Response method and apparatus, electronic device, and storage medium

Abstract

The present disclosure provides a response method and device, electronic equipment and storage medium, relates to the technical field of artificial intelligence, in particular to the technical field of large language model, generation model and intelligent question answering. The specific implementation scheme is: in response to receiving input information from a target object, determining a scene type according to the input information; in response to determining that the scene type belongs to a preset scene type, determining an object description feature of the target object corresponding to the scene type; generating an object description text according to the object description feature; and generating a response information according to the object description text and a retrieval text, wherein the retrieval text is obtained by retrieving the input information.

Description

Technical Field

[0001] This disclosure relates to the field of artificial intelligence technology, and more particularly to the fields of large language models, generative models, and intelligent question answering technology. More specifically, this disclosure provides a response method, apparatus, electronic device, storage medium, and computer program product. Background Technology

[0002] In recent years, large model technologies have demonstrated excellent generation, understanding, and reasoning capabilities, but they still face challenges in question-answering tasks, such as slow knowledge updates and high illusion rates. To address these issues, search-augmented generation (RAG) provides large models with relevant documents / fragments, helping them better acquire real-time knowledge and generate more accurate answers. Summary of the Invention

[0003] This disclosure provides a response method, apparatus, electronic device, storage medium, and computer program product.

[0004] According to the first aspect, a response method is provided, the method comprising: in response to receiving input information from a target object, determining a scene type based on the input information; in response to determining that the scene type belongs to a preset scene type, determining an object description feature of the target object corresponding to the scene type; generating object description text based on the object description feature; and generating response information based on the object description text and the search text, wherein the search text is obtained by searching based on the input information.

[0005] According to a second aspect, an information interaction device is provided, comprising: a scene type determination module, configured to determine a scene type based on input information received from a target object; an object description feature determination module, configured to determine object description features of the target object corresponding to the scene type in response to determining that the scene type belongs to a preset scene type; an object description text determination module, configured to generate object description text based on the object description features; and a first response module, configured to generate response information based on the object description text and the search text, wherein the search text is obtained after searching based on the input information.

[0006] According to a third aspect, an electronic device is provided, comprising: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to perform a method provided according to the present disclosure.

[0007] According to a fourth aspect, a non-transitory computer-readable storage medium is provided that stores computer instructions for causing a computer to perform the methods provided in this disclosure.

[0008] According to a fifth aspect, a computer program product is provided, comprising a computer program stored on at least one of a readable storage medium and an electronic device, wherein the computer program, when executed by a processor, implements the method provided in this disclosure.

[0009] It should be understood that the description in this section is not intended to identify key or essential features of the embodiments of this disclosure, nor is it intended to limit the scope of this disclosure. Other features of this disclosure will become readily apparent from the following description. Attached Figure Description

[0010] The accompanying drawings are provided to better understand this solution and do not constitute a limitation of this disclosure. Wherein:

[0011] Figure 1 This is an exemplary system architecture diagram of a response method and apparatus applicable according to an embodiment of the present disclosure;

[0012] Figure 2 This is a flowchart of a response method according to an embodiment of the present disclosure;

[0013] Figure 3 This is a system schematic diagram of a response method according to an embodiment of the present disclosure;

[0014] Figure 4 This is a system schematic diagram of a response method according to an embodiment of the present disclosure;

[0015] Figure 5 This is a block diagram of a response device according to an embodiment of the present disclosure; and

[0016] Figure 6 This is a block diagram of an electronic device using a response method according to an embodiment of the present disclosure. Detailed Implementation

[0017] The exemplary embodiments of this disclosure are described below with reference to the accompanying drawings, including various details of the embodiments to aid understanding, and should be considered merely exemplary. Therefore, those skilled in the art will recognize that various changes and modifications can be made to the embodiments described herein without departing from the scope and spirit of this disclosure. Similarly, for clarity and brevity, descriptions of well-known functions and structures are omitted in the following description.

[0018] Knowledge-based intelligent question answering is currently generally implemented using RAG (Related Aspects of Language) technology. This involves searching relevant documents to allow a large model to acquire more up-to-date and relevant knowledge, thereby generating more suitable answers. However, current RAG technology is often generic and lacks awareness of individual user characteristics, resulting in generic, redundant answers. Large models include, for example, Large Language Models (LLMs).

[0019] One method to enhance answer personalization based on user history involves using the user's historical conversations as part of the LLM's prompts. This allows the LLM to reference user information when generating answers, thus enabling partial personalization. For example, if user A previously mentioned liking a certain singer, that singer can be added as an interest tag for user A. When user A subsequently requests song recommendations from the LLM, songs by that singer can be suggested.

[0020] However, tags mined from users' historical conversations cannot accurately represent their characteristics because they require explicit expression from users, increasing their operational costs. Secondly, the large amount and complexity of information in conversation data can lead to inaccurate tag mining. Finally, if users no longer update their tags, the data becomes severely outdated.

[0021] The collection, storage, use, processing, transmission, provision, and disclosure of user personal information involved in the technical solution disclosed herein comply with the provisions of relevant laws and regulations and do not violate public order and good morals.

[0022] Figure 1 This is a schematic diagram of an exemplary system architecture to which a response method and apparatus can be applied according to an embodiment of this disclosure. It should be noted that... Figure 1 The examples shown are merely examples of system architectures that can be applied to the embodiments of this disclosure, in order to help those skilled in the art understand the technical content of this disclosure, but do not mean that the embodiments of this disclosure cannot be used in other devices, systems, environments or scenarios.

[0023] like Figure 1 As shown, the system architecture 100 according to this embodiment may include terminal devices 101, 102, and 103, a network 104, and a server 105. The network 104 serves as a medium for providing a communication link between the terminal devices 101, 102, and 103 and the server 105. The network 104 may include various connection types, such as wired and / or wireless communication links, etc.

[0024] Users can use terminal devices 101, 102, and 103 to interact with server 105 via network 104 to receive or send messages, etc. Terminal devices 101, 102, and 103 can be various electronic devices, including but not limited to smartphones, tablets, laptops, etc.

[0025] Server 105 can be a server that provides various services, such as a backend management server that supports websites browsed by users using terminal devices 101, 102, and 103 (for example only). The backend management server can analyze and process data such as received user requests and feed the processing results back to the terminal devices.

[0026] The response method provided in this disclosure can generally be executed by server 105. Correspondingly, the information interaction device provided in this disclosure can generally be located in server 105. The response method provided in this disclosure can also be executed by a server or server cluster that is different from server 105 and capable of communicating with terminal devices 101, 102, 103 and / or server 105. Correspondingly, the information interaction device provided in this disclosure can also be located in a server or server cluster that is different from server 105 and capable of communicating with terminal devices 101, 102, 103 and / or server 105.

[0027] Figure 2 This is a flowchart of a response method according to an embodiment of the present disclosure.

[0028] like Figure 2 As shown, the response method 200 includes operations S210 to S240.

[0029] In operation S210, in response to receiving input information from the target object, the scene type is determined based on the input information.

[0030] The target object can be a user, and the input information can be a question entered by the user. Based on the semantics of the input information, we can determine which scenario type the input information belongs to.

[0031] The embodiments disclosed herein can be applied to enterprise office scenarios. The scenario types in enterprise office scenarios can include human resources scenarios, financial scenarios, system scenarios, office scenarios, etc. Human resources scenarios can include specific scenarios such as leave requests and policies; financial scenarios can include specific scenarios such as expense reimbursements and invoicing; system scenarios can include specific scenarios such as networks, assets, and departments; and office scenarios can include specific scenarios such as meetings, to-do items, and document retrieval.

[0032] The input information can be classified into text categories to obtain the scene type corresponding to the input information.

[0033] Large-scale modeling (LLM) can also be used to determine the scenario type. For example, a prompt word can be written, which requires determining the scenario type based on the semantics of the question. If the scenario is determined to be human resources, finance, system, office, etc., the LLM can then be requested to obtain the specific scenario within that scenario, such as requesting leave or finding a document.

[0034] In operation S220, in response to determining that the scene type belongs to a preset scene type, the object description features of the target object corresponding to the scene type are determined.

[0035] The preset scenarios can include the aforementioned scenarios such as personnel, finance, systems, and office work, as well as specific scenarios within each scenario. If the semantics of the input information determine that the current problem belongs to a target scenario within the preset scenarios, the object description features corresponding to that target scenario can be determined.

[0036] A target object can have multiple object description features, and there can be at least one object description feature corresponding to the scene type. Multiple object description features can include: basic description features, historical operation description features, role description features, and social relationship description features, etc.

[0037] The descriptive features of a target object can also be called a user profile. Multiple descriptive features of a target object can form a multi-dimensional user profile of a target user. For example, the basic descriptive features of a target object are a basic profile, the historical operation descriptive features of a target object are a work profile, the role descriptive features of a target user are a role profile, and the social relationship descriptive features of a target user are a social relationship profile.

[0038] User profiles across multiple dimensions can be pre-built for the target user. The basic profile can include the user's basic information within the company, such as department, organizational structure, and length of service. The role profile can include the user's role and work direction. The work profile can include the projects, platforms, and products the user has worked on. The social relationship profile can include the user's social relationships and superior-subordinate relationships within the company.

[0039] The various scenario types mentioned above can be pre-established with corresponding user profiles across multiple dimensions. For example, the department scenario corresponds to the basic profile, while the document search scenario corresponds to the work profile, role profile, and social relationship profile. These correspondences can be stored in a scenario-profile mapping table.

[0040] For example, if a user's question is "How do I get to Department X, and on which floor is it?", we can determine that the scenario type is a department scenario, and obtain the basic profile based on the scenario-profile mapping table.

[0041] For example, if a user's question is "find documents in project Y", the scenario type can be determined as a document search scenario. Based on the scenario-profile mapping table, the work profile, role profile, and social relationship profile can be obtained.

[0042] In the technical solution disclosed herein, the acquisition and use of user-related feature information or profile information are known and agreed to by the users, comply with the provisions of relevant laws and regulations, and do not violate public order and good morals.

[0043] In operation S230, object description text is generated based on object description characteristics.

[0044] Object description features can be in a structured data format (such as JOIN format). For ease of subsequent processing, in one example, the object description features can be directly described using natural language to obtain the object description text.

[0045] In another example, since the object description features corresponding to a scene type may contain many features, to more accurately answer user questions, object description features relevant to the user question can be identified from the scene type's object description features. These user question-related object description features can be called target profile information. This target profile information can then be converted into object description text in natural language.

[0046] For example, basic descriptive features (or basic profiles) include department information, organizational structure information, and years of service. Given the input information as "How to get to Department X, which floor?", by searching the basic profile, the user's department information can be obtained. This user-specific information constitutes the target profile information. The user's department information can then be converted into text, resulting in a descriptive text for the user's department. This descriptive text for the user's department is the object description text required for this response task.

[0047] In operation S240, response information is generated based on the object description text and the search text, wherein the search text is obtained after searching based on the input information.

[0048] The system can retrieve the knowledge base based on input information and obtain the search text. This search text can be a document or information fragment related to the input information. For example, if the input information is "How to get to Department X, which floor is it on?", the system can retrieve information about Department X in multiple geographical locations, such as Department X in location A, Department X in location B, and so on.

[0049] Object description text can be text converted from object description features corresponding to a scenario type, containing target profile information of the target object in a specific scenario. For example, if the input information is "How to get to Department X, on which floor?", and the scenario type is determined to be a department scenario, searching the basic profile in the department scenario can obtain the user's department information. This department information is the target profile information. Converting the structured department information into text yields the object description text.

[0050] The response information can be an answer to the input information. Input information, search text, and object description text can be input into the LLM to obtain the response information.

[0051] For example, you can input the information "How to get to Department X, which floor?", object description text (a description of the user's department), and search text (information about Department X in multiple geographical locations). Since the description text of the user's department describes its geographical location, the search text can be used to determine the Department X information in the user's location, which is the information the user is actually looking for. This results in a response that is more closely aligned with the user's needs.

[0052] The embodiments of this disclosure determine the scene type based on the input information. If the scene type belongs to a preset scene type, the object description features corresponding to the scene type are determined. Object description text is generated based on the object description features. Response information is generated based on the object description text and the search text. Since the search text is a knowledge document or information fragment related to the input information, and the object description text characterizes the user characteristics in a specific scene, generating response information based on the object description text and the search text to reply to input information in a specific scene can generate response information that better matches the needs of the target object corresponding to the input information, thereby improving the matching degree of the response information.

[0053] Figure 3 This is a system schematic diagram of a response method according to an embodiment of the present disclosure.

[0054] like Figure 3 As shown, the system in this embodiment may include a portrait retrieval module 310, a document retrieval module 320, a personalization detection module 330, and a generation module 340.

[0055] The user has a user identifier 301. On the one hand, after the user enters a user question 302, the profile retrieval module 310 can perform a profile retrieval based on the user identifier 301 and the user question 302.

[0056] For example, based on user question 302, the scenario type can be determined. The scenario type can be any of the following scenarios: personnel scenario, financial scenario, system scenario, office scenario. If the scenario type is determined to belong to any of the above scenarios, the specific scenario can be further determined. For example, the specific scenario can be a leave request scenario in personnel scenario, a department scenario in system scenario, a document search scenario in office scenario, etc.

[0057] Multiple user profiles can be determined based on user identifier 301. Based on the specific scenario type to which user question 302 belongs, the user profile 311 corresponding to that specific scenario type can be determined from the profile database. For example, the department scenario corresponds to the basic profile, and the document search scenario corresponds to the work profile, role profile, and social relationship profile.

[0058] A user profile database can be built offline based on enterprise domain user data (303). Enterprise domain user data (303) can include basic information about each user within the enterprise, user-written documents, weekly reports, code, and other documents, as well as user communication information and hierarchical relationship information. Based on enterprise domain user data (303), multi-dimensional user profiles for each user can be constructed. These multi-dimensional user profiles can include basic profiles, role profiles, work profiles, and social relationship profiles. Storing these multi-dimensional user profiles of each user within the enterprise in a database yields the user profile database.

[0059] According to embodiments of this disclosure, the construction of a user profile includes determining basic descriptive features based on the basic information of the target object; determining historical operation descriptive features and role descriptive features from the historical operation data of the target object; and determining social relationship descriptive features based on the historical social information of the target object.

[0060] The target audience can be the target user, and the basic information can include the user's basic information within the company, such as department, organizational structure, and length of service. Basic information can be directly retrieved from the company's database, and basic descriptive features can be generated based on this information to form a basic profile.

[0061] Historical behavioral data can include user-written documents, weekly reports, code, and other materials. LLM (Local Management Model) can be used to extract role description features from this data, creating a user profile. For example, an LLM prompt can be constructed, allowing the LLM to output a user's role description features based on their historical behavioral data. These role description features can include the user's role information and work direction information.

[0062] LLM can also be used to extract descriptive features of users' historical behavior from historical behavioral data, creating a user profile. For example, an LLM prompt can be constructed, allowing the LLM to output descriptive features of a user's historical behavior based on their historical behavioral data. These descriptive features can include information about projects, platforms, and products they have experienced.

[0063] To ensure the effectiveness of extracting the aforementioned character description features and historical behavior description features, several similar examples can be constructed for LLM reference, enabling LLM to output more accurate character profiles and job profiles.

[0064] Historical social information can include a user's past communication records, shared meeting information with other users, and hierarchical relationships. Social relationship description features of a target user can be generated based on their historical social information, forming a social relationship profile of the target user.

[0065] Hierarchical relationships can generally be obtained from company databases. Explicit social relationships, such as users' historical communication information and shared meeting information, can be mined, and a social network can be constructed based on the mined information. Each user in a social network can be considered a node, and the connections between users are edges. For example, if user A's historical social records include interactions with user B, and user A has a hierarchical relationship with user C, then two edges can be established: one between user A and user B, and another between user A and user C. Next, the semantic representation (implicit representation) of each user in the social network can be calculated. Then, the semantic similarity between the target user and other users can be calculated. The semantic representations of the other users with the highest similarity to the target user are selected as the social relationship description features of the target user, i.e., a social relationship profile.

[0066] After determining the user profile 311 corresponding to the scenario type, the profile retrieval module 310 can retrieve the target profile data by searching the user profile 311.

[0067] According to embodiments of this disclosure, target description features are obtained from object description features based on input information; and object description text is generated based on the target description features.

[0068] For example, user profile 311 may contain a large amount of user profile data. However, solving user problem 302 does not require all the data in user profile 311. Therefore, user profile 311 can be retrieved based on user problem 302 to obtain target profile information corresponding to user problem 302. The target profile information can be stored in a structured data format. To facilitate subsequent LLM processing, the retrieved target profile information can be described using natural language to obtain object description text, i.e., profile text 312.

[0069] On the other hand, after the user inputs user question 302, the document retrieval module 320 can perform a general search based on user question 302, such as retrieving documents or knowledge fragments related to user question 302 from the knowledge base to obtain the search text 321.

[0070] Next, personalized response information can be generated based on the portrait text 312 retrieved by the portrait retrieval module 310 and the retrieval text 321 retrieved by the document retrieval module 320.

[0071] According to embodiments of this disclosure, a generation mode is determined based on input information, object description text, and search text; and in response to the generation mode being a personalized generation mode, response information is generated based on the object description text and search text.

[0072] Personalized generation mode refers to generating response information that matches the characteristics of different users. In personalized generation mode, because user description features (user profiles) are combined, it is possible to generate response information that matches the characteristics of different users, achieving a personalized response effect.

[0073] In personalized generation tasks, failing to distinguish between the subjective and objective aspects of user question 302 and providing personalized answers to all questions can also have negative effects. For example, if user question 302 is "Which dynasty did historical figure A belong to?", using a personalized answer for such an objective question would actually affect the accuracy of the response.

[0074] In view of this, embodiments of the present disclosure introduce a personalization detection module 303 to determine whether a personalized response should be provided. If the determination result is yes, the response information is generated in a personalized generation mode. If the determination result is no, the response information is generated in a normal generation mode.

[0075] In one example, text can be categorized based on input information, object description text, and search text, with the categorization result indicating whether a personalized response should be provided.

[0076] In another example, LLM can be used to determine personalized answers. For instance, input information, object description text, and search text are fed into the LLM, with prompts indicating whether the LLM output is suitable for a personalized answer. To improve the accuracy of the LLM's determination, samples can be built to fine-tune the LLM, enabling it to accurately determine whether the current question is suitable for a personalized answer.

[0077] According to embodiments of this disclosure, based on input information, object description text and search text are matched to obtain matching results, whereby the matching results characterize whether the object description text enhances the personalization of the response information; and based on the matching results, a generation mode is determined. The generation modes include a personalized generation mode and a normal generation mode.

[0078] For subjective questions from users, the user's object description text can generally enhance the personalization of the response. However, for objective questions, the object description text cannot enhance the personalization of the response. Therefore, the generation pattern can be determined by judging whether the object description text enhances the personalization of the response.

[0079] For example, if the generated response information based on the object description text and the search text is inconsistent with the response information generated based solely on the search text, and the response information generated based on the object description text and the search text is more in line with the user's characteristics, then it can be determined that the object description text has personalized the answer to user question 302, that is, the object description text has enhanced the personalization of the response information 341, and the generation mode can be determined to be a personalized generation mode.

[0080] If the generated response information based on the object description text and the search text is consistent with the response information generated based solely on the search text, it can be determined that the object description text has not personalized the answer to user question 302. That is, the object description text has not enhanced the personalization of the response information 341, and the generation mode can be determined to be the normal generation mode.

[0081] LLM can be used to determine whether the object description text enhances the personalization of the response information. For example, a prompt can be written that merges user question 302 with search text 321 to obtain merged text. This merged text is then matched with the object description text, and the LLM is instructed to determine, based on the matching results, whether the response information 341 generated with the object description text is consistent with the response information 341 generated solely based on the merged text, and whether the response information is more in line with user characteristics. If yes, then "yes" can be output; otherwise, "no" can be output.

[0082] In personalized generation mode, user question 302, profile text 312, and search text 321 can be input into generation module 340, which can then generate personalized response information 341. In normal generation mode, user question 302 and search text 321 can be input into generation module 340, which can then generate standard response information 341.

[0083] According to embodiments of this disclosure, a personalized retrieval module determines whether a personalized response is required. If a personalized response is required, personalized response information is generated based on the retrieved profile text and the retrieval text. If a personalized response is not required, ordinary response information is generated based on the retrieval text. This approach improves both the personalization effect and the accuracy of the response information.

[0084] Figure 4 This is a system schematic diagram of a response method according to another embodiment of the present disclosure.

[0085] like Figure 4 As shown, the system in this embodiment includes a scene discrimination module 410, a document retrieval module 420, a portrait retrieval module 430, a personalization detection module 440, a personalization generation module 450, and a normal generation module 460.

[0086] After the user inputs a question, the document retrieval module 420 can perform a document retrieval based on the user's question and obtain the retrieval text 421.

[0087] In addition, after the user inputs a question, the scenario discrimination module 410 can determine whether the user's question belongs to a preset scenario, and if it is determined to belong to a preset scenario, it can determine which specific scenario it belongs to.

[0088] According to embodiments of this disclosure, the preset scenarios include task scenarios and transaction scenarios; the response method of embodiments of this disclosure further includes: establishing at least one of the following: a correspondence between the task scenarios and the historical operation description features; a correspondence between the transaction scenarios and the basic description features, wherein the correspondence is used to determine the object description features of the target object corresponding to the scenario type.

[0089] Task scenarios can include office scenarios, such as finding documents and attending meetings. Transaction scenarios can include personnel scenarios, financial scenarios, and system scenarios. Personnel scenarios include leave requests and policy changes, while financial scenarios include expense reimbursements and invoicing. System scenarios include departments and assets. All of the above task and transaction scenarios are preset scenarios.

[0090] For each of the above preset scenarios, a correspondence between the scenario and multiple user profiles can be pre-established based on the user's needs in that scenario.

[0091] For example, in task-based scenarios, user questions are often related to the user's work content, role information, and social relationships. Therefore, a correspondence can be established between task scenarios and at least one of the work profile, role profile, and social relationship profile. For instance, in a document-finding scenario, a user might ask "find the document in project Y" or "find the document sent by Z." Therefore, a relationship can be established between the document-finding scenario and the work profile, social relationship profile, and role profile.

[0092] For example, in transactional scenarios, user questions are often related to the user's basic information, role information, and hierarchical information. Therefore, a correspondence can be established between transactional scenarios and at least one of the basic user profile, role profile, and social relationship profile. For instance, in a departmental scenario, a user might ask, "How do I get to Department X? Which floor is it on?" A correspondence can be established between the departmental scenario and the basic user profile.

[0093] The correspondence between the above-mentioned scenario types and user profiles can be established according to the actual needs of users in each scenario, and the embodiments disclosed herein do not limit this.

[0094] After the scene discrimination module 410 identifies the specific scene and determines the corresponding user profile based on the specific scene, the profile retrieval module 430 can retrieve the user profile corresponding to the scene type to obtain the target profile data. The target profile data can be described in natural language to obtain the profile text 431.

[0095] Next, the personalization detection module 450 can determine whether personalized generation is needed based on the search text 421 and the profile text 431. If personalized generation is needed, the personalization generation module 450 generates personalized response information based on the user question, search text 421, and profile text 431. If personalized generation is not needed, the normal generation module 460 generates normal response information based on the user question and search text 431.

[0096] It should be noted that the scenario identification module 410 may determine that the user's question does not belong to any of the aforementioned preset scenarios. In this case, the scenario identification module 410 can output that the scenario type is empty. If the scenario is empty, the profile retrieval can be skipped, and the normal generation module 460 can be executed directly, that is, normal response information can be generated based on the user's question and the retrieval text 421.

[0097] According to embodiments of this disclosure, in response to receiving a user question, a scenario category is determined. If the user question belongs to a preset scenario type, a user profile is retrieved. If the user question does not belong to a preset scenario type, response information can be directly generated based on the retrieved text without retrieving a user profile, which can reduce computational load and improve response efficiency.

[0098] According to embodiments of this disclosure, a response device is also provided.

[0099] Figure 5 This is a block diagram of a response device according to an embodiment of the present disclosure.

[0100] like Figure 5 As shown, the response device 500 includes a scene type determination module 510, an object description feature determination module 520, an object description text determination module 530, and a first response module 540.

[0101] The scene type determination module 510 is used to determine the scene type based on the input information received from the target object.

[0102] The object description feature determination module 520 is used to determine the object description features of the target object corresponding to the scene type in response to determining that the scene type belongs to the preset scene type.

[0103] The object description text determination module 530 is used to generate object description text based on object description features.

[0104] The first response module 540 is used to generate response information based on the object description text and the search text, wherein the search text is obtained by searching based on the input information.

[0105] The first response module 540 includes a generation mode determination unit and a first response information determination unit.

[0106] The generation pattern determination unit is used to determine the generation pattern based on the input information, object description text, and retrieval text.

[0107] The first response information determination unit is used to generate response information based on the object description text and the search text in response to determining that the generation mode is a personalized generation mode.

[0108] The generation pattern determination unit includes a matching subunit and a generation pattern determination subunit.

[0109] The matching subunit is used to match the object description text with the search text based on the input information to obtain the matching result. The matching result characterizes whether the object description text enhances the personalization of the response information.

[0110] The generation mode determination subunit is used to determine the generation mode based on the matching results. The generation mode includes one of the personalized generation mode and the normal generation mode.

[0111] The first response module 540 also includes a second response information determination unit.

[0112] The second response information determination unit is used to generate response information based on the retrieved text in response to determining that the generation mode is normal generation mode.

[0113] The response device 500 also includes a second response module.

[0114] The second response module is used to generate response information based on the search text when the scene type is determined not to belong to the preset scene type.

[0115] The object description text determination module 530 includes a retrieval unit and an object description text generation unit.

[0116] The retrieval unit is used to obtain target description features from object description features based on input information.

[0117] The object description text generation unit is used to generate object description text based on target description features.

[0118] According to embodiments of this disclosure, the object description features include basic description features, historical operation description features, role description features, and social relationship description features. The response device 500 further includes a first feature determination module, a second feature determination module, and a third feature determination module.

[0119] The first feature determination module is used to determine basic descriptive features based on the basic information of the target object.

[0120] The second feature determination module is used to determine historical operation description features and role description features from the historical operation data of the target object.

[0121] The third feature determination module is used to determine social relationship description features based on the target object's historical social information.

[0122] The preset scenarios include task scenarios and transaction scenarios. The response device 500 also includes a correspondence establishment module.

[0123] The correspondence establishment module is used to establish at least one of the following: the correspondence between task scenarios and historical operation description features; the correspondence between transaction scenarios and basic description features; wherein, the correspondence is used to determine the object description features of the target object corresponding to the scenario type.

[0124] According to embodiments of this disclosure, this disclosure also provides an electronic device, a readable storage medium, and a computer program product.

[0125] Figure 6 A schematic block diagram of an example electronic device 600 that can be used to implement embodiments of the present disclosure is shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device may also represent various forms of mobile devices, such as personal digital processors, cellular phones, smartphones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely illustrative and are not intended to limit the implementation of the present disclosure described and / or claimed herein.

[0126] like Figure 6 As shown, device 600 includes a computing unit 601, which can perform various appropriate actions and processes based on a computer program stored in read-only memory (ROM) 602 or a computer program loaded from storage unit 608 into random access memory (RAM) 603. RAM 603 may also store various programs and data required for the operation of device 600. The computing unit 601, ROM 602, and RAM 603 are interconnected via bus 604. Input / output (I / O) interface 605 is also connected to bus 604.

[0127] Multiple components in device 600 are connected to I / O interface 605, including: input unit 606, such as keyboard, mouse, etc.; output unit 607, such as various types of monitors, speakers, etc.; storage unit 608, such as disk, optical disk, etc.; and communication unit 609, such as network card, modem, wireless transceiver, etc. Communication unit 609 allows device 600 to exchange information / data with other devices through computer networks such as the Internet and / or various telecommunications networks.

[0128] The computing unit 601 can be a variety of general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of the computing unit 601 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various special-purpose artificial intelligence (AI) computing chips, various computing units running machine learning model algorithms, a digital signal processor (DSP), and any suitable processor, controller, microcontroller, etc. The computing unit 601 performs the various methods and processes described above, such as the response method. For example, in some embodiments, the response method may be implemented as a computer software program tangibly contained in a machine-readable medium, such as storage unit 608. In some embodiments, part or all of the computer program may be loaded and / or installed on device 600 via ROM 602 and / or communication unit 609. When the computer program is loaded into RAM 603 and executed by the computing unit 601, one or more steps of the response method described above may be performed. Alternatively, in other embodiments, the computing unit 601 may be configured to perform the response method by any other suitable means (e.g., by means of firmware).

[0129] Various embodiments of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field-programmable gate arrays (FPGAs), application-specific integrated circuits (ASICs), application-specific standard products (ASSPs), systems-on-a-chip (SoCs), complex programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various embodiments may include implementations in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which may be a dedicated or general-purpose programmable processor, capable of receiving data and instructions from a storage system, at least one input device, and at least one output device, and transmitting data and instructions to the storage system, the at least one input device, and the at least one output device.

[0130] The program code used to implement the methods of this disclosure may be written in any combination of one or more programming languages. This program code may be provided to a processor or controller of a general-purpose computer, special-purpose computer, or other programmable data processing apparatus, such that when executed by the processor or controller, the program code causes the functions / operations specified in the flowcharts and / or block diagrams to be implemented. The program code may be executed entirely on a machine, partially on a machine, as a standalone software package partially on a machine and partially on a remote machine, or entirely on a remote machine or server.

[0131] 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.

[0132] To provide interaction with a user, the systems and techniques described herein can be implemented on a computer having: a display device for displaying information to the user (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor); and a keyboard and pointing device (e.g., a mouse or trackball) through which the user provides input to the computer. Other types of devices can also be used to provide interaction with the user; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including sound input, voice input, or tactile input).

[0133] The systems and technologies described herein can be implemented in computing systems that include backend components (e.g., as a data server), or computing systems that include middleware components (e.g., an application server), or computing systems that include frontend components (e.g., a user computer with a graphical user interface or web browser through which a user can interact with implementations of the systems and technologies described herein), or any combination of such backend, middleware, or frontend components. The components of the system can be interconnected via digital data communication of any form or medium (e.g., a communication network). Examples of communication networks include local area networks (LANs), wide area networks (WANs), and the Internet.

[0134] Computer systems can include clients and servers. Clients and servers are generally located far apart and typically interact through communication networks. Client-server relationships are created by computer programs running on the respective computers and having a client-server relationship with each other.

[0135] It should be understood that the various forms of processes shown above can be used to rearrange, add, or delete steps. For example, the steps described in this disclosure can be executed in parallel, sequentially, or in different orders, as long as the desired result of the technical solution disclosed in this disclosure can be achieved, and this is not limited herein.

[0136] The specific embodiments described above do not constitute a limitation on the scope of protection of this disclosure. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this disclosure should be included within the scope of protection of this disclosure.

Claims

1. A response method, comprising: In response to receiving input information from the target object, the scene type is determined based on the input information; In response to determining that the scenario type belongs to a preset scenario type under the enterprise office scenario, based on the correspondence between the preset scenario type and multiple dimensions of object description features, the target object description features corresponding to the scenario type are determined from the multiple dimensions of object description features of the target object. The multiple dimensions of object description features include basic description features representing the basic information of the object, historical operation description features representing the work content of the object, role description features representing the role and work direction of the object, and social relationship description features representing the social relationship of the object. The structured target object description features are converted into natural language description text to generate object description text; The input text and the retrieved text are fused to obtain the fused text. The fused text and the object description text are input into a large language model. The large language model is used to perform semantic matching on the object description text and the fused text. Based on the matching results, it is determined whether the response information generated in combination with the object description text enhances the degree of personalization. In response to the judgment result indicating enhanced personalization, the generation mode is determined to be a personalized generation mode, and response information is generated based on the object description text and the search text; In response to the judgment result that no personalization has been enhanced, the generation mode is determined to be the normal generation mode, and the response information is generated based on the search text, wherein the search text is obtained by searching the knowledge base based on the input information.

2. The method according to claim 1, further comprising: In response to determining that the scene type does not belong to the preset scene type, the response information is generated based on the search text.

3. The method according to claim 1, wherein, The step of generating object description text based on the target object description features includes: Based on the input information, target features are obtained from the target object description features; and The object description text is generated based on the target features.

4. The method according to claim 1, further comprising: Based on the basic information of the target object, the basic descriptive features are determined; The historical operation description features and role description features are determined from the historical operation data of the target object; as well as The social relationship description features are determined based on the target object's historical social information.

5. The method according to claim 4, wherein, The preset scenarios include task scenarios and transaction scenarios; the method further includes: Establish at least one of the following: The correspondence between the task scenario and the historical operation description features; The correspondence between the transaction scenario and the basic descriptive features; The correspondence is used to determine the object description features of the target object corresponding to the scene type.

6. A response device, comprising: The scene type determination module is used to determine the scene type based on the input information received from the target object in response to receiving the input information. The object description feature determination module is used to, in response to determining that the scene type belongs to a preset scene type under the enterprise office scene, determine the target object description feature corresponding to the scene type from the multiple dimensions of the object description feature of the target object according to the correspondence between the preset scene type and multiple dimensions of object description features. The multiple dimensions of object description features include basic description features representing the basic information of the object, historical operation description features representing the work content of the object, role description features representing the role and work direction of the object, and social relationship description features representing the social relationship of the object. The object description text determination module is used to convert the target object description features into natural language description text to generate object description text. The generation pattern judgment module is used to merge the input text and the retrieved text to obtain the merged text, input the merged text and the object description text into the large language model, use the large language model to match the object description text and the merged text, and judge whether the generated response information with the object description text is enhanced in terms of personalization based on the matching result. The first response module is used to respond to the judgment result that the degree of personalization has been enhanced, determine that the generation mode is a personalized generation mode, and generate response information based on the object description text and the search text. The second response module is used to determine that the generation mode is a normal generation mode in response to the judgment result that there is no enhancement of personalization, and to generate the response information based on the search text, wherein the search text is obtained by searching the knowledge base based on the input information.

7. The apparatus according to claim 6, further comprising: The third response module is used to generate the response information based on the search text in response to determining that the scene type does not belong to the preset scene type.

8. The apparatus according to claim 6, wherein the object description text determination module comprises: The retrieval unit is used to obtain target features from the target object description features based on the input information; as well as An object description text generation unit is used to generate the object description text based on the target features.

9. The apparatus according to claim 6, further comprising: The first feature determination module is used to determine the basic descriptive features based on the basic information of the target object; The second feature determination module is used to determine the historical operation description features and role description features from the historical operation data of the target object; as well as The third feature determination module is used to determine the social relationship description features based on the target object's historical social information.

10. The apparatus according to claim 9, wherein the preset scenario includes a task scenario and a transaction scenario; the apparatus further includes: The mapping relationship establishment module is used to establish at least one of the following: The correspondence between the task scenario and the historical operation description features; The correspondence between the transaction scenario and the basic descriptive features; The correspondence is used to determine the object description features of the target object corresponding to the scene type.

11. An electronic device, comprising: At least one processor; as well as A memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor to enable the at least one processor to perform the method of any one of claims 1 to 5.

12. A non-transitory computer-readable storage medium storing computer instructions, wherein, The computer instructions are used to cause the computer to perform the method according to any one of claims 1 to 5.

13. A computer program product comprising a computer program stored on at least one of a readable storage medium and an electronic device, the computer program implementing the method according to any one of claims 1 to 5 when executed by a processor.

Images