A method, apparatus, and medium for intelligent agent knowledge retrieval based on multi-knowledge hybrid recall.

Abstract

A knowledge retrieval method, apparatus, and medium based on multi-knowledge hybrid recall for intelligent agents are disclosed, relating to the fields of large models, intelligent agents, and artificial intelligence. The method includes: matching question-answer pairs in a question-answer database based on the knowledge retrieval question; if no matching question-answer pairs are found, matching terms in a terminology database based on the knowledge retrieval question; and performing knowledge retrieval in the question-answer database and / or a database based on the terminology matching results and the knowledge retrieval question to obtain the knowledge retrieval results. By performing question-answer pair matching in the question-answer database, corresponding answers can be quickly output when a matching question-answer pair is found, thereby improving the stability of the response and reducing the long waiting time caused by calling large models. When no matching question-answer pairs are found in the question-answer database, terminology matching in the terminology database can supplement the explanations of terms in the knowledge retrieval question beforehand, improving the accuracy and reliability of the knowledge retrieval results.

Description

Technical Field

[0001] This disclosure relates to the fields of large models, intelligent agents, and artificial intelligence, and specifically to an intelligent agent knowledge retrieval method, apparatus, and medium based on multi-knowledge hybrid recall. Background Technology

[0002] With the continuous development of artificial intelligence technology, intelligent agents are being used more and more widely in daily life. For example, in knowledge retrieval scenarios, in order to quickly obtain the information needed by users from massive amounts of data, the powerful semantic understanding and generation capabilities of the large models associated with the intelligent agent can be used to quickly sort and filter the massive amounts of data, thereby providing users with targeted answers or related knowledge expansion.

[0003] However, when using related technologies to retrieve knowledge based on large models associated with intelligent agents, there are problems such as unstable and inaccurate knowledge retrieval results. Summary of the Invention

[0004] This summary section is provided to briefly introduce the concepts, which will be described in detail in the detailed description section below. This summary section is not intended to identify key or essential features of the claimed technical solution, nor is it intended to limit the scope of the claimed technical solution.

[0005] Firstly, this disclosure provides an intelligent agent knowledge retrieval method based on multi-knowledge hybrid recall, including:

[0006] Determine the knowledge retrieval link of the intelligent agent, wherein the knowledge retrieval link is used to characterize the knowledge retrieval logic of the intelligent agent in a preset retrieval resource base, wherein the preset retrieval resource base includes at least one of a knowledge base, a terminology base, a question and answer base, and a database.

[0007] When the knowledge retrieval link represents the knowledge retrieval logic of the agent in the knowledge base, the terminology base, the question-and-answer base, and the database, in response to receiving a knowledge retrieval question, the agent performs question-and-answer pair matching in the question-and-answer base based on the knowledge retrieval question;

[0008] If a question-answer pair is matched, the answer in the matched question-answer pair will be output as the knowledge retrieval result of the knowledge retrieval question.

[0009] If no question-answer pair is found, term matching is performed in the terminology database based on the knowledge retrieval question, and knowledge retrieval is performed in the question-answer database and / or the database based on the term matching result and the knowledge retrieval question to obtain knowledge retrieval results for the knowledge retrieval question.

[0010] Secondly, this disclosure provides an intelligent agent knowledge retrieval device based on multi-knowledge hybrid recall, comprising:

[0011] A determination module is used to determine the knowledge retrieval link of the intelligent agent. The knowledge retrieval link is used to characterize the knowledge retrieval logic of the intelligent agent in a preset retrieval resource base. The preset retrieval resource base includes at least one of a knowledge base, a terminology base, a question and answer base, and a database.

[0012] The matching module is configured to, in response to receiving a knowledge retrieval question, perform question-answer pair matching in the question-answer database based on the knowledge retrieval question, when the knowledge retrieval link represents the knowledge retrieval logic of the agent in the knowledge base, the terminology base, the question-answer base, and the database;

[0013] The first processing module is used to output the answer in the matched question-answer pair as the knowledge retrieval result of the knowledge retrieval question when a question-answer pair is matched.

[0014] The second processing module is used to perform term matching in the terminology database based on the knowledge retrieval question when no question-answer pair is matched, and to perform knowledge retrieval in the question-answer database and / or the database according to the term matching result and the knowledge retrieval question, so as to obtain the knowledge retrieval result for the knowledge retrieval question.

[0015] Thirdly, this disclosure provides a computer-readable medium having a computer program stored thereon, which, when executed by a processing device, implements the steps of the method described in the first aspect.

[0016] Fourthly, this disclosure provides an electronic device, comprising:

[0017] A storage device on which computer programs are stored;

[0018] A processing device for executing the computer program in the storage device to implement the steps of the method in the first aspect.

[0019] Fifthly, this disclosure provides a computer program product, including a computer program that, when executed by a processor, implements the steps of the method described in the first aspect.

[0020] Through the above technical solutions, the knowledge retrieval process can be structured so that the agent can perform question-and-answer pair matching in the knowledge base, terminology base, question-and-answer base, and database. First, based on the knowledge retrieval question, a question-and-answer pair can be matched in the question-and-answer base. If a matching question-and-answer pair is found, the corresponding answer can be quickly output. This not only improves the stability of the response but also reduces the long waiting time caused by calling large models, thus enhancing the user experience. If no matching question-and-answer pair is found in the question-and-answer base, terminology matching can be performed in the terminology base based on the knowledge retrieval question. This allows for preliminary explanations of terms in the knowledge retrieval question, enabling the agent to more accurately understand the knowledge retrieval question and improve the accuracy and reliability of knowledge retrieval results when performing knowledge retrieval in the question-and-answer base and / or database. In addition, since the preset retrieval resource base includes at least one of the following: knowledge base, terminology base, question and answer base, and database, the knowledge retrieval link of the intelligent agent can be flexibly configured according to business needs when determining the knowledge retrieval link. Compared with the high-code approach to building the knowledge retrieval link in related technologies, it can reduce the development burden caused by changes in business needs, achieve better compatibility and reusability, significantly reduce subsequent development costs, and improve the adaptability to new application scenarios.

[0021] Other features and advantages of this disclosure will be described in detail in the following detailed description section. Attached Figure Description

[0022] The above and other features, advantages, and aspects of the embodiments of this disclosure will become more apparent from the accompanying drawings and the following detailed description. Throughout the drawings, the same or similar reference numerals denote the same or similar elements. It should be understood that the drawings are schematic, and the originals and elements are not necessarily drawn to scale. In the drawings:

[0023] Figure 1 This is a flowchart illustrating an intelligent agent knowledge retrieval method based on multi-knowledge hybrid recall according to an exemplary embodiment of this disclosure;

[0024] Figure 2 This is a flowchart illustrating an intelligent agent knowledge retrieval method based on multi-knowledge hybrid recall according to an exemplary embodiment of the present disclosure;

[0025] Figure 3 This is a flowchart illustrating another intelligent agent knowledge retrieval method based on multi-knowledge hybrid recall according to an exemplary embodiment of the present disclosure;

[0026] Figure 4 This is a flowchart illustrating yet another intelligent agent knowledge retrieval method based on multi-knowledge hybrid recall according to an exemplary embodiment of the present disclosure;

[0027] Figure 5 This is a schematic diagram illustrating the configuration of a knowledge retrieval link according to an exemplary embodiment of this disclosure;

[0028] Figure 6 This is a schematic diagram illustrating the configuration of a retrieval resource repository node according to an exemplary embodiment of the present disclosure;

[0029] Figure 7 This is a schematic diagram illustrating the configuration of a search resource library corresponding to an initial search resource library node according to an exemplary embodiment of the present disclosure;

[0030] Figure 8 This is a schematic diagram illustrating another knowledge retrieval link configuration according to an exemplary embodiment of the present disclosure;

[0031] Figure 9 This is a structural block diagram of an intelligent agent knowledge retrieval device based on multi-knowledge hybrid recall, according to an exemplary embodiment of the present disclosure.

[0032] Figure 10 This is a schematic diagram of the structure of an electronic device according to an exemplary embodiment of the present disclosure. Detailed Implementation

[0033] Embodiments of this disclosure will now be described in more detail with reference to the accompanying drawings. While some embodiments of this disclosure are shown in the drawings, it should be understood that this disclosure can be implemented in various forms and should not be construed as limited to the embodiments set forth herein. Rather, these embodiments are provided to provide a more thorough and complete understanding of this disclosure. It should be understood that the accompanying drawings and embodiments of this disclosure are for illustrative purposes only and are not intended to limit the scope of protection of this disclosure.

[0034] It should be understood that the steps described in the method embodiments of this disclosure may be performed in different orders and / or in parallel. Furthermore, the method embodiments may include additional steps and / or omit the steps shown. The scope of this disclosure is not limited in this respect.

[0035] The term "comprising" and its variations as used herein are open-ended inclusions, meaning "including but not limited to". The term "based on" means "at least partially based on". The term "one embodiment" means "at least one embodiment"; the term "another embodiment" means "at least one additional embodiment"; the term "some embodiments" means "at least some embodiments". Definitions of other terms will be given in the description below.

[0036] It should be noted that the concepts of "first" and "second" mentioned in this disclosure are used only to distinguish different devices, modules or units, and are not used to limit the order of functions performed by these devices, modules or units or their interdependencies.

[0037] It should be noted that the terms "a" and "a plurality of" used in this disclosure are illustrative rather than restrictive, and those skilled in the art should understand that, unless otherwise expressly indicated in the context, they should be understood as "one or more".

[0038] The names of messages or information exchanged between multiple devices in the embodiments of this disclosure are for illustrative purposes only and are not intended to limit the scope of such messages or information.

[0039] It is understood that before using the technical solutions disclosed in the various embodiments of this disclosure, users should be informed of the types, scope of use, and usage scenarios of the personal information involved in this disclosure in an appropriate manner in accordance with relevant laws and regulations, and user authorization should be obtained.

[0040] For example, upon receiving a user's active request, a prompt message is sent to the user to explicitly inform them that the requested operation will require the acquisition and use of the user's personal information. This allows the user to independently choose whether to provide personal information to the software or hardware, such as the electronic device, application, server, or storage medium performing the operations of this disclosed technical solution, based on the prompt message.

[0041] As an optional but non-limiting implementation, in response to a user's active request, sending a prompt message to the user can be done via a pop-up window, where the prompt message can be presented in text format. Furthermore, the pop-up window can also include a selection control allowing the user to choose "agree" or "disagree" to provide personal information to the electronic device.

[0042] It is understood that the above notification and user authorization process are merely illustrative and do not constitute a limitation on the implementation of this disclosure. Other methods that comply with relevant laws and regulations may also be applied to the implementation of this disclosure.

[0043] Meanwhile, it is understood that the data involved in this technical solution (including but not limited to the data itself, the acquisition or use of the data) shall comply with the requirements of relevant laws, regulations and related provisions.

[0044] As mentioned in the background section, when performing knowledge retrieval based on large models associated with intelligent agents, related technologies suffer from problems such as unstable and inaccurate knowledge retrieval results.

[0045] For example, in scenarios involving high-frequency enterprise questions, ordinary knowledge retrieval processes directly invoke large models to retrieve knowledge based on user-input questions. However, the output of these large models often suffers from inconsistent results due to uncertainties in contextual understanding and generation logic, failing to meet enterprises' requirements for consistent answers. Furthermore, in query scenarios involving specialized terminology, ordinary knowledge retrieval processes cannot provide prior explanations of these terms, potentially causing large models to misunderstand their meanings and thus reducing the accuracy and reliability of knowledge retrieval.

[0046] Furthermore, when defining the knowledge retrieval path, related technologies typically involve the business side first designing the path based on their own use cases, and then customizing it using high-level code. However, knowledge retrieval paths implemented in this way suffer from poor compatibility and are difficult to promote and reuse due to deep coupling with business logic. Once business requirements change, or if the application needs to be moved to other scenarios, a significant amount of redevelopment work is required, resulting in high development costs and low efficiency.

[0047] In view of this, this disclosure provides a method, apparatus and medium for intelligent agent knowledge retrieval based on multi-knowledge hybrid recall, in order to solve the above-mentioned technical problems.

[0048] The embodiments of this disclosure will be further explained below with reference to the accompanying drawings.

[0049] Figure 1 This is a flowchart illustrating an intelligent agent knowledge retrieval method based on multi-knowledge hybrid recall according to an exemplary embodiment of this disclosure, with reference to... Figure 1 The intelligent agent knowledge retrieval method based on multi-knowledge hybrid recall may include the following steps:

[0050] S101: Determine the knowledge retrieval link of the agent. The knowledge retrieval link is used to characterize the knowledge retrieval logic of the agent in the preset retrieval resource base. The preset retrieval resource base includes at least one of the following: knowledge base, terminology base, question and answer base, and database.

[0051] The knowledge base can be a retrieval resource library that supports the collection, segmentation, and indexing of knowledge content from different data sources, and provides RAG (Retrieval-Augmented Generation) capabilities for intelligent agents. Because the knowledge base can cover knowledge from different data sources and provide RAG support to intelligent agents through efficient retrieval and indexing capabilities, it is suitable for various scenarios that require responses based on local knowledge content. The database can be a retrieval resource library that supports parsing and storing table content according to user-defined rules and supports direct connections to external databases. Because the database can efficiently store, manage, and retrieve structured data, it is suitable for data query scenarios. The question-and-answer library can be a retrieval resource library that supports importing FAQs (Frequently Asked Questions) and indexing questions. FAQs can include high-frequency question pairs and question interception pairs. Because it can quickly output the corresponding answer when a user's question matches content in the question-and-answer library, it is suitable for business scenarios involving high-frequency question processing and / or question interception. A terminology database can be a retrieval resource that supports both structured and unstructured terms. Since a terminology database can include explanations of terms, it can be applied to business knowledge gaps or data analysis scenarios in specific fields to help agents understand knowledge retrieval questions input by users.

[0052] Since knowledge bases, terminology databases, question-and-answer databases, and databases can adapt to different business scenarios, users can freely construct knowledge retrieval links based on the applicable business scenarios of each knowledge base, terminology database, question-and-answer database, and database, as well as their own needs, to obtain knowledge retrieval links that meet their specific requirements.

[0053] S102: In the context of the knowledge retrieval link representing the agent's knowledge retrieval logic in the knowledge base, terminology base, question-answer base, and database, in response to receiving a knowledge retrieval question, question-answer pair matching is performed in the question-answer base based on the knowledge retrieval question.

[0054] For example, after receiving a knowledge retrieval question, the similarity between each question-answer pair in the question-answer database and the knowledge retrieval question can be calculated first. After obtaining the similarity between each question-answer pair and the knowledge retrieval question, question-answer pair matching can be performed in the question-answer database based on the similarity corresponding to each question-answer pair and the preset similarity threshold.

[0055] Given that question-answer pairs typically consist of a question and an answer, calculating the similarity between the question-answer pair and the knowledge retrieval question presents challenges due to the complexity and diversity of the answer portion, which often has a relatively weak relevance to the knowledge retrieval question. This can lead to biases in similarity calculations, affecting the accuracy and effectiveness of matching. Furthermore, processing both the question and answer in a question-answer pair simultaneously increases computational load, reduces the efficiency of similarity calculation, and ultimately impacts knowledge retrieval efficiency and user experience. Therefore, to overcome the influence of the answer portion in similarity calculation, one possible approach is to calculate only the similarity between the question and the knowledge retrieval question within the question-answer pair. This reduces computational load, improves the efficiency of similarity calculation, and overcomes matching biases caused by the complexity of the answer portion, thereby enhancing the accuracy of knowledge retrieval and user experience. In other words, in a possible approach, the question-answer database can include multiple question-answer pairs, each including a question and an answer. Accordingly, in response to receiving a knowledge retrieval question, question-answer pair matching is performed in the question-answer database based on the knowledge retrieval question, which may include:

[0056] In response to receiving a knowledge retrieval question, for each question-answer pair, the similarity between the question and the knowledge retrieval question is determined; based on the similarity and a second similarity threshold, question-answer pair matching is performed in the question-answer database.

[0057] The similarity between the question and the knowledge retrieval question can be obtained by calculating the text edit distance between the two questions, or by calculating the semantic vector similarity between them. Of course, it can also be calculated in other ways, and this disclosure does not impose any restrictions on this.

[0058] The question-answer pair matching in the question-answer database, based on similarity and a second similarity threshold, can be performed as follows: Target question-answer pairs are filtered in the database based on the second similarity threshold and the similarity corresponding to each question-answer pair, wherein the similarity of the target question-answer pair is greater than or equal to the second similarity threshold. If no target question-answer pair is found in the database, it indicates that no knowledge retrieval results were obtained for the knowledge retrieval question after knowledge retrieval in the database. Therefore, further knowledge retrieval can be performed by executing S104. If a target question-answer pair is found in the database, it indicates that knowledge retrieval results for the knowledge retrieval question have been obtained after knowledge retrieval in the database. Therefore, knowledge retrieval results for the knowledge retrieval question can be obtained by executing S103, such as... Figure 2 As shown.

[0059] S103: If a question-answer pair is matched, the answer in the matched question-answer pair will be output as the knowledge retrieval result of the knowledge retrieval question.

[0060] The matched question-answer pairs can be one or more. When multiple question-answer pairs are matched, the answers of multiple question-answer pairs can be output together as the knowledge retrieval result of the knowledge retrieval question. Alternatively, the answer of one of the question-answer pairs can be randomly output as the knowledge retrieval result of the knowledge retrieval question. Or, the answer of the question-answer pair with the highest corresponding similarity among multiple question-answer pairs can be output as the knowledge retrieval result of the knowledge retrieval question. Of course, there can be other methods as well, and this embodiment of the disclosure does not impose any restrictions on them.

[0061] S104: If no question-answer pair is found, perform term matching in the terminology base based on the knowledge retrieval question, and perform knowledge retrieval in the question-answer base and / or database according to the term matching results and the knowledge retrieval question to obtain knowledge retrieval results for the knowledge retrieval question.

[0062] In this embodiment, term matching in a terminology database based on a knowledge retrieval question can be performed as follows: first, the knowledge retrieval question is segmented into multiple words; then, for each word, the semantic similarity between the word and terms in the terminology database is calculated, and term matching is performed in the terminology database based on the semantic similarity and a preset similarity threshold. Alternatively, the knowledge retrieval question can be keyword-extracted to obtain target keywords; then, the semantic similarity between the target keywords and terms in the terminology database is calculated, and term matching is performed in the terminology database based on the semantic similarity and a preset similarity threshold. Of course, term matching in the terminology database can also be performed in other ways, and this embodiment does not impose any limitations on this.

[0063] It should be understood that a terminology database generally includes terms and their associated explanations. The result of term matching typically includes either a term matched in the database or a term not matched. When a term is matched in the database, it indicates that there may be a term requiring explanation in the knowledge retrieval question. Therefore, to enable the agent to more accurately understand the knowledge retrieval question and perform subsequent knowledge retrieval, it can combine the terminology explanations associated with the target term with the knowledge retrieval question to perform knowledge retrieval in a question-and-answer database and / or a general database, thereby improving the accuracy of knowledge retrieval. When no target term is matched in the database, it indicates that there may not be a term requiring explanation in the knowledge retrieval question. Therefore, it can directly perform knowledge retrieval in a question-and-answer database and / or a general database based on the knowledge retrieval question to obtain knowledge retrieval results specific to the knowledge retrieval question.

[0064] Through the above technical solutions, the knowledge retrieval process can be structured so that the agent can perform question-and-answer pair matching in the knowledge base, terminology base, question-and-answer base, and database. First, based on the knowledge retrieval question, a question-and-answer pair can be matched in the question-and-answer base. If a matching question-and-answer pair is found, the corresponding answer can be quickly output. This not only improves the stability of the response but also reduces the long waiting time caused by calling large models, thus enhancing the user experience. If no matching question-and-answer pair is found in the question-and-answer base, terminology matching can be performed in the terminology base based on the knowledge retrieval question. This allows for preliminary explanations of terms in the knowledge retrieval question, enabling the agent to more accurately understand the knowledge retrieval question and improve the accuracy and reliability of knowledge retrieval results when performing knowledge retrieval in the question-and-answer base and / or database. In addition, since the preset retrieval resource base includes at least one of the following: knowledge base, terminology base, question and answer base, and database, the knowledge retrieval link of the intelligent agent can be flexibly configured according to business needs when determining the knowledge retrieval link. Compared with the high-code approach to building the knowledge retrieval link in related technologies, it can reduce the development burden caused by changes in business needs, achieve better compatibility and reusability, significantly reduce subsequent development costs, and improve the adaptability to new application scenarios.

[0065] To facilitate understanding of the agent knowledge retrieval method based on multi-knowledge hybrid recall provided in this disclosure, the possible implementation methods in this disclosure are described below.

[0066] In some possible ways, the terminology database includes terms and their associated explanations. Accordingly, term matching based on a knowledge retrieval question within the terminology database can include:

[0067] The knowledge retrieval question is segmented into multiple words; for each word, the semantic similarity between the word and the terms in the terminology database is determined, and term matching is performed in the terminology database based on the semantic similarity and the first similarity threshold.

[0068] Accordingly, based on the term matching results and the knowledge retrieval question, a knowledge retrieval is performed in the question-answering base and / or database to obtain knowledge retrieval results for the knowledge retrieval question, which may include:

[0069] If the semantic similarity between the target word segment and the target term is greater than or equal to the first similarity threshold, a new knowledge retrieval question is obtained based on the terminology explanation and knowledge retrieval question associated with the target term; based on the new knowledge retrieval question, knowledge retrieval is performed in the question-and-answer base and / or database to obtain knowledge retrieval results for the knowledge retrieval question.

[0070] The semantic similarity between word segmentation and terms in the terminology database can be obtained by calculating the cosine similarity of word vectors between word segmentation and terms, or by calculating the edit distance between word segmentation and terms, or by other methods. This disclosure does not impose any restrictions on this method.

[0071] The term matching process in the terminology corpus, based on semantic similarity and a first similarity threshold, can be as follows: For each word segment, target terms are filtered from the terminology corpus based on the first similarity threshold and the similarity between the word segment and every term in the corpus. The similarity between the target term and the word segment must be greater than or equal to the first similarity threshold. If no target term is found in the terminology corpus, it indicates that there may be no term requiring explanation in the knowledge retrieval question. Therefore, knowledge retrieval can be directly performed in the question-and-answer database and / or other databases based on the knowledge retrieval question to obtain knowledge retrieval results for the knowledge retrieval question. If a target term is found in the terminology corpus, it indicates that there may be a term requiring explanation in the knowledge retrieval question. Therefore, a new knowledge retrieval question can be obtained based on the term explanations associated with the target term and the knowledge retrieval question; then, knowledge retrieval can be performed based on the new knowledge retrieval question, such as... Figure 3 As shown. For example, after obtaining a new knowledge retrieval question, a knowledge retrieval can be performed in a question-and-answer base and / or database to obtain knowledge retrieval results for the knowledge retrieval question, such as... Figure 4 As shown.

[0072] In possible approaches, the terminology may include semantic terms and / or expression terms, and the terminology explanation may include natural language explanations for semantic terms and / or expression explanations for expression terms. Accordingly, based on the terminology explanations associated with the target term and the knowledge retrieval question, a new knowledge retrieval question is derived, which may include:

[0073] When the target term is a semantic term, the knowledge retrieval problem is rewritten based on the natural language interpretation of the target term association using a large model of agent association, resulting in a new knowledge retrieval problem; when the target term is an expression term, the knowledge retrieval problem is expanded based on the expression interpretation of the target term association using a large model of agent association, resulting in a new knowledge retrieval problem.

[0074] It should be understood that semantic terms can refer to the meaning, interpretation, and business logic description of data or concepts, such as the definition of an indicator and / or the calculation method of an indicator. Expression terms can refer to specific forms used to achieve a certain function or calculation, such as the calculation formula of an indicator or the corresponding structured query statement.

[0075] For semantic terms, a first prompt word template can be pre-set. When a target semantic term is matched in the terminology database, the target semantic term, the associated natural language explanation, and the knowledge retrieval question can be filled into the corresponding position of the first prompt word template to obtain the first prompt word. Then, the first prompt word can be input into the large model associated with the agent. The large model can then rewrite the knowledge retrieval question based on the first prompt word to obtain a new knowledge retrieval question.

[0076] For example, a knowledge retrieval question might be: "What dairy products can be stored at low temperatures?", the target semantic term might be "low temperature", and the associated natural language interpretation of the target semantic term might be "below 12°C". Then, after rewriting the knowledge retrieval question based on the associated natural language interpretation of the target semantic term, the new knowledge retrieval question could be:

[0077] "What dairy products can be stored at low temperatures (below 12°C)?"

[0078] For expression terms, a second prompt word template can be pre-set. When a target expression term is matched in the terminology database, the target expression term, the expression explanation associated with the target expression term, and the knowledge retrieval question can be filled into the corresponding position of the second prompt word template to obtain the second prompt word. Then, the second prompt word can be input into the large model associated with the agent. The large model expands the knowledge retrieval question based on the second prompt word to obtain a new knowledge retrieval question.

[0079] For example, the knowledge retrieval question could be: "What is the customer conversion rate?", the target expression term could be "conversion rate", and the expression interpretation associated with the target expression term could be: "count(distinct(case when {{whether to purchase product}}=='yes' and {{customer stage}}=='lead' then {{username}}else NULL end)) / count(distinct(case when {{customer stage}}=='lead' then {{username}} else NULL end))". Then, by expanding the knowledge retrieval question based on the expression interpretation associated with the target expression term, the new knowledge retrieval question could be:

[0080] The query statement for conversion rate is: "count(distinct(case when {{whether to purchase product}}=='yes' and {{customer stage}}=='lead' then {{username}}else NULL end)) / count(distinct(case when {{customer stage}}=='lead' then {{username}} else NULL end))". Please use this query statement to output the customer's conversion rate.

[0081] The above method allows for term matching within a terminology database based on the knowledge retrieval question. When a target term is matched, its associated explanations can be used to supplement the terminology in the knowledge retrieval question, enabling the agent to more accurately understand the question and improving the accuracy and reliability of knowledge retrieval results in question-answering databases and / or other databases. Furthermore, in data query scenarios, large models sometimes generate inaccurate queries when producing complex statements. Adding expression terms to the terminology database allows large models to generate more accurate queries, further improving the accuracy and reliability of data queries.

[0082] Determining the knowledge retrieval path for an agent, among possible approaches, may include:

[0083] In response to the configuration operation of the retrieval resource library for knowledge retrieval by the agent on the first configuration page, a retrieval resource library node is obtained. The first configuration page is used for users to customize the knowledge retrieval link of the agent. The retrieval resource library node includes at least one of the following: knowledge base node, terminology base node, question and answer base node, and database node. In response to the connection configuration operation between the resource library nodes, the knowledge retrieval link of the agent is obtained.

[0084] For example, the first configuration page may include a first display area and a second display area. The first display area can be used to display preset search resource library types, and also allows the user to select the desired search resource library type. The search resource library types may include knowledge base types, terminology base types, question-and-answer base types, and database types, etc. The second display area can be used to display the search resource library nodes corresponding to the user-selected search resource library type, and allows the user to edit the search resource library nodes displayed in the second display area. Therefore, when a user needs to customize the knowledge retrieval link of the intelligent agent, they can select the desired search resource library type in the first display area according to actual needs. After selecting the desired search resource library type, the corresponding search resource library nodes are displayed in the second display area. Then, the search resource library nodes corresponding to the desired search resource library type can be connected according to actual needs to obtain the knowledge retrieval link of the intelligent agent. For example, if the required resource base types are question-and-answer base, terminology base, and knowledge base, the user can click on these types in the first display area, and then the question-and-answer base node, terminology base node, and knowledge base node will be displayed in the second display area. Alternatively, the user can drag and drop these types to the second display area, thus displaying them there. Then, by establishing connections between these nodes, the knowledge retrieval chain of the intelligent agent is obtained, such as... Figure 5 As shown.

[0085] Since displaying preset search resource library types on the first configuration page reduces the editing area of ​​that page, users may find it difficult to clearly build and adjust the knowledge retrieval link when it is complex, increasing the difficulty and time cost of configuration and thus affecting user experience. Therefore, to increase the editing area on the first configuration page, an add control can be displayed. When this control is triggered, preset search resource library types can be displayed, and the desired search resource library type can be selected from the preset types, after which the preset search resource library types can be de-displayed. In other words, in a possible approach, in response to the configuration operation of the search resource library used for knowledge retrieval by the agent on the first configuration page, the retrieved search resource library node can include:

[0086] In response to the triggering operation of adding a control to a node in the first configuration page, the preset search resource library types are displayed. The node adding control is used to add search resource library types for the agent to perform knowledge retrieval. The search resource library types include knowledge base type, terminology base type, question and answer base type, and database type. In response to the selection operation of the first search resource library type among the preset search resource library types, the target search resource library node belonging to the first search resource library type is obtained.

[0087] For example, such as Figure 6 As shown, a node addition control 601 can be displayed at the bottom of the first configuration page. When the user triggers the node addition control 601 through clicking or other operations, a pull-up window 602 can be displayed. The pull-up window 602 is used to display preset search resource library types, such as knowledge base type, terminology base type, question and answer base type, and database type. When the user clicks on the terminology base type, the search resource library node belonging to the terminology base type can be obtained and displayed on the first configuration page.

[0088] It should be understood that a retrieval resource repository node represents the position of the corresponding retrieval resource repository type in the knowledge retrieval chain, and the retrieval resource repository corresponding to a retrieval resource repository type can include one or more retrieval resource repositories. When the retrieval resource repository type includes multiple retrieval resource repositories, in order to further configure the knowledge retrieval chain according to user needs, after selecting the first retrieval resource repository type, an initial retrieval resource repository node that does not contain a retrieval resource repository can be displayed first. Then, through the configuration operation of the retrieval resource repository, the target retrieval resource repository node belonging to the first retrieval resource repository type can be obtained. This can further improve the matching degree between the knowledge retrieval chain and user needs, thereby improving the accuracy and reliability of knowledge retrieval results. That is to say, in possible ways, in response to the selection operation of the first retrieval resource repository type among the preset retrieval resource repository types, obtaining the target retrieval resource repository node belonging to the first retrieval resource repository type can include:

[0089] In response to the selection of a first search resource library type from the preset search resource library types, an initial search resource library node belonging to the first search resource library type is displayed on the first configuration page, wherein the search resource library corresponding to the initial search resource library node is empty; in response to the triggering operation of the initial search resource library node, a second configuration page for the initial search resource library node is displayed, and in response to the configuration operation of the search resource library corresponding to the initial search resource library node on the second configuration page, a target search resource library node belonging to the first search resource library type is obtained.

[0090] For example, continuing with the previous example, when a user clicks on the terminology database type, the initial terminology database node corresponding to that type can be displayed on the first configuration page, such as... Figure 7As shown, when a user triggers the initial terminology database node through clicking or other operations, a second configuration page can be displayed on the first configuration page. The second configuration page displays at least a resource configuration control 701 for configuring the search resource database of the initial terminology database node. When the drop-down control in the resource configuration control 701 is triggered through clicking or other operations, a drop-down menu can be displayed, and the drop-down menu displays multiple preset terminology databases. When one terminology database is selected, a search resource database node of the terminology database type is obtained.

[0091] Using the above method, the knowledge retrieval link of the intelligent agent can be customized on the first configuration page according to user needs. Compared with the knowledge retrieval link construction method in related technologies, the knowledge retrieval link construction method disclosed in this paper can effectively reduce the development burden caused by changes in business requirements, achieve better compatibility and reusability, significantly reduce subsequent development costs, and improve the adaptability to new application scenarios.

[0092] Determining the knowledge retrieval path for an agent, among possible approaches, may include:

[0093] In response to the configuration operation of the retrieval resource library for knowledge retrieval by the agent on the third configuration page, the target retrieval resource library is obtained. The third configuration page is used for the user to configure the knowledge retrieval link of the agent according to the retrieval logic template. The target retrieval resource library includes at least one of the following: knowledge base, terminology base, question and answer base, and database. The knowledge retrieval link of the agent is determined based on the target retrieval resource library and the retrieval logic template.

[0094] It should be understood that there can be one or more retrieval logic templates, and this disclosure does not impose any restrictions on this. When there are multiple retrieval logic templates, the user can select one of them to configure the knowledge retrieval link for the agent, or select multiple retrieval logic templates to configure the knowledge retrieval link for the agent. When multiple retrieval logic templates are selected to configure the knowledge retrieval link for the agent, a test page can be displayed after configuration to allow the user to test the knowledge retrieval effect of the configured multiple knowledge retrieval links and determine the knowledge retrieval link for the agent based on the test results.

[0095] In possible ways, obtaining the target retrieval resource library in response to the configuration operation of the retrieval resource library for knowledge retrieval by the agent on the third configuration page may include:

[0096] The third configuration page displays preset search resource library types. Each preset search resource library type is associated with an add control. The preset search resource library types include knowledge base type, terminology base type, question and answer base type, and database type. The add control is used to add the search resource library corresponding to the search resource library type as a search resource library for the intelligent agent to perform knowledge retrieval. In response to the triggering operation of the add control associated with the second search resource library type in the preset search resource library types, the fourth configuration page for the second search resource library type is displayed. In response to the resource configuration operation of the second search resource library type on the fourth configuration page, the target search resource library is obtained.

[0097] For example, such as Figure 8 As shown, the third configuration page can display preset search resource library types, such as knowledge base, terminology database, question and answer database, and display the associated add control 801 to the right of each preset search resource library type. When the user triggers the add control 801 associated with the second search resource library type through clicking or other operations, a fourth configuration page for the second search resource library type can be displayed. The fourth configuration page can display multiple search resource libraries associated with the second search resource library type, and for each search resource library, an add control 802 can be displayed to identify the search resource library as the search resource library corresponding to the second search resource library type in the knowledge retrieval chain. When the user triggers the target add control 802 through clicking or other operations, the search resource library corresponding to the target add control 802 is identified as the search resource library corresponding to the second search resource library type. For example, when a user triggers the add control 801 associated with the question-and-answer database type through clicking or other operations, a fourth configuration page for the question-and-answer database type can be displayed. The fourth configuration page can display the first, second, and third question-and-answer databases associated with the question-and-answer database type. When a user triggers the add control 802 associated with the second question-and-answer database through clicking or other operations, the second question-and-answer database can be used as the target retrieval resource database for the question-and-answer database type and can be displayed on the third configuration page.

[0098] It is worth noting that, in order to make the configured knowledge retrieval link more in line with the user's actual needs, when configuring based on the retrieval logic template, only part of the content in the retrieval logic template can be configured. For the content not configured in the retrieval logic template, it can be skipped directly when generating the knowledge retrieval link.

[0099] For example, if the retrieval logic template is: first, perform question-answer pair matching in the question-answer base based on the knowledge retrieval question; if no question-answer pair is matched, perform term matching in the terminology base based on the knowledge retrieval question; and then perform matching in the knowledge base based on the term matching result and the knowledge retrieval question. If question-answer pairs are not configured when configuring the knowledge retrieval link based on the retrieval logic template, then when determining the agent's knowledge retrieval link based on the configuration content and the retrieval logic template, the following knowledge retrieval link can be obtained: perform term matching in the terminology base based on the knowledge retrieval question, and then perform matching in the knowledge base based on the term matching result and the knowledge retrieval question.

[0100] The above method allows for the configuration of the agent's knowledge retrieval link on the second configuration page based on the retrieval logic template. This simplifies the configuration of the knowledge retrieval link, improves its efficiency, and enhances the user experience. Furthermore, compared to related technologies that construct knowledge retrieval links using high-code methods, the knowledge retrieval link construction method in this disclosure effectively reduces the development burden caused by changes in business requirements, achieves better compatibility and reusability, significantly reduces subsequent development costs, and improves adaptability to new application scenarios.

[0101] Based on the same concept, embodiments of this disclosure also provide an intelligent agent knowledge retrieval device based on multi-knowledge hybrid recall, such as... Figure 9 As shown, the intelligent agent knowledge retrieval device 900 based on multi-knowledge hybrid recall may include:

[0102] The determination module 901 is used to determine the knowledge retrieval link of the intelligent agent. The knowledge retrieval link is used to characterize the knowledge retrieval logic of the intelligent agent in the preset retrieval resource base. The preset retrieval resource base includes at least one of the knowledge base, terminology base, question and answer base and database.

[0103] Matching module 902 is used to perform question-answer pair matching in the question-answer database in response to receiving a knowledge retrieval question, given that the knowledge retrieval link represents the intelligent agent's knowledge retrieval logic in the knowledge base, terminology base, question-answer base, and database;

[0104] The first processing module 903 is used to output the answer in the matched question-answer pair as the knowledge retrieval result of the knowledge retrieval question when a question-answer pair is matched.

[0105] The second processing module 904 is used to perform term matching in the terminology base based on the knowledge retrieval question when no question-answer pair is matched, and to perform knowledge retrieval in the question-answer base and / or database according to the term matching results and the knowledge retrieval question, so as to obtain the knowledge retrieval results for the knowledge retrieval question.

[0106] The aforementioned intelligent agent knowledge retrieval device 900, based on multi-knowledge hybrid recall, can represent the agent's knowledge retrieval logic in knowledge bases, terminology bases, question-and-answer bases, and databases. First, it performs question-and-answer pair matching in the question-and-answer base based on the knowledge retrieval question. If a matching question-and-answer pair is found, the corresponding answer can be quickly output. This not only improves the stability of the response but also reduces long waiting times caused by calling large models, enhancing the user experience. If no matching question-and-answer pair is found in the question-and-answer base, it can perform terminology matching in the terminology base based on the knowledge retrieval question. This allows for preliminary explanations of terms in the knowledge retrieval question, enabling the agent to more accurately understand the knowledge retrieval question and improve the accuracy and reliability of knowledge retrieval results when performing knowledge retrieval in the question-and-answer base and / or database. In addition, since the preset retrieval resource base includes at least one of the following: knowledge base, terminology base, question and answer base, and database, the knowledge retrieval link of the intelligent agent can be flexibly configured according to business needs when determining the knowledge retrieval link. Compared with the high-code approach to building the knowledge retrieval link in related technologies, it can reduce the development burden caused by changes in business needs, achieve better compatibility and reusability, significantly reduce subsequent development costs, and improve the adaptability to new application scenarios.

[0107] In one possible manner, the terminology database includes terms and related term explanations; accordingly, the second processing module 904 may include:

[0108] The first processing submodule is used to perform word segmentation on the knowledge retrieval question to obtain multiple words;

[0109] The second processing submodule is used to determine the semantic similarity between each word segment and the terms in the terminology database, and to perform term matching in the terminology database based on the semantic similarity and the first similarity threshold.

[0110] The third processing submodule is used to obtain a new knowledge retrieval question based on the terminology explanation and knowledge retrieval question associated with the target term if the semantic similarity between the target word segment and the target term is greater than or equal to the first similarity threshold.

[0111] The retrieval submodule is used to perform knowledge retrieval in the question-and-answer base and / or database based on the new knowledge retrieval question, and obtain the knowledge retrieval results for the knowledge retrieval question.

[0112] In possible embodiments, the terminology includes semantic terms and / or expression terms, and the term interpretation includes natural language interpretation for semantic terms and / or expression interpretation for expression terms. Accordingly, the third processing submodule may include:

[0113] The first processing unit is used to rewrite the knowledge retrieval problem based on the natural language interpretation of the target term association through a large model of agent association when the target term is a semantic term, so as to obtain a new knowledge retrieval problem.

[0114] The second processing unit is used to expand the knowledge retrieval problem by interpreting the expression associated with the target term through a large model of agent association when the target term is an expression term, thereby obtaining a new knowledge retrieval problem.

[0115] In one possible manner, the question-and-answer database includes multiple question-and-answer pairs, and each question-and-answer pair includes a question and an answer; accordingly, the matching module 902 may include:

[0116] The first determination submodule is used to, in response to receiving a knowledge retrieval question, determine the similarity between the question and the knowledge retrieval question for each question-answer pair;

[0117] The matching submodule is used to perform question-answer pair matching in the question-answer database based on similarity and a second similarity threshold.

[0118] In some possible ways, the determining module 901 may include:

[0119] The first configuration submodule is used to respond to the configuration operation of the retrieval resource library for knowledge retrieval by the agent on the first configuration page and obtain the retrieval resource library node. The first configuration page is used for users to customize the knowledge retrieval link of the agent. The retrieval resource library node includes at least one of the following: knowledge base node, terminology base node, question and answer base node, and database node.

[0120] The second configuration submodule is used to obtain the knowledge retrieval link of the agent in response to the connection configuration operation between the resource library nodes.

[0121] In some possible ways, the first configuration submodule may include:

[0122] The first display unit is used to respond to the trigger operation of adding controls to nodes in the first configuration page and display the preset search resource library types. The node adding control is used to add search resource library types for the intelligent agent to perform knowledge retrieval. The search resource library types include knowledge base type, terminology base type, question and answer base type and database type.

[0123] The selection unit is used to respond to the selection operation of the first search resource type in the preset search resource type to obtain the target search resource node belonging to the first search resource type.

[0124] Among the possible approaches, the selection unit may include:

[0125] The display subunit is used to respond to the selection operation of the first search resource library type in the preset search resource library type, and to display the initial search resource library node belonging to the first search resource library type in the first configuration page, wherein the search resource library corresponding to the initial search resource library node is empty;

[0126] The processing subunit is configured to, in response to a triggering operation on the initial search resource repository node, display a second configuration page for the initial search resource repository node, and, in response to a configuration operation on the search resource repository corresponding to the initial search resource repository node in the second configuration page, obtain a target search resource repository node belonging to the first search resource repository type.

[0127] In some possible ways, the determining module 901 may include:

[0128] The third configuration submodule is used to respond to the configuration operation of the retrieval resource library for knowledge retrieval by the agent on the third configuration page and obtain the target retrieval resource library. The third configuration page is used to allow the user to configure the knowledge retrieval link of the agent according to the retrieval logic template. The target retrieval resource library includes at least one of the following: knowledge base, terminology base, question and answer base, and database.

[0129] The second determination submodule is used to determine the knowledge retrieval link of the intelligent agent based on the target retrieval resource library and retrieval logic template.

[0130] In some possible ways, the third configuration submodule may include:

[0131] The second display unit is used to display preset search resource library types on the third configuration page. Each preset search resource library type is associated with an add control. The preset search resource library types include knowledge base type, terminology base type, question and answer base type and database type. The add control is used to add the search resource library corresponding to the search resource library type as a search resource library for the intelligent agent to perform knowledge retrieval.

[0132] The third processing unit is used to respond to the triggering operation of the add control associated with the second search resource library type in the preset search resource library type, display the fourth configuration page for the second search resource library type, and respond to the resource configuration operation of the second search resource library type on the fourth configuration page to obtain the target search resource library.

[0133] Based on the same concept, embodiments of this disclosure also provide a computer-readable medium having a computer program stored thereon, which, when executed by a processing device, implements the steps of any of the above-described intelligent agent knowledge retrieval methods based on multi-knowledge hybrid recall.

[0134] Based on the same concept, this disclosure also provides an electronic device that may include:

[0135] A storage device on which computer programs are stored;

[0136] A processing device is used to execute a computer program stored in a storage device to implement the steps of any of the above-described intelligent agent knowledge retrieval methods based on multi-knowledge hybrid recall.

[0137] Based on the same concept, embodiments of this disclosure also provide a computer program product, including a computer program that, when executed by a processor, implements the steps of any of the above-described intelligent agent knowledge retrieval methods based on multi-knowledge hybrid recall.

[0138] The following is for reference. Figure 10 The diagram illustrates a structural schematic of an electronic device 1000 suitable for implementing embodiments of the present disclosure. The terminal devices in the embodiments of the present disclosure may include, but are not limited to, mobile terminals such as mobile phones, laptops, digital broadcast receivers, PDAs (personal digital assistants), PADs (tablet computers), PMPs (portable multimedia players), in-vehicle terminals (e.g., in-vehicle navigation terminals), and fixed terminals such as digital TVs and desktop computers. Figure 10 The electronic device shown is merely an example and should not be construed as limiting the functionality and scope of the embodiments disclosed herein.

[0139] like Figure 10 As shown, the electronic device 1000 may include a processing unit (e.g., a central processing unit, a graphics processing unit, etc.) 1001, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 1002 or a program loaded from a storage device 1008 into a random access memory (RAM) 1003. The RAM 1003 also stores various programs and data required for the operation of the electronic device 1000. The processing unit 1001, ROM 1002, and RAM 1003 are interconnected via a bus 1004. An input / output (I / O) interface 1005 is also connected to the bus 1004.

[0140] Typically, the following devices can be connected to the I / O interface 1005: input devices 1006 including, for example, a touchscreen, touchpad, keyboard, mouse, camera, microphone, accelerometer, gyroscope, etc.; output devices 1007 including, for example, a liquid crystal display (LCD), speaker, vibrator, etc.; storage devices 1008 including, for example, magnetic tape, hard disk, etc.; and communication devices 1009. Communication device 1009 allows electronic device 1000 to communicate wirelessly or wiredly with other devices to exchange data. Although Figure 10 An electronic device 1000 with various devices is shown; however, it should be understood that it is not required to implement or possess all of the devices shown. More or fewer devices may be implemented or possessed alternatively.

[0141] In particular, according to embodiments of this disclosure, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments of this disclosure include a computer program product comprising a computer program carried on a non-transitory computer-readable medium, the computer program containing program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via communication device 1009, or installed from storage device 1008, or installed from ROM 1002. When the computer program is executed by processing device 1001, it performs the functions defined in the methods of embodiments of this disclosure.

[0142] It should be noted that the computer-readable medium described in this disclosure can be a computer-readable signal medium or a computer-readable storage medium, or any combination thereof. A computer-readable storage medium can be, for example,—but not limited to—an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of a computer-readable storage medium may include, but are not limited to: an electrical connection having one or more wires, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof. In this disclosure, a computer-readable storage medium can be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device. In this disclosure, a computer-readable signal medium can include a data signal propagated in baseband or as part of a carrier wave, carrying computer-readable program code. Such propagated data signals can take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. A computer-readable signal medium can be any computer-readable medium other than a computer-readable storage medium, which can send, propagate, or transmit a program for use by or in connection with an instruction execution system, apparatus, or device. The program code contained on the computer-readable medium can be transmitted using any suitable medium, including but not limited to: wires, optical fibers, RF (radio frequency), etc., or any suitable combination thereof.

[0143] In some implementations, communication can be conducted using any currently known or future-developed network protocol such as HTTP (Hypertext Transfer Protocol), and can be interconnected with digital data communication (e.g., communication networks) of any form or medium. Examples of communication networks include local area networks (“LANs”), wide area networks (“WANs”), the Internet (e.g., the Internet of Things), and end-to-end networks (e.g., ad hoc end-to-end networks), as well as any currently known or future-developed networks.

[0144] The aforementioned computer-readable medium may be included in the aforementioned electronic device; or it may exist independently and not assembled into the electronic device.

[0145] The aforementioned computer-readable medium carries one or more programs. When the electronic device executes the aforementioned one or more programs, the electronic device causes the electronic device to: determine a knowledge retrieval link for an intelligent agent, the knowledge retrieval link being used to characterize the intelligent agent's knowledge retrieval logic in a preset retrieval resource base, the preset retrieval resource base including at least one of a knowledge base, a terminology base, a question-and-answer base, and a database; when the knowledge retrieval link characterizes the intelligent agent's knowledge retrieval logic in the knowledge base, terminology base, question-and-answer base, and database, in response to receiving a knowledge retrieval question, perform question-and-answer pair matching in the question-and-answer base based on the knowledge retrieval question; if a question-and-answer pair is matched, output the answer in the matched question-and-answer pair as the knowledge retrieval result for the knowledge retrieval question; if no question-and-answer pair is matched, perform terminology matching in the terminology base based on the knowledge retrieval question, and perform knowledge retrieval in the question-and-answer base and / or database based on the terminology matching result and the knowledge retrieval question, to obtain the knowledge retrieval result for the knowledge retrieval question.

[0146] Computer program code for performing the operations of this disclosure can be written in one or more programming languages ​​or a combination thereof, including but not limited to object-oriented programming languages ​​such as Java, Smalltalk, and C++, as well as conventional procedural programming languages ​​such as the "C" language or similar programming languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving remote computers, the remote computer can be connected to the user's computer via any type of network—including a local area network (LAN) or a wide area network (WAN)—or can be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0147] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of this disclosure. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, can be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0148] The modules described in the embodiments of this disclosure can be implemented in software or hardware. The names of the modules are not, in some cases, intended to limit the functionality of the module itself.

[0149] The functions described above in this document can be performed at least in part by one or more hardware logic components. For example, exemplary types of hardware logic components that can be used, without limitation, include: field-programmable gate arrays (FPGAs), application-specific integrated circuits (ASICs), application-specific standard products (ASSPs), system-on-a-chip (SoCs), complex programmable logic devices (CPLDs), and so on.

[0150] In the context of this disclosure, a machine-readable medium can be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, apparatus, or device. A machine-readable medium can be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium can be, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination of the foregoing. More specific examples of machine-readable storage media include electrical connections based on one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the foregoing.

[0151] The above description is merely a preferred embodiment of this disclosure and an explanation of the technical principles employed. Those skilled in the art should understand that the scope of this disclosure is not limited to technical solutions formed by specific combinations of the above-described technical features, but should also cover other technical solutions formed by arbitrary combinations of the above-described technical features or their equivalents without departing from the above-described concept. For example, technical solutions formed by substituting the above features with (but not limited to) technical features disclosed in this disclosure that have similar functions.

[0152] Furthermore, while the operations are described in a specific order, this should not be construed as requiring these operations to be performed in the specific order shown or in a sequential order. In certain environments, multitasking and parallel processing may be advantageous. Similarly, while several specific implementation details are included in the above discussion, these should not be construed as limiting the scope of this disclosure. Certain features described in the context of individual embodiments may also be implemented in combination in a single embodiment. Conversely, various features described in the context of a single embodiment may also be implemented individually or in any suitable sub-combination in multiple embodiments.

[0153] Although the subject matter has been described using language specific to structural features and / or methodological logic, it should be understood that the subject matter defined in the appended claims is not necessarily limited to the specific features or actions described above. Rather, the specific features and actions described above are merely illustrative forms of implementing the claims. Regarding the apparatus in the above embodiments, the specific manner in which the various modules perform their operations has been described in detail in the embodiments relating to the method, and will not be elaborated upon here.

Claims

1. A knowledge retrieval method for intelligent agents based on multi-knowledge hybrid recall, characterized in that, include: A knowledge retrieval link for an intelligent agent is determined. The knowledge retrieval link is used to characterize the knowledge retrieval logic of the intelligent agent in a preset retrieval resource base. The preset retrieval resource base includes at least one of a knowledge base, a terminology base, a question-and-answer base, and a database. The question-and-answer base includes multiple question-and-answer pairs, and each question-and-answer pair includes a question and the answer corresponding to the question. The terminology base includes terms and terminology explanations associated with the terms. The terms include semantic terms and / or expression terms. The terminology explanations include natural language explanations for the semantic terms and / or expression explanations for the expression terms. When the knowledge retrieval link represents the knowledge retrieval logic of the agent in the knowledge base, the terminology base, the question-and-answer base, and the database, in response to receiving a knowledge retrieval question, the agent performs question-and-answer pair matching in the question-and-answer base based on the knowledge retrieval question; If a question-answer pair is matched, the answer in the matched question-answer pair will be output as the knowledge retrieval result of the knowledge retrieval question. If no question-answer pair is found, term matching is performed in the terminology database based on the knowledge retrieval question to obtain the explanation of the target term. When the target term is interpreted as a natural language interpretation of the semantic term, the knowledge retrieval problem is rewritten based on the target term interpretation using the large model associated with the agent to obtain a new knowledge retrieval problem; or, when the target term is interpreted as an expression interpretation of the expression term, the knowledge retrieval problem is expanded based on the target term interpretation using the large model associated with the agent to obtain a new knowledge retrieval problem. Based on the new knowledge retrieval question, a knowledge retrieval is performed in the knowledge base and / or the database to obtain knowledge retrieval results for the knowledge retrieval question.

2. The intelligent agent knowledge retrieval method based on multi-knowledge hybrid recall according to claim 1, characterized in that, The process of matching terms in the terminology database based on the knowledge retrieval question to obtain the explanation of the target term includes: The knowledge retrieval problem is processed by word segmentation to obtain multiple word segments; For each word segment, the semantic similarity between the word segment and a term in the terminology database is determined, and term matching is performed in the terminology database based on the semantic similarity and a first similarity threshold. If the semantic similarity between the target word segment and the target term is greater than or equal to the first similarity threshold, the term interpretation of the target term will be used as the target term interpretation.

3. The intelligent agent knowledge retrieval method based on multi-knowledge hybrid recall according to claim 1, characterized in that, The response to receiving a knowledge retrieval question, and the matching of question-answer pairs in the question-answer database based on the knowledge retrieval question, includes: In response to receiving a knowledge retrieval question, for each question in the question-answer pair, the similarity between the question and the knowledge retrieval question is determined; Based on the aforementioned similarity and the second similarity threshold, question-answer pair matching is performed in the question-answer database.

4. The intelligent agent knowledge retrieval method based on multi-knowledge hybrid recall according to any one of claims 1-3, characterized in that, The knowledge retrieval link for determining the intelligent agent includes: In response to the configuration operation of the retrieval resource library for knowledge retrieval by the agent on the first configuration page, a retrieval resource library node is obtained, wherein the first configuration page is used for users to customize the knowledge retrieval link of the agent, and the retrieval resource library node includes at least one of a knowledge base node, a terminology base node, a question and answer base node, and a database node. In response to the connection configuration operation between the resource repository nodes, the knowledge retrieval link of the agent is obtained.

5. The intelligent agent knowledge retrieval method based on multi-knowledge hybrid recall according to claim 4, characterized in that, The response to the configuration operation of the retrieval resource library for knowledge retrieval by the agent on the first configuration page, obtaining the retrieval resource library node, includes: In response to the triggering operation of adding a control to a node in the first configuration page, a preset search resource library type is displayed. The node adding control is used to add a search resource library type for the agent to perform knowledge retrieval. The search resource library type includes knowledge base type, terminology base type, question and answer base type, and database type. In response to the selection operation of the first search resource type among the preset search resource types, a target search resource node belonging to the first search resource type is obtained.

6. The intelligent agent knowledge retrieval method based on multi-knowledge hybrid recall according to claim 5, characterized in that, The step of responding to the selection operation of the first search resource type among the preset search resource types, and obtaining the target search resource node belonging to the first search resource type, includes: In response to the selection operation of the first search resource library type among the preset search resource library types, an initial search resource library node belonging to the first search resource library type is displayed on the first configuration page, wherein the search resource library corresponding to the initial search resource library node is empty; In response to the triggering operation of the initial search resource library node, a second configuration page for the initial search resource library node is displayed, and in response to the configuration operation of the search resource library corresponding to the initial search resource library node in the second configuration page, a target search resource library node belonging to the first search resource library type is obtained.

7. The intelligent agent knowledge retrieval method based on multi-knowledge hybrid recall according to any one of claims 1-3, characterized in that, The knowledge retrieval link for determining the intelligent agent includes: In response to the configuration operation of the retrieval resource library for knowledge retrieval by the agent on the third configuration page, a target retrieval resource library is obtained. The third configuration page is used for the user to configure the knowledge retrieval link of the agent according to the retrieval logic template. The target retrieval resource library includes at least one of a knowledge base, a terminology base, a question and answer base, and a database. Based on the target retrieval resource library and the retrieval logic template, the knowledge retrieval link of the intelligent agent is determined.

8. The intelligent agent knowledge retrieval method based on multi-knowledge hybrid recall according to claim 7, characterized in that, The process of configuring the retrieval resource library for knowledge retrieval by the agent on the third configuration page, and obtaining the target retrieval resource library, includes: The third configuration page displays preset search resource library types. Each preset search resource library type is associated with an add control. The preset search resource library types include knowledge base type, terminology base type, question and answer base type, and database type. The add control is used to add the search resource library corresponding to the search resource library type as the search resource library for the agent to perform knowledge retrieval. In response to the triggering operation of the add control associated with the second search resource library type in the preset search resource library type, a fourth configuration page for the second search resource library type is displayed, and in response to the resource configuration operation of the second search resource library type on the fourth configuration page, the target search resource library is obtained.

9. A knowledge retrieval device for intelligent agents based on multi-knowledge hybrid recall, characterized in that, include: A determination module is used to determine the knowledge retrieval link of an intelligent agent. The knowledge retrieval link is used to characterize the knowledge retrieval logic of the intelligent agent in a preset retrieval resource library. The preset retrieval resource library includes at least one of a knowledge base, a terminology base, a question-and-answer base, and a database. The question-and-answer base includes multiple question-and-answer pairs, and each question-and-answer pair includes a question and the answer corresponding to the question. The terminology base includes terms and terminology explanations associated with the terms. The terms include semantic terms and / or expression terms. The terminology explanations include natural language explanations for the semantic terms and / or expression explanations for the expression terms. The matching module is configured to, in response to receiving a knowledge retrieval question, perform question-answer pair matching in the question-answer database based on the knowledge retrieval question, when the knowledge retrieval link represents the knowledge retrieval logic of the agent in the knowledge base, the terminology base, the question-answer base, and the database; The first processing module is used to output the answer in the matched question-answer pair as the knowledge retrieval result of the knowledge retrieval question when a question-answer pair is matched. The second processing module is used to, when no question-answer pair is matched, perform term matching in the terminology base based on the knowledge retrieval question to obtain a target term explanation; if the target term explanation is a natural language explanation for the semantic term, rewrite the knowledge retrieval question based on the target term explanation using the large model associated by the agent to obtain a new knowledge retrieval question; or if the target term explanation is an expression explanation for the expression term, expand the knowledge retrieval question based on the target term explanation using the large model associated by the agent to obtain a new knowledge retrieval question; and perform knowledge retrieval in the knowledge base and / or the database according to the new knowledge retrieval question to obtain knowledge retrieval results for the knowledge retrieval question.

10. A computer-readable medium having a computer program stored thereon, characterized in that, When executed by a processing device, the computer program performs the steps of the method according to any one of claims 1-8.

11. An electronic device, characterized in that, include: A storage device on which computer programs are stored; A processing device for executing the computer program in the storage device to implement the steps of the method according to any one of claims 1-8.

12. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1-8.

Images