficiency of data processing and improve the accuracy of the related vocabulary determined. By strongly correlating the content in the word library with the business scenario (question and answer scenario), the question and answer effect in the business scenario can be optimized.

[0072] In some embodiments, the method further comprises:

[0073] In different business scenarios, initial data is obtained from at least one data source in each business scenario;

[0074] When the initial data includes the target vocabulary, analyzing the text content in the initial data to obtain the association information between each vocabulary in the initial data and the target vocabulary;

[0075] Based on the association information, associated words are determined from the initial data, and based on the association relationship between the associated words and the target words and / or the description information of the target words, a vocabulary of the business scenario is constructed.

[0076] In the embodiments of the present disclosure, for different business scenarios, initial data can be obtained from at least one data source in the business scenario. The business scenario can also be called a question-and-answer scenario. Exemplarily, it can include: a consulting scenario, an Internet scenario, a pharmaceutical industry scenario, etc.

[0077] For each business scenario, there are multiple ways to obtain data sources. For example, for an Internet scenario, initial data can be obtained from within the enterprise and from open source websites.

[0078] After obtaining the initial data, the text content in the initial data can be analyzed to obtain the association information between each word in the initial data and the target word, and the associated words can be determined from the initial data based on the association information. The association information can include: the degree of association between each word in the initial data and the target word.

[0079] In other embodiments, description information of the target vocabulary can be obtained while obtaining the association information between each vocabulary in the initial data and the target vocabulary; or description information of the target vocabulary can be obtained before obtaining the association information between each vocabulary in the initial data and the target vocabulary; or description information of the target vocabulary can be obtained after obtaining the association information between each vocabulary in the initial data and the target vocabulary.

[0080] In some embodiments, initial data can be obtained from an external industry thesaurus by associating it with an external industry thesaurus, and the thesaurus can be pre-configured in the question-and-answer system based on the initial data, without the need for the business side to import it by itself. For example, in the medical industry, complex drug names can be corresponded to by abbreviations, and the "Cola combination" refers to the combined use of two drugs, Lenvatinib and Pembrolizumab. This industry-wide knowledge can be obtained from a series of pharmaceutical and medical-related paper websites or public databases.

[0081] In other embodiments, initial data can be obtained through the set reference or vocabulary structure relationship within the group (company), and the vocabulary library can be obtained based on the initial data. For example, the vocabulary can be manually imported on the business side to obtain the vocabulary library; or the vocabulary library can be automatically extracted and constructed based on the documents and historical chat sessions within the group.

[0082] Taking corporate jargon as an example, since some corporate jargon and expressions similar to corporate jargon appear in historical chat sessions, the large model can be used to analyze group chat sessions to extract corporate jargon and its description information (explanation). For example, in the chat session, it was found that "XX Membership Package" and "XX Membership Package pro" appeared in parallel, and there was no inclusion relationship between the two. Then the corresponding relationship (association relationship) of "Membership Package-XX Membership Package / XX Membership Package pro (parallel words)" can be extracted, where Membership Package is the target word, and "XX Membership Package" and "XX Membership Package pro" are associated words of the target word.

[0083] In other embodiments, entity words in a document can be automatically extracted based on the paragraph structure of the document. For example, the relationship structure (association relationship) of words can be extracted through the title structure in the document. For example, the first-level title is: "Basic Dimensions of Environmental Assessment", and the entity "Environmental Assessment" is extracted. There are multiple second-level titles under the first-level title, "Talent Training", "Organizational Relationship", and the corresponding relationship (association relationship) of "Environmental Assessment-Talent Training / Organizational Relationship (Subordinate Word)" can be extracted, among which environmental assessment is the target word, and "Talent Training" and "Organizational Relationship" are associated words of the target word.

[0084] In some embodiments, in order to improve the accuracy of the vocabulary constructed based on the automatically extracted vocabulary, after the vocabulary is obtained, the constructed vocabulary can be verified and checked based on the business scenario.

[0085] In other embodiments, the vocabulary may be classified based on the source of the initial data, and the vocabulary may include words of general knowledge with industry attributes.

[0086] In the disclosed embodiment, a vocabulary can be constructed for different business scenarios. By matching the vocabulary with the business scenario, after determining the business scenario, the vocabulary of the business scenario can be directly searched based on the target vocabulary. Compared with storing all data in a database, the efficiency of processing data can be improved, and the accuracy of the determined related vocabulary can be improved. By strongly correlating the content in the vocabulary with the business scenario (question and answer scenario), the question and answer effect in the business scenario can be optimized.

[0087] In some embodiments, the method further comprises:

[0088] In different business scenarios, based on the target vocabulary in the preset object, determine the associated vocabulary of the target vocabulary from the vocabulary library of the business scenario;

[0089] Determining a first evaluation value of the vocabulary of the business scenario based on the number of preset objects of the associated vocabulary and the total number of preset objects; wherein the first evaluation value is used to indicate the performance of the vocabulary of the business scenario;

[0090] When the first evaluation value is less than the preset threshold, the vocabulary of the business scenario is updated until the first evaluation value is greater than or equal to the preset threshold.

[0091] It should be noted that the preset object can be a query object for testing, for example, a test sentence. After determining the target vocabulary in the preset object, the associated vocabulary of the target vocabulary can be determined from the vocabulary of the business scenario based on the target vocabulary in the preset object. In the case where there is no vocabulary associated with the target vocabulary in the vocabulary of the business scenario, the associated vocabulary of the target vocabulary cannot be determined. Therefore, some preset objects can determine the associated vocabulary, while some preset objects cannot determine the associated vocabulary.

[0092] The present disclosure can count whether the number of preset objects of associated vocabulary is determined, obtain the number of preset objects of associated vocabulary, and determine the first evaluation value of the vocabulary of the business scenario based on the number of preset objects of associated vocabulary and the total number of preset objects. The first evaluation value can be used to indicate the success rate of rewriting the target vocabulary in the preset objects and the coverage rate of the target vocabulary in the vocabulary.

[0093] In some embodiments, a preset object including a target vocabulary can be input into a question-and-answer system. After the business scenario is determined, it can be determined whether the vocabulary of the business scenario includes associated vocabulary of the target vocabulary. If the vocabulary of the business scenario includes associated vocabulary of the target vocabulary and the target vocabulary is successfully rewritten based on the associated vocabulary, it is determined that associated vocabulary of the preset object exists in the vocabulary of the business scenario.

[0094] For example, the industry vocabulary set up for educational scenarios includes proper nouns about college majors or special terms in the education industry. For example, EPFL is the abbreviation of the Federal Institute of Technology in Lausanne, and there is a corresponding relationship between EPFL and the Federal Institute of Technology in Lausanne in the vocabulary. During the test, the preset objects include the colloquial abbreviation "EPFL". If "EPFL" is successfully rewritten as "Federal Institute of Technology in Lausanne" in the question-and-answer system, it means that the vocabulary of the consulting scenario covers the target vocabulary.

[0095] In some embodiments, the word library may be subjected to multiple rounds of testing, and after the multiple rounds of testing, the test results may be annotated. Exemplarily, multiple initial evaluation values ​​may be obtained, and a first evaluation value may be obtained based on the multiple initial evaluation values. For example, the first evaluation value may be obtained based on the average of the multiple initial evaluation values; or, the first evaluation value may be obtained based on the median of the multiple initial evaluation values.

[0096] In some embodiments, the first evaluation value may be determined based on the ratio between the number of preset objects of the associated vocabulary and the total number of preset objects. Assuming there are N preset objects and M preset objects successfully rewrite the target vocabulary, the calculation formula of the first evaluation value may be as follows:

[0097]

[0098] In formula (1), A1 represents the first evaluation value; M represents the number of preset objects that have successfully rewritten the target vocabulary; and N represents the total number of preset objects.

[0099] In the embodiment of the present disclosure, when the first evaluation value is less than a preset threshold, the vocabulary of the business scenario may be updated until the first evaluation value is greater than or equal to the preset threshold.

[0100] It should be noted that, along with the updating of industry knowledge and the company's business knowledge and product knowledge, the first evaluation value of the vocabulary of the business scenario can be regularly tested and tracked (the first evaluation value is determined according to a preset monitoring period), and when the first evaluation value is less than a preset threshold, the vocabulary of the business scenario can be updated. For example, inaccurate rewriting results can be calibrated in real time through manual intervention to improve the performance of the vocabulary of the business scenario.

[0101] In the disclosed embodiment, after obtaining the vocabulary, the quality of the constructed vocabulary can be evaluated to determine the performance of the vocabulary for each business scenario, and the vocabulary with lower performance can be updated until the performance of the vocabulary reaches the reference index, so that each vocabulary can maintain a high performance. By improving the success rate of rewriting the target vocabulary in the query object, the question and answer effect of the question and answer system can be improved.

[0102] It should be noted that the performance of the vocabulary in the business scenario is used to indicate the efficiency and effect of data processing based on the vocabulary in the process of executing the question-answering task. In some embodiments, the performance of the vocabulary may include at least one of the following: accuracy; recall rate; speed of the vocabulary in processing data; coverage of the vocabulary; adaptability of the vocabulary; scalability of the vocabulary; robustness of the vocabulary; maintenance cost of the vocabulary; and compatibility of the vocabulary.

[0103] In the disclosed embodiment, after obtaining the associated words of the target word, the target word in the first query object can be adjusted based on the associated words to obtain a second query object. Here, adjusting the target word in the first query object based on the associated words can be understood as rewriting the target word based on the associated words. Exemplarily, the target word in the first query object can be replaced based on the associated words to obtain the second query object; or the associated words and the target words can be fused to obtain a new word, and the target word in the first query object can be replaced based on the new word to obtain the second query object.

[0104] It should be noted that the step of adjusting the target vocabulary of the first query object based on the associated vocabulary can be set after the intent distribution based on the first query object and between the retrieval of the target corpus based on the first query object, that is, the adjustment step can be set after the question-answering system distributes the intent and between the retrieval of the target corpus. The target vocabulary in the first query object input by the user is rewritten to avoid the situation where the original text of the document material in the first query object cannot correspond to that in the target corpus, resulting in the inability to find relevant document fragments (query results). In some embodiments, the rewriting basis of the first query object can be a vocabulary of the business scenario.

[0105] Exemplarily, the target word in the first query object can be replaced based on the associated words to obtain the second query object; or the associated words and the target words can be fused to obtain new words, and the target words in the first query object can be replaced based on the new words to obtain the second query object. Exemplarily, the first query object is: What is the ranking of EPFL? The target word is: "EPFL", and the associated word is "EPFL". If the target word in the first query object is replaced based on the associated words, the second query object can be: What is the ranking of EPFL?

[0106] After obtaining the second query object, a target query result can be obtained based on the second query object. Exemplarily, the target corpus can be searched based on the second query object to obtain a search result (recall fragment), and then the search result is sent to a generation module of the question-answering system, the generation module understands the search result, obtains the target query result, and outputs the target query result.

[0107] In some embodiments, a target query result is obtained based on the second query object, including: determining a query result to be processed based on the retrieval result of the first query object in the target corpus and the association type between the associated vocabulary and the target vocabulary; and generating a target query result based on the query result to be processed.

[0108] It should be noted that the search result of the target corpus based on the first query object may include: the first query result is retrieved from the target corpus based on the first query object; or the first query result is not retrieved from the target corpus based on the first query object. It can be understood as the first query result is retrieved from the target corpus based on the use of the first query object; or the first query result is not retrieved from the target corpus based on the use of the first query object.

[0109] The association type between the associated words and the target words may include: the association type between the associated words and the target words is a synonym, the association type between the associated words and the target words is a near-synonym, the association type between the associated words and the target words is a peer relationship, the association type between the associated words and the target words is a hierarchical relationship, etc.

[0110] In the case where the first query result has different retrieval results and / or association types for the target corpus, the corresponding determined target query result will also be different. Exemplarily, in the case where the first query result is retrieved from the target corpus based on the first query object, the target query result can be generated based on the first query result and / or the second query result. In the case where the first query result is not retrieved from the target corpus based on the first query object, the target query result can be generated directly based on the second query result.

[0111] In the disclosed embodiment, based on the search results of the first query object on the target corpus and the association type between the associated vocabulary and the target vocabulary, the query results to be processed are determined, and the target query results are generated based on the query results to be processed; based on the search results of the first query object on the target corpus and the associated vocabulary, the query results to be processed are jointly determined, and the target query results are generated based on the query results to be processed. Compared with obtaining the target query results directly based on the first query results or based on the second query results, the accuracy of the obtained target query results can be improved.

[0112] Taking the education consulting industry as an example, if the first query object is: "What is the QS world ranking of EFPL?". Since there is no abbreviation of EFPL in the target corpus, the first query result S cannot be retrieved (the recall segment is empty). When the first query object is rewritten as the second query object "What is the QS world ranking of EPFL?" based on the associated vocabulary, the big model can understand that EFPL = EPFL, and can accurately output the target query result: "According to the content in the document, the QS world ranking of EFPL (EPFL) is XX".

[0113] In some embodiments, based on the retrieval results of the first query object in the target corpus and the association type between the associated vocabulary and the target vocabulary, the query results to be processed are determined, including: when the first query result is not retrieved based on the first query object and the association type is the first type, searching the target corpus based on the second query object to obtain a second query result; and determining the second query result as the query result to be processed.

[0114] It should be noted that failure to retrieve the first query result from the target corpus based on the first query object can be understood as: the matching degree of all query results (recalled fragments) retrieved from the target corpus based on the first query object and the first query object is lower than the preset matching degree threshold (the recalled results do not meet the preset conditions), or failure to retrieve any query result from the target corpus based on the first query object (recall is empty).

[0115] The first type of association relationship may be a relationship indicating that the target word and the associated word are synonyms, a synonym relationship, or a cascade relationship. Exemplarily, when the first query result is not retrieved based on the first query object, and the target word and the associated word are synonyms, the target corpus may be retrieved based on the second query object to obtain a second query result, and the second query result may be determined as the query result to be processed.

[0116] Exemplarily, when the first query result S is not retrieved from the target corpus based on the first query object (for example, the recalled fragment is empty), the target vocabulary in the first query object can be replaced with a synonym to obtain a second query object. If a second query result S1 is found based on the second query object, the second query result can be determined as a query result to be processed, and the second query result can be input into a large model, i.e., a large language model (LLM), to obtain a target query result.

[0117] In the embodiment of the present disclosure, if the first query object does not retrieve the first query result and the association type is the first type, the target corpus can be retrieved based on the second query object to obtain the second query result, and the second query result can be determined as the query result to be processed, and then the target query result can be obtained based on the query result to be processed. In the case that the recall segment of the first query object is empty, the second query result can be obtained based on the second query object obtained after rewriting the first query object, and the target query result can be obtained based on the second query result, which can improve the accuracy of the target query result.

[0118] In some embodiments, a target corpus is retrieved based on a second query object to obtain a second query result, including: when there are at least two associated words, for each associated word, based on the type of association between the associated word and the target word, determining the priority of the associated word; based on the priority of the associated words, adjusting the target words in the first query object based on each associated word in a traversal manner to obtain a second query object, and searching the target corpus based on the second query object until the second query result is retrieved.

[0119] It should be noted that the first type of association relationship may be a relationship indicating that the target word and the associated word are near-synonymous, synonymous, or cascaded.

[0120] In an embodiment of the present disclosure, the priority of associated words can be determined based on the association type. In some embodiments, the semantic relevance between the target word and the associated word can be indicated based on the association type, wherein the semantic relevance between the target word and the associated word is positively correlated with the priority of the associated word. Exemplarily, when the association type indicates that the semantic relevance between the target word and the associated word is greater than a preset relevance, the priority of the associated word is higher; when the association type indicates that the semantic relevance between the target word and the associated word is less than or equal to the preset relevance, the priority of the associated word is lower.

[0121] For example, the priority of an associated word that is a synonym of the target word is higher than the priority of an associated word that is a near synonym of the target word; the priority of an associated word that is a near synonym of the target word is higher than the priority of an associated word that is a superior-subordinate relationship of the target word.

[0122] In the disclosed embodiment, when multiple associated words are determined, the target words in the first query object may be rewritten in a traversal manner based on the priority of each associated word until the second query result is retrieved from the target corpus.

[0123] In some embodiments, searching the target corpus based on the second query object until a second query result is retrieved includes: if the second query result is not retrieved by the second query object obtained based on the associated vocabulary of the first priority, searching the target corpus by the second query object obtained based on the associated vocabulary of the second priority until all the associated vocabulary of the first type are traversed and / or the second query result is retrieved.

[0124] It should be noted that the first priority is higher than the second priority.

[0125] Exemplarily, the first query object can be adjusted based on associated words that are synonymous with the target word to obtain a second query object. If the second query result S1 is not retrieved from the target corpus based on the second query object, the first query object is further adjusted based on associated words that are superior to the target word to obtain a second query object. If a second query result St is retrieved from the target corpus based on the second query object, the second query result St is determined as the query result to be processed, and the traversal is terminated.

[0126] In other embodiments, when the second query result is still not determined after traversing the associated words of the first type, a prompt message may be output to indicate that the recall has failed.

[0127] In the disclosed embodiment, for each associated word, the priority of the associated word is determined based on the type of association between the associated word and the target word; based on the priority of the associated word, the target word in the first query object is adjusted based on each associated word to obtain a second query object, and the target corpus is searched based on the adjusted second query object until the second query result is retrieved. The accuracy of determining the target query result can be improved by rewriting the first query object in a traversal manner and searching the target corpus based on the rewritten second query object.

[0128] In some embodiments, based on the retrieval results of the first query object in the target corpus and the association type between the associated vocabulary and the target vocabulary, the query results to be processed are determined, including: when the first query result is retrieved based on the first query object and the association type is the second type, the first query result is adjusted based on the second query result to obtain the query result to be processed.

[0129] It should be noted that the first type of association relationship may be a relationship indicating that the target word and the associated word are parallel words.

[0130] When the first query result is retrieved based on the first query result and the association type is the second type, the first query result may be adjusted based on the second query result to obtain the query result to be processed.

[0131] For example, after obtaining the second query result S2, the duplicate content of the first query result S and the second query result S2 can be determined, and the duplicate content can be deleted from the first query result S to obtain the query result to be processed (S-S2). By removing the fragments in the first query result S that may interfere with the large model, the target query result can be made more accurate.

[0132] In some other embodiments, the step of retrieving the target corpus based on the first query object may be performed in parallel with the step of retrieving the target corpus based on the second query object. In some other embodiments, the target corpus may be retrieved based on the first query object first, and then based on the second query object.

[0133] In other embodiments, when the second query result S2 is not retrieved, the first query result S can be determined as a query result to be processed, and the query result to be processed can be input into the big model to obtain the target query result.

[0134] In other embodiments, when the question-answering system receives a query object that does not include the target vocabulary, it can directly search the target corpus based on the query statement, input the retrieved query results into the big model, and output the query results based on the big model.

[0135] In some embodiments, the method further comprises:

[0136] Upon receiving a first query object including a target word, obtaining description information of the target word;

[0137] Based on the second query object, a target query result is obtained, including:

[0138] Based on the second query object and the description information, a target query result is obtained.

[0139] In the embodiment of the present disclosure, when the first query object is received, description information of the target vocabulary can be obtained. Exemplarily, semantic analysis is performed on the first query object to obtain a semantic analysis result, and the current business scenario is determined based on the semantic analysis result. After the current business scenario is determined, the vocabulary of the current business scenario can be determined.

[0140] After determining the vocabulary of the current business scenario, the target vocabulary can be compared with each vocabulary pre-stored in the vocabulary of the current business scenario to determine a preset vocabulary matching the target vocabulary, and the description information of the preset vocabulary can be determined as the description information of the target vocabulary. The preset vocabulary matching the target vocabulary can include: a vocabulary whose similarity with the target vocabulary is less than a preset similarity threshold, for example, a preset vocabulary that is the same as the target vocabulary.

[0141] After obtaining the description information of the target vocabulary, the query result to be processed and the description information obtained based on the second query object in any of the above embodiments can be input into the large model, and the large model obtains the target query result based on the second query result and the description information.

[0142] In the disclosed embodiment, after obtaining the query result to be processed based on any of the above embodiments, when inputting the query result to the big model to generate the answer, the query result to be processed and the description information of the target vocabulary in the vocabulary (the vocabulary explanation included in the vocabulary) can be input into the big model together to help the big model better understand the content in the query result to be processed and generate a more accurate target query result. Compared with the query result after only inputting the query statement and then processing it in the related art, the accuracy of the target query result can be improved.

[0143] In some embodiments, the method further comprises:

[0144] Obtain the total number of query objects received within the target time period, the number of triggers for adjusting the query objects, and the parameters for processing the query objects;

[0145] A second evaluation value is determined based on at least one of the total number of times, the number of triggers, and the parameter.

[0146] Exemplarily, the total number of query objects received by the question-answering system within a target time period, the number of triggers for adjusting the query objects, and the parameters for processing the query objects by the question-answering system are obtained; based on at least one of the total number, the number of triggers, and the parameters, a second evaluation value of the question-answering system is determined; wherein the second evaluation value is used to indicate the performance of the question-answering system. The query object may include: a query statement.

[0147] In the disclosed embodiment, after the question-answering system obtained based on any of the above embodiments is launched, that is, after the question-answering system is put into use, the total number of query objects received by the question-answering system within the target time period, the number of triggers for adjusting the query objects, and the parameters for processing the query objects by the question-answering system can be obtained. The parameters for processing the query objects may include: the duration and speed of the question-answering system processing the query objects.

[0148] Determining a second evaluation value of the question-answering system based on the total number of query objects received by the question-answering system within the target time period and the number of triggers for adjusting the query objects includes: determining a ratio of the total number of query objects received by the question-answering system within the target time period to the number of triggers for adjusting the query objects, and determining the second evaluation value based on the ratio. Assuming that the number of triggers is P and the total number is Q, the calculation formula for the second evaluation value can be as follows:

[0149]

[0150] In formula (2), A2 represents the second evaluation value; P represents the number of triggers for adjusting the query object within the target time period; and Q represents the total number of query objects received within the target time period.

[0151] In other embodiments, the optimization effect of the question-answering system may also be monitored through relevant parameters such as the retrieval recall rate, answer accuracy rate, etc. of the question-answering system.

[0152] In the embodiments of the present disclosure, the problem of inaccurate and irrelevant answers due to the inability of the large model to understand some professional knowledge and the company's own knowledge in the related technology is solved, and the recall rate of the entire question-answering system and the overall answer accuracy are improved. Compared with the solution of fine-tuning the large model, the present disclosure not only provides a method for optimizing the question-answering system with lower technical cost and easier operation, but also provides a solution for building a vocabulary library for business scenarios, helping enterprises build various business scenarios, and helping users better manage knowledge through the question-answering system.

[0153] Figure 4 A data processing method according to an exemplary embodiment is shown in FIG. Figure 2 .like Figure 4 As shown, when the question-answering system receives a first query object 401 including a target vocabulary, related vocabulary of the target vocabulary can be determined from the vocabulary library 402, and the target vocabulary in the first query object 401 can be rewritten based on the related vocabulary to obtain a second query object 403, and the target corpus can be retrieved based on the second query term 403 to obtain a second query result, and then a query result to be processed 404 can be determined based on the result of retrieving the target corpus based on the first query object 401 and the second query result, and the query result to be processed 404 can be input into the large model to obtain a target query result 405.

[0154] Figure 5 A data processing method according to an exemplary embodiment is shown in FIG. Figure 3 .like Figure 5 As shown, the method mainly includes the following steps:

[0155] In step 501, it is determined whether a first query result S is retrieved.

[0156] Here, if the first query result S is retrieved, step 502 is continued to be executed, and if the first query result S is not retrieved, step 504 is executed.

[0157] In step 502, the first query result S is adjusted based on the second query result S2.

[0158] In step 503, the duplicate content of the second query result S2 is deleted from the first query result S to obtain a query result to be processed S', where S'=(S-S2).

[0159] In step 504, the first query object is rewritten based on the associated words of the first priority to obtain a second query object.

[0160] In step 504, it is determined whether the second query result S1 is retrieved.

[0161] Here, if the second query result S1 is retrieved, step 506 is continued to be executed, and if the second query result S1 is not retrieved, step 507 is executed.

[0162] In step 506, the second query result S1 is determined as the query result to be processed S', where S'=-S1.

[0163] In step 507, the first query object is rewritten based on the associated words of the second priority to obtain a second query object.

[0164] In step 508, whether the second query result St is retrieved.

[0165] Here, if the second query result St is retrieved, step 509 is continued to be executed, and if the second query result St is not retrieved, it is determined that there is no recall, and a prompt message is output. For example, a prompt message of "target word not found" may be output.

[0166] In step 509, the second query result St is determined as the query result to be processed S', where S'=St.

[0167] In step 510, the large model obtains the target query result.

[0168] Here, in the case where the target vocabulary has description information, the large model can obtain the target query result based on the query result to be processed and the description information of the target vocabulary.

[0169] In the embodiments of the present disclosure, firstly, a vocabulary can be constructed for different business scenarios. By matching the vocabulary with the business scenario, after determining the business scenario, the vocabulary of the business scenario can be directly searched based on the target vocabulary. Compared with storing all data in a database, the efficiency of processing data can be improved, and the accuracy of the determined related vocabulary can be improved. By strongly correlating the content in the vocabulary with the business scenario (question and answer scenario), the question and answer effect in the business scenario can be optimized.

[0170] In the second aspect, the link of adjusting the target vocabulary of the first query object based on the associated vocabulary can be set after the question-answering system distributes the intent based on the first query object and between retrieving the target corpus based on the first query object. The target vocabulary in the first query object input by the user is rewritten to avoid the situation where the original text of the document material in the first query object cannot correspond to the original text in the target corpus, resulting in the inability to find the relevant document fragment (query result). In some embodiments, the rewriting basis of the first query object can be a vocabulary of the business scenario.

[0171] Thirdly, it not only strengthens the retrieval effect in the pre-recall stage, but also transmits the pending query results and scanning information to the big model in the big model output stage, so that the big model can understand the correspondence between the expression of the target vocabulary in the first query object and the pending query results (recalled document fragments).

[0172] Figure 6 FIG. 1 is a block diagram of a data processing device according to an exemplary embodiment. Figure 6 As shown, the device 600 mainly includes:

[0173] A first acquisition module 601 is configured to acquire associated words having an associated relationship with the target word when receiving a first query object including the target word;

[0174] An adjustment module 602 is configured to adjust the target word in the first query object based on the associated word to obtain a second query object;

[0175] The generating module 603 is configured to obtain a target query result based on the second query object.

[0176] In some embodiments, the generating module 603 is configured to:

[0177] Determining a query result to be processed based on the search result of the first query object in the target corpus and the association type between the associated vocabulary and the target vocabulary;

[0178] Based on the query result to be processed, the target query result is generated.

[0179] In some embodiments, the generating module 603 is configured to:

[0180] If no first query result is retrieved based on the first query object and the association type is the first type, searching the target corpus based on the second query object to obtain a second query result;

[0181] The second query result is determined as the query result to be processed.

[0182] In some embodiments, the generating module 603 is configured to:

[0183] In the case where there are at least two associated words, for each of the associated words, determining a priority of the associated words based on the type of association between the associated words and the target word;

[0184] Based on the priorities of the associated words, the target words in the first query object are adjusted based on each of the associated words in a traversal manner to obtain the second query object, and the second query object searches the target corpus until the second query result is retrieved.

[0185] In some embodiments, the generating module 603 is configured to:

[0186] If the second query result is not retrieved by the second query object obtained based on the associated words of the first priority, the target corpus is searched by the second query object obtained based on the associated words of the second priority until all the associated words of the first type are traversed and / or the second query result is retrieved.

[0187] In some embodiments, the generating module 603 is configured to:

[0188] When the first query result is retrieved based on the first query object and the association type is the second type, the first query result is adjusted based on the second query result to obtain the query result to be processed.

[0189] In some embodiments, the first acquisition module 601 is configured to:

[0190] In case of receiving the first query object, determining a current business scenario based on the first query object, and determining a vocabulary of the current business scenario;

[0191] The words in the vocabulary of the current business scenario that have the association relationship with the target word are determined as the associated words of the target word in the current business scenario.

[0192] In some embodiments, the apparatus 600 further includes:

[0193] A second acquisition module is configured to acquire initial data from at least one data source in different business scenarios respectively;

[0194] an analysis module, configured to analyze text content in the initial data to obtain association information between each word in the initial data and the target word when the initial data includes the target word;

[0195] The construction module is configured to determine the associated vocabulary from the initial data based on the associated information, and construct the vocabulary of the business scenario based on the association relationship between the associated vocabulary and the target vocabulary and / or the description information of the target vocabulary.

[0196] In some embodiments, the apparatus 600 further includes:

[0197] A first determination module is configured to determine, in different business scenarios, based on the target vocabulary in a preset object, an associated vocabulary of the target vocabulary from a vocabulary library of the business scenario;

[0198] A second determination module is configured to determine a first evaluation value of the vocabulary of the business scenario based on the number of preset objects of the associated vocabulary and the total number of the preset objects; wherein the first evaluation value is used to indicate the performance of the vocabulary of the business scenario;

[0199] The updating module is configured to update the vocabulary of the business scenario when the first evaluation value is less than a preset threshold value until the first evaluation value is greater than or equal to the preset threshold value.

[0200] In some embodiments, the apparatus 600 further includes:

[0201] A third acquisition module is configured to acquire description information of the target vocabulary when receiving the first query object including the target vocabulary;

[0202] The generating module 603 is configured to:

[0203] The target query result is obtained based on the second query object and the description information.

[0204] In some embodiments, the apparatus further comprises:

[0205] A fourth acquisition module is configured to acquire a total number of query objects received within a target time period, a triggering number of adjustments to the query objects, and parameters for processing the query objects;

[0206] The third determination module is configured to determine a second evaluation value based on the total number of times, the triggering number of times and at least one of the parameters.

[0207] Regarding the device in the above embodiment, the specific manner in which each module performs operations has been described in detail in the embodiment of the method, and will not be elaborated here.

[0208] Based on the same inventive concept, an embodiment of the present disclosure further provides an electronic device, which may be a computer or a terminal in one or more of the above embodiments. Figure 7 FIG. 1 is a schematic diagram showing the structure of an electronic device according to an exemplary embodiment. Figure 7As shown, the electronic device 700 adopts general computer hardware and includes a processor 701 , a memory 702 , a bus 703 , an input device 704 , and an output device 705 .

[0209] In some possible implementations, the memory 702 may include computer storage media in the form of volatile and / or non-volatile memory, such as read-only memory and / or random access memory. The memory 702 may store an operating system, application programs, other program modules, executable code, program data, user data, etc.

[0210] Input device 704 can be used to input commands and information to the electronic device, such as a keyboard or a pointing device, such as a mouse, trackball, touch pad, microphone, joystick, game pad, satellite TV antenna, scanner or similar device. Input device 704 can be connected to processor 701 via bus 703.

[0211] The output device 705 can be used to output information to the electronic device 700 . In addition to the monitor, the output device 705 can also be other peripheral output devices, such as a speaker and / or a printing device. The output device 705 can also be connected to the processor 701 via the bus 703 .

[0212] The electronic device 700 may be connected to a network, such as a local area network (LAN), via the antenna 706. In a networked environment, executable instructions may be stored in a remote storage device, rather than being limited to local storage.

[0213] When the processor 701 in the electronic device 700 executes the executable code or application stored in the memory 702, the electronic device 700 can implement the image processing method in the above embodiments. The specific execution process refers to the above embodiments and will not be repeated here.

[0214] The memory 702 may store information for implementing Figure 6 The executable instructions of the functions of the first acquisition module 601, the adjustment module 602 and the generation module 603. Figure 6 The functions / implementation processes of the first acquisition module 601, the adjustment module 602 and the generation module 603 can be realized by Figure 7 The processor 701 in the embodiment calls the executable instructions stored in the memory 702 to implement the above-mentioned implementation process and functions.

[0215] Based on the same inventive concept, the embodiment of the present disclosure further provides a storage medium. The storage medium stores instructions. When the instructions are executed on a computer, the instructions are used to execute the data processing method in one or more of the above embodiments.

[0216] Based on the same inventive concept, the embodiments of the present disclosure further provide a computer program or a computer program product. When the computer program product is executed on a computer, the computer implements the data processing method in one or more of the above embodiments.

[0217] Those skilled in the art will readily appreciate other embodiments of the present disclosure after considering the specification and practicing the invention disclosed herein. The present disclosure is intended to cover any variations, uses or adaptations of the present disclosure that follow the general principles of the present disclosure and include common knowledge or customary techniques in the art that are not disclosed in the present disclosure. The description and examples are to be considered exemplary only, and the true scope and spirit of the present disclosure are indicated by the claims.

[0218] It should be understood that the present disclosure is not limited to the exact structures that have been described above and shown in the drawings, and that various modifications and changes may be made without departing from the scope thereof. The scope of the present disclosure is limited only by the appended claims.

Claims

1. A data processing method, characterized in that: include: When receiving a first query object including a target word, obtaining associated words having an associated relationship with the target word; Adjusting the target vocabulary in the first query object based on the associated vocabulary to obtain a second query object; Based on the second query object, a target query result is obtained.

2. The method according to claim 1, characterized in that The obtaining a target query result based on the second query object includes: Determining a query result to be processed based on the search result of the first query object in the target corpus and the association type between the associated vocabulary and the target vocabulary; Based on the query result to be processed, the target query result is generated.

3. The method according to claim 2, characterized in that The determining of the query result to be processed based on the search result of the first query object in the target corpus and the association type between the associated vocabulary and the target vocabulary includes: If no first query result is retrieved based on the first query object and the association type is the first type, searching the target corpus based on the second query object to obtain a second query result; The second query result is determined as the query result to be processed.

4. The method according to claim 3, characterized in that The step of searching the target corpus based on the second query object to obtain a second query result includes: In the case where there are at least two associated words, for each of the associated words, determining a priority of the associated words based on the type of association between the associated words and the target word; Based on the priorities of the associated words, the target words in the first query object are adjusted based on each of the associated words in a traversal manner to obtain the second query object, and the target corpus is searched based on the second query object until the second query result is retrieved.

5. The method according to claim 4, characterized in that The searching the target corpus based on the second query object until the second query result is retrieved includes: In the event that the second query result is not retrieved by the second query object obtained based on the associated vocabulary of the first priority, the target corpus is searched by the second query object obtained based on the associated vocabulary of the second priority until all associated vocabulary of the first type are traversed and / or the second query result is retrieved.

6. The method according to claim 3, characterized in that The determining of the query result to be processed based on the search result of the first query object in the target corpus and the association type between the associated vocabulary and the target vocabulary includes: When the first query result is retrieved based on the first query object and the association type is the second type, the first query result is adjusted based on the second query result to obtain the query result to be processed.

7. The method according to claim 1, characterized in that The step of, upon receiving a first query object including a target word, obtaining associated words associated with the target word includes: In case of receiving the first query object, determining a current business scenario based on the first query object, and determining a vocabulary of the current business scenario; The words in the vocabulary of the current business scenario that have the association relationship with the target word are determined as the associated words of the target word in the current business scenario.

8. The method according to claim 7, characterized in that The method further comprises: In the case of different business scenarios, initial data is obtained from at least one data source in the business scenario respectively; In the case where the initial data includes the target vocabulary, analyzing the text content in the initial data to obtain association information between each vocabulary in the initial data and the target vocabulary; Based on the association information, the associated vocabulary is determined from the initial data, and based on the association relationship between the associated vocabulary and the target vocabulary and / or the description information of the target vocabulary, a vocabulary library of the business scenario is constructed.

9. The method according to claim 8, characterized in that The method further comprises: In the case of different business scenarios, based on the target vocabulary in the preset object, determining the associated vocabulary of the target vocabulary from the vocabulary library of the business scenario; Determining a first evaluation value of the vocabulary of the business scenario based on the number of preset objects of the associated vocabulary and the total number of the preset objects; wherein the first evaluation value is used to indicate the performance of the vocabulary of the business scenario; When the first evaluation value is less than a preset threshold, the vocabulary of the business scenario is updated until the first evaluation value is greater than or equal to the preset threshold.

10. The method according to any one of claims 1 to 9, characterized in that: The method further comprises: Upon receiving the first query object including the target vocabulary, obtaining description information of the target vocabulary; The obtaining a target query result based on the second query object includes: The target query result is obtained based on the second query object and the description information.

11. The method according to any one of claims 1 to 9, characterized in that: The method further comprises: Obtaining the total number of query objects received within a target time period, the number of triggers for adjusting the query objects, and parameters for processing the query objects; A second evaluation value is determined based on at least one of the total number of times, the triggering number of times, and the parameter.

12. A data processing device, characterized in that: include: A first acquisition module is configured to acquire associated words having an associated relationship with the target word when receiving a first query object including the target word; an adjustment module, configured to adjust the target vocabulary in the first query object based on the associated vocabulary to obtain a second query object; The generating module is configured to obtain a target query result based on the second query object.

13. A computer program product comprising a computer program or instructions, characterized in that When the computer program or instruction is executed by a processor, the steps of the method according to any one of claims 1 to 11 are implemented.

Images