Information interaction method and device, electronic equipment, storage medium and program product

Abstract

The embodiment of the invention provides an information interaction method and device, electronic equipment, a storage medium and a program product. The method comprises the steps of obtaining first inquiry information, converting the first inquiry information into second inquiry information, and performing retrieval in a target database according to the first inquiry information and the second inquiry information so as to obtain multiple pieces of first recall information; determining a problem solving state of the first inquiry information and retrieval feedback information in the problem solving state according to the first inquiry information, the plurality of pieces of first recall information and a first intelligent agent; and in response to the event of which the problem solving state is that the problem is solved, determining problem reply information of the first inquiry information according to the retrieval feedback information, and displaying the problem reply information. According to the technical scheme, through dual optimization of preposed semantic enhancement and postposed state verification, the response speed of a simple problem is increased in magnitude while the processing capacity of the complex problem is kept.

Description

TECHNICAL FIELD

[0001] Embodiments of the present disclosure relate to the technical field of computer processing, and particularly relate to an information interaction method and device, electronic equipment, storage medium and program product. BACKGROUND

[0002] With the continuous development of computer technology, as an emerging information interaction method, intelligent agents have gradually become one of the core technologies in various application scenarios. In the service field, intelligent agents are widely used in intelligent customer service, knowledge management, and personalized recommendation, etc., and can provide efficient and accurate services based on user needs. Especially in complex problem processing, intelligent agents significantly improve information processing efficiency and user experience through multi-task collaboration and real-time decision-making capabilities.

[0003] However, in the related art, the general reply feedback mechanism commonly used for various types of interaction content has obvious limitations. For example, although the solution based on the multi-agent architecture has enhanced comprehensive processing capabilities, its mechanism of relying on multiple model calls results in a long response time, especially when dealing with simple problems, which still consumes a large amount of computing resources, seriously affecting user experience. SUMMARY

[0004] Embodiments of the present disclosure provide an information interaction method, device, electronic equipment, storage medium and program product to solve the technical problem of slow response speed and high resource consumption of intelligent agent framework even when solving simple problems in the related art.

[0005] In a first aspect, embodiments of the present disclosure provide an information interaction method, which comprises:

[0006] obtaining first inquiry information, converting the first inquiry information into second inquiry information, and performing retrieval in a target database according to the first inquiry information and the second inquiry information to obtain a plurality of first recall information;

[0007] determining a problem solving state of the first inquiry information and retrieval feedback information in the problem solving state according to the first inquiry information, the plurality of first recall information and a first intelligent agent;

[0008] in response to an event that the problem solving state is that the problem has been solved, determining problem reply information of the first inquiry information according to the retrieval feedback information, and displaying the problem reply information.

[0009] In a second aspect, embodiments of the present disclosure also provide an information interaction device, which comprises:

[0010] The inquiry information acquisition module is configured to acquire first inquiry information, convert the first inquiry information into second inquiry information, and perform a search in a target database according to the first inquiry information and the second inquiry information to obtain a plurality of first recall information.

[0011] The search information feedback module is configured to determine a problem solving state of the first inquiry information and search feedback information in the problem solving state according to the first inquiry information, the plurality of first recall information, and a first intelligent agent.

[0012] The reply information display module is configured to, in response to the problem solving state being an event that the problem has been solved, determine problem reply information of the first inquiry information according to the search feedback information, and display the problem reply information.

[0013] In a third aspect, the embodiments of the present disclosure further provide an electronic device, which comprises:

[0014] one or more processors;

[0015] a storage device configured to store one or more programs,

[0016] When the one or more programs are executed by the one or more processors, the one or more processors implement the information interaction method according to any of the embodiments of the present disclosure.

[0017] In a fourth aspect, the embodiments of the present disclosure further provide a storage medium containing computer executable instructions, which, when executed by a computer processor, are configured to perform the information interaction method according to any of the embodiments of the present disclosure.

[0018] In a fifth aspect, the embodiments of the present disclosure further provide a computer program product comprising a computer program, which, when executed by a processor, implements the information interaction method according to any of the embodiments of the present disclosure.

[0019] The technical scheme of the embodiments of the present disclosure is that first inquiry information is obtained, the first inquiry information is converted into second inquiry information, the first inquiry information and the second inquiry information are used for searching in a target database to obtain a plurality of first recall information, a double search triggering mechanism is constructed by converting the original inquiry (the first inquiry information) into the search instruction (the second inquiry information) with more complete semantics, the target database can be matched from two dimensions of expression form and potential demand at the same time, and the hit rate of high correlation information (the first recall information) is significantly improved. The problem solving state of the first inquiry information and the search feedback information in the problem solving state are determined according to the first inquiry information, the plurality of first recall information and the first intelligent agent, which is different from the consumptive thinking of the traditional architecture that needs multiple rounds of model self-questioning and self-answering. The present solution forms a closed-loop decision through the instant state evaluation (the problem solving state) of the first intelligent agent on the recall information. This single decision mechanism makes the response speed of simple problems close to the level of real-time interaction. According to the search feedback information, the problem reply information of the first inquiry information is determined and the problem reply information is displayed in response to the event that the problem solving state is that the problem is solved. In the present technical solution, the first intelligent agent compares the coverage of the original inquiry and the search result, and when it is judged that the recall information already contains a complete solution, the subsequent processing process is terminated immediately, and the reply information is generated directly. The present solution essentially reconstructs the value transmission chain of the intelligent agent system: the "model thinking time" in the traditional architecture is converted into "effective information matching time", and through the double optimization of front semantic enhancement and back state verification, the processing speed of simple problems is improved by an order of magnitude while the processing ability of complex problems is maintained. For the scene with low customer service error rate and high response time requirement, this end-to-end processing process makes the user not need to perceive the multi-system cooperation process behind, meets the dialogue expectation, and balances the accuracy and efficiency, thereby significantly improving the information interaction experience. BRIEF DESCRIPTION OF DRAWINGS

[0020] The above and other features, advantages, and aspects of the embodiments of the present disclosure will become more apparent with reference to the following detailed description in conjunction with the accompanying drawings. 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 sizes of the components and elements are not necessarily drawn to scale.

[0021] Figure 1 A flowchart of an information interaction method provided by an embodiment of the present disclosure is shown in the figure;

[0022] Figure 2 A flowchart of another information interaction method provided by an embodiment of the present disclosure is shown in the figure;

[0023] Figure 3 A flowchart of another information interaction method provided by an embodiment of the present disclosure is shown in the figure;

[0024] Figure 4 A flowchart of another information interaction method provided by an embodiment of the present disclosure;

[0025] Figure 5 A structural diagram of an information interaction apparatus provided by an embodiment of the present disclosure;

[0026] Figure 6 A structural diagram of an electronic device for implementing an embodiment of the present disclosure provided by an embodiment of the present disclosure. DETAILED DESCRIPTION

[0027] Embodiments of the present disclosure will be described in more detail by referring to the drawings. Although certain embodiments of the present disclosure are shown in the drawings, it is understood that the present disclosure can be implemented in various forms, and should not be construed as being limited to the embodiments set forth herein, but rather the embodiments are provided to more thoroughly and completely understand the present disclosure. It is understood that the drawings and embodiments of the present disclosure are for exemplary purposes only, and are not intended to limit the scope of protection of the present disclosure.

[0028] It should be understood that each of the steps recited in the method embodiments of the present disclosure can be executed in different orders, and / or in parallel. In addition, the method embodiments can include additional steps and / or omit the execution of the steps shown. The scope of the present disclosure is not limited in this respect.

[0029] The term "comprising" and variations thereof as used in the present disclosure are open-ended, that is, "including but not limited to". The term "based on" is "based, at least in part, 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". Related definitions of other terms will be given in the description below.

[0030] It should be noted that the concepts of "first", "second", etc. mentioned in the present disclosure are only used to distinguish different devices, modules or units, and are not intended to limit the order or interdependence of the functions performed by these devices, modules or units.

[0031] It should be noted that the modification of "one" or "multiple" mentioned in the present disclosure is illustrative and not limiting, and those skilled in the art should understand that, unless otherwise explicitly indicated in the context, it should be understood as "one or more".

[0032] The names of the messages or information exchanged between the plurality of devices in the embodiments of the present disclosure are only for illustrative purposes, and are not intended to limit the scope of the messages or information.

[0033] It can be understood that, before using the technical solutions disclosed in the embodiments of the present disclosure, the type, use range, use scenario, etc. of the personal information involved in the present disclosure should be informed to the user and the authorization of the user should be obtained through appropriate means according to relevant laws and regulations.

[0034] For example, in response to receiving the active request of the user, the prompt information is sent to the user to explicitly prompt the user that the operation requested to be performed will require obtaining and using the personal information of the user. Thus, the user can autonomously select whether to provide the personal information to the software or hardware such as the electronic device, application program, server or storage medium, etc. performing the operation of the technical solutions of the present disclosure according to the prompt information.

[0035] As an optional but non-limiting implementation manner, in response to receiving the active request of the user, the prompt information can be sent to the user in the form of a pop-up window, and the prompt information can be presented in the form of text in the pop-up window. In addition, the pop-up window can also carry selection controls for the user to select "agree" or "disagree" to provide the personal information to the electronic device.

[0036] It can be understood that the above notification and obtaining of the authorization of the user are only illustrative, and do not limit the implementation manners of the present disclosure, and other manners meeting the relevant laws and regulations can also be applied to the implementation manners of the present disclosure.

[0037] It can be understood that the data (including but not limited to the data itself, the acquisition or use of the data) involved in the present technical solutions should comply with the requirements of the relevant laws and regulations and the relevant provisions.

[0038] Figure 1 A flowchart of an information interaction method provided by the embodiments of the present disclosure is shown. The embodiments of the present disclosure are applicable to an information interaction scene in which simple questions and complex questions are interwoven. The method can be executed by an information interaction device, which can be implemented in the form of software and / or hardware. Optionally, the information interaction device is implemented by an electronic device, which can be a mobile terminal, a PC terminal or a server, etc. Figure 1 As shown in the figure, the method of the present embodiment can specifically include:

[0039] S110, obtaining first inquiry information, converting the first inquiry information into second inquiry information, and performing retrieval in a target database according to the first inquiry information and the second inquiry information to obtain a plurality of first recall information.

[0040] In the embodiments of the present disclosure, the first query information can be understood as the original question expression in the information interaction process, for example, the query information input by the user. The first query information is usually natural language text, which can contain one or more problems such as colloquial expression, information missing or ambiguity. The second query information can be understood as the standardized query information after semantic enhancement, which can contain explicit expression of the implicit semantics of the first query information. In a popular way, the second query information can be a more complete, clear and semantically enhanced expression of the first query information. The target database can be understood as a storage space for storing retrievable data. The target database can specifically be a collection of heterogeneous data sources. According to the data form, the target database can include structured data (such as tables, etc.), unstructured data (such as product manuals, etc.), and program data interfaces (such as order systems), etc. According to the data type or data source, the target database can store document data and / or question and answer data, etc. The first recall information can be understood as a preliminary retrieval result set, which includes retrieval results in the target database according to the first query information and retrieval results in the target database according to the second query information.

[0041] As an optional implementation manner of the embodiments of the present disclosure, obtaining the first query information can include but is not limited to at least one of the following operations: obtaining the first query information input through a preset information input control; obtaining the first query information photographed through a preset image shooting control; obtaining the first query information uploaded through a preset image uploading control; receiving the first query information transmitted by a target data interface (including a program data interface); obtaining the first query information recorded through a preset audio recording control; obtaining the first query information uploaded through a preset audio uploading control; extracting the first query information from a multimedia resource, etc.

[0042] As an optional implementation of the embodiments of the present disclosure, the converting the first query information into second query information comprises: inputting the first query information into a third intelligent agent to obtain the second query information. The third intelligent agent can be understood as an intelligent agent with semantic understanding and reconstruction capabilities. The third intelligent agent can be used for deep analysis and semantic reconstruction of the input information. In simple terms, the third intelligent agent can convert the first query information into second query information that is more specific, more semantically clear, and more standardized in structure. The third intelligent agent can perform one or more operations such as converting to explicit semantic expression, supplementing context, and adjusting expression method (such as replacing colloquial vocabulary with professional terms in the database, etc.) to convert the first query information into the second query information. This scheme effectively solves the "answering the wrong question" problem caused by user expression ambiguity in general question and answer systems through deep semantic processing of special intelligent agents, and is particularly suitable for high-precision scenarios. Moreover, the third intelligent agent can automatically, quickly, accurately and comprehensively convert the first query information into the second query information, improving the intelligent level of information interaction and data processing efficiency, and improving the information interaction experience.

[0043] In some embodiments of the present disclosure, the searching in the target database according to the first query information and the second query information can be specifically that the original question (first query information) and the reconstructed question (second query information) are used as retrieval keywords, and a global search is performed in the target database by using a literal search (preset word matching) and / or vector search as a search method to quickly recall associated content of the first query information and the second query information. The vector search can be understood as determining information relevance by semantic similarity. It should be noted that the associated content of the first query information and the second query information can be searched in parallel. Such a search method combines the original question and the reconstructed question, which can guarantee the recall rate, improve the precision rate, and avoid the limitations of a single search strategy. Moreover, the content matching the first query information can be quickly recalled.

[0044] In some embodiments of the present disclosure, after obtaining a plurality of first recall information, the plurality of first recall information recalled can be weighted and sorted according to one or more information such as field matching degree (complete matching > partial matching), data freshness (preferentially querying the data version closest to the time), and source reliability (for example, documents published by the owner of the query subject > user-generated content). In order to generate more accurate and reliable question and answer information subsequently.

[0045] In some embodiments of the present disclosure, the first query information can be pre-processed before being searched according to the first query information, wherein the pre-processing can include text cleaning and / or context association, etc. The text cleaning can include, but is not limited to, at least one of filtering preset characters and correcting misspelled words (e.g., "maintenance period" -> "warranty period"). The context association can be to extract the dialogue entity (subject object associated with the question) in the previous round of dialogue as supplementary information in the continuous dialogue (multi-round dialogue) scenario.

[0046] In some embodiments of the present disclosure, an agent refers to a computing entity with autonomous perception, decision-making and execution capabilities. The agent can run without human intervention and can perceive environmental changes in real time and respond. The agent can include a perception layer, a cognitive layer and a decision layer. The perception layer can be used to process multi-modal input, such as processing one or more types of data such as voice (e.g., telephone customer service noise reduction), image (medical image analysis) and sensor data (warehouse robot obstacle avoidance). The cognitive layer can analyze user ambiguous expressions through a language model. The decision layer can implement decision-making by combining one or more architectures such as rule engines, probabilistic models and scenario constraints.

[0047] As described above, the target database can include various types of searchable data. In other words, the data sources of the searchable data in the target database are complex, often with multiple granularities, structured, unstructured and mixed characteristics. Therefore, in some embodiments of the present disclosure, in order to maximize the adaptation of the recall strategy to the data structure, the source data in the target database can be reprocessed.

[0048] As an optional technical solution of some embodiments of the present disclosure, when the searchable data in the target database includes document data, before searching in the target database according to the first query information and the second query information, the document data content can be summarized according to a document processing model to obtain document description information of the document data, and the document data and the document description information corresponding thereto are stored in the target database.

[0049] The document data can be understood as data recorded in a document. The document data can be understood as an original information carrier that needs to be processed. The document processing model can be understood as a pre-trained machine learning model that can be used for document processing (analysis and understanding). The original document can be analyzed and summarized by the document processing model. For example, the document processing model can include at least one of a natural language processing model, a multi-modal model, and an abstract generation model. The natural language processing model can be used for semantic understanding and text feature extraction. The multi-modal model can be used to process a composite document containing one content or multiple mixed contents such as images, tables, texts, and codes. The abstract generation model can be used to generate condensed content by an extractive or generative method.

[0050] Specifically, the document data can be pre-processed (such as format conversion and / or text extraction) first, and then the natural language processing technology is used to extract key entities, topics and core semantics to generate structured description information (such as abstracts, keywords, topic labels, etc.). The document data and the document description information are associated by identification. The document data and the document description information can adopt a hybrid storage strategy, for example, the original document is stored in object storage, and the document description information is stored in a database (such as a vector database) supporting efficient retrieval. A multi-level index (full-text index, semantic vector index, etc.) can also be established to optimize query performance. In addition, the quality evaluation module can also be used to ensure the accuracy of the description information and support continuous iterative optimization. By using the technical solution, the structured description information of the document data can be obtained, the information query of different granularities can be supported, the information query efficiency can be improved, and a standardized data basis for subsequent information query is provided.

[0051] As another optional technical solution of the embodiment of the present disclosure, the searchable data in the target database includes document data; before the searching in the target database according to the first inquiry information and the second inquiry information, the document data can be divided to obtain a plurality of initial document segments, a target document segment is determined according to the initial document segments, and the target document segment is stored in the target database. By using the technical solution, the document data is fragmented, in the subsequent processing process, the parallel query of information can be supported, and by using the pre-segmented segment-level index, the key information area can be located without scanning the full text during the searching. The information query efficiency is improved. Compared with the overall compressed document, the segment-level compression can select the best compression algorithm according to the content type (text / table / code). The loose coupling characteristics between the document segments support dynamic recombination, lay a foundation for the subsequent explainable search and interactive analysis, support accurate result focusing, and effectively improve the response speed, resource utilization and functional expansibility.

[0052] The content in the plurality of initial document segments can be pieced together to form the entire content of the document data, or can be part of the content of the document data. That is, the plurality of initial document segments can be obtained by dividing at least part of the content in the document data. At least part of the content in the plurality of initial document segments is different. The advantage of such an arrangement is that more content in the document data can be embodied through the plurality of initial document segments.

[0053] In some embodiments of the present disclosure, the dividing of the document data to obtain the plurality of initial document segments can specifically include: dividing the document data into the plurality of initial document segments according to a data format of content data included in the document data. The data format can include, but is not limited to, at least one of an image, a table, a link, a preset word, and a text, etc. By using the technical solution, data in different data formats can be classified and stored to support the retrieval of data in different formats. The slicing based on the format characteristics can accurately align the document logical structure during index construction, which helps to improve the accuracy of retrieval and the success rate of data matching.

[0054] As an optional implementation, the document data can be divided according to the document type. For example, a text document can be divided according to paragraph markers, title levels, or fixed character windows; a table document can be divided according to worksheets or logical row groups; a rich media document can be divided according to the boundaries of text and image regions identified by layout analysis; a code document can be divided according to syntax tree nodes (including functions, class definitions, etc.); and the like.

[0055] As another optional implementation, a pre-trained machine learning model can be used to automatically detect the format type of a document, and then a corresponding format slicing rule template can be loaded to divide the document data into the plurality of initial document segments according to the data format of the content data included in the document data.

[0056] In some other embodiments of the present disclosure, the document data can also be input into a document segmentation model to divide the document data into the plurality of initial document segments based on document semantics. The document segmentation model can be obtained by training a pre-established machine learning model based on sample documents and their corresponding expected document segments. By using the technical solution, the document can be quickly and automatically segmented, and the plurality of initial document segments obtained by the division can maintain their original semantics in the document as much as possible, thereby ensuring the accuracy of subsequent information retrieval.

[0057] In some embodiments of the present disclosure, the determining the target document segment according to the initial document segment can specifically include: obtaining meta information of the document data, and adding the meta information to the initial document segment to obtain the target document segment. With this technical solution, the meta information can be added to the initial document segment, the discrete content segment can be converted into an intelligent data unit with context awareness ability through the meta information, a multi-dimensional index system (including time and space dimensions, semantic dimensions and relationship dimensions) can be constructed, and the retrieval system can realize stereoscopic query. The dynamically associated meta information can give the segment self-explanation ability, and improve the semantic matching degree of information. Moreover, the intelligent routing mechanism based on the meta data can optimize the storage access path, and through the analysis of the meta data, the information retrieval efficiency can be improved.

[0058] In some embodiments of the present disclosure, the meta information of the data refers to high-level information describing the attributes of the data, that is, "information about the data". The meta information of the data usually does not directly contain the specific content of the data, but annotates or explains the source, structure, context, relationship, update time, etc. of the data, which is similar to the "label" or "instruction manual" of the data. The meta information is usually used to represent one or more attribute information of the source attribute, the structure attribute, the semantic attribute and the use attribute of the data. The source attribute can include at least one attribute of the creator of the data, the publishing time, the version number and the reference file. The structure attribute can include at least one attribute of the organization form (such as chapter division, node relationship in knowledge graph) and logical framework (such as argument chain of premise and conclusion) of the data. The semantic attribute can include at least one attribute of the term definition, the domain label and the multilingual comparison label of the data. The use attribute can include at least one attribute of the use permission and the applicable scenario of the data.

[0059] In some embodiments of the present disclosure, the determining the target document segment according to the initial document segment can specifically include: summarizing the content of the initial document segment according to a document processing model to obtain segment description information of the initial document segment, and storing the initial document segment and the segment description information corresponding thereto into the target database. With this technical solution, first, the semantic fingerprint of the content segment can be described by the information, the unstructured text can be converted into a computable feature vector, and the semantic retrieval efficiency based on deep learning can be significantly improved; second, the index volume can be compressed by the lightweight characteristics of the description information through the dual storage mode; and finally, the dynamic mapping relationship between the segment description information and the document segment constitutes a traceable knowledge unit, and when the document is updated, only the description information of the affected segment needs to be recalculated, thereby reducing the time consumption of index maintenance. This technical solution upgrades the traditional keyword matching to the dual retrieval dimensions of "semantic anchor point + content entity", which not only retains the recall rate advantage of full-text retrieval, but also obtains the correlation reasoning ability of the knowledge graph level.

[0060] As another optional technical solution of the embodiments of the present disclosure, in the case that the searchable data in the target database includes question and answer data, before the searching in the target database according to the first inquiry information and the second inquiry information, the method further includes: generating question data and reply data according to the question and answer data, and storing the question data and its reply data in the target database in correspondence. With this technical solution, the precise mapping relationship between questions and replies can be established, the bidirectional index structure is constructed based on the association storage of question and answer pairs, the time complexity of question searching and answer backtracking is reduced, and the response speed of information searching is significantly improved. The coupling storage of question and answer data preserves the semantic integrity of the conversation context, and supports the coherence maintenance of multi-round conversations through implicit association fields. Moreover, the dynamic updating association mapping mechanism ensures that the question and answer pairs can be updated atomically when the database is maintained. This technical solution can convert discrete text data into knowledge units with logical association, which not only meets the engineering requirements of high-concurrency searching, but also provides a structured corpus basis for the subsequent training and information searching of conversational models.

[0061] Optionally, the generating of the question data and the reply data according to the question and answer data includes: pre-processing the question and answer data, and generating the question data and the reply data according to the pre-processed question and answer data. The pre-processing includes at least one of the following: elimination of invalid data, filtering of question and answer data that does not meet the expected quality, etc.

[0062] Optionally, the generating of the question data and the reply data according to the question and answer data includes: inputting the question and answer data into a semantic recognition model to obtain a plurality of question and answer pairs containing the question data and the reply data. The semantic recognition model can be obtained by training a pre-established machine learning model according to sample question and answer data and its corresponding expected question and answer pairs. The question data and the reply data in the question and answer pairs can be in a one-to-one, one-to-many or many-to-one relationship. The semantic recognition model can be used to divide the question data and the reply data from the question and answer data, and can further reprocess the question data and the reply data to generate structured question and answer data. This can provide more accurate data basis for subsequent information searching.

[0063] S120, determining the problem solving state of the first inquiry information and the retrieval feedback information in the problem solving state according to the first inquiry information, the plurality of first recall information and the first intelligent agent.

[0064] In the embodiments of the present disclosure, the first intelligent agent can be understood as a module for generating retrieval decisions. The problem solving state is used to represent the solving state or solving degree of the question in the first query information with respect to the first recall information. In a general way, the problem solving state is used to represent whether the first recall information is sufficient to solve the question in the first query information. The retrieval feedback information can be understood as the output after the problem solving state is determined, and can include reply-related information of the first query information or operation instructions for the next step.

[0065] Specifically, the problem solving state can be a binary classification result, and specifically can include that the question has been solved and that the question has not been solved. For example, in the case where the problem solving state includes that the question has been solved, the retrieval feedback information can include a plurality of the first recall information, a first recall result generated according to the plurality of the first recall information, or question reply data generated according to the plurality of the first recall information. In the case where the problem solving state includes that at least part of the question has not been solved, the retrieval feedback information can include at least one third query information, target retrieval data in the target database, and a retrieval manner for the target retrieval data. The third query information can be understood as a question in the first query information that has not been solved. That is, the third query information can be a question to be further retrieved generated by the first intelligent agent. The target retrieval data in the target database can be target retrieval data in the retrievable data that has not been retrieved, and / or the retrievable data that has not been retrieved in a preset manner. That is, the target retrieval data can be at least part of the retrievable data in the target database. The retrieval manner for the target retrieval data can be one or more of the preset retrieval manners. For example, the retrieval manner at least includes preset word retrieval and / or semantic retrieval. That is, preset word matching and / or similarity matching of semantic feature vectors, etc.

[0066] Specifically, determining the problem solving state of the first query information and the retrieval feedback information in the problem solving state according to the first query information, the plurality of the first recall information, and the first intelligent agent can include: inputting the first query information and the plurality of the first recall information into the first intelligent agent to analyze whether the first recall information completely covers the question demand of the first query information by the first intelligent agent, in the case where it is determined that the question has been solved, the retrieval feedback information for generating question reply information of the first query information can be output, and in the case where it is determined that at least part of the question has not been solved, accurate supplementary query instructions (subsequent supplementary retrieval strategies) can be generated for the unsolved part. Further, the retrieval range and the time limit requirement can be limited. The advantage of such a setting is that both timely and rapid response to simple questions and accurate and comprehensive reply to complex questions can be ensured.

[0067] In an embodiment of the present disclosure, specifically, the meta-information of the plurality of pieces of the first recall information can be acquired, and each piece of the first recall information can be spliced with its meta-information to obtain a plurality of pieces of second recall information; and the data description information and the retrieval execution information of the plurality of pieces of retrievable data can be acquired; then, the problem solving state of the first query information and the retrieval feedback information in the problem solving state can be determined according to the first query information, the plurality of pieces of second recall information, the data description information and the retrieval execution information of the plurality of pieces of retrievable data, and the first agent.

[0068] In an embodiment of the present disclosure, in the case where the target database includes retrievable data of multiple data types, the problem solving state of the first query information and the retrieval feedback information in the problem solving state can be determined according to the first query information, the plurality of pieces of first recall information, and the first agent, specifically, the data description information and the retrieval execution information of the plurality of pieces of retrievable data can be acquired, and the problem solving state of the first query information and the retrieval feedback information in the problem solving state can be determined according to the first query information, the plurality of pieces of first recall information, the data description information and the retrieval execution information of the plurality of pieces of retrievable data, and the first agent.

[0069] The retrievable data refers to data stored in the target database that can be used for retrieval. The data description information can be understood as description information of the retrievable data. For example, the data description information can include at least one of the description information of the applicable retrieval scene of the retrievable data, the applicable retrieval method in the retrieval scene, the source of the retrievable data, the update time, the reliability, and the format type. The retrieval execution information can be understood as information generated by the retrieval operation that has been performed in the process of retrieving the first query information. Specifically, the retrieval execution information can refer to information generated in the process of obtaining the first recall information and the subsequent retrieval operation performed. For example, the retrieval execution information can include the retrievable data that has been retrieved. Further, the retrieval execution information can include the retrieval method of the retrievable data that has been retrieved, etc.

[0070] In an embodiment of the present disclosure, the data description information of the retrievable data can be determined according to the data type corresponding to the retrievable data. In other words, the retrievable data of the same data type can use the same data description information. More specifically, the data type corresponding to the retrievable data can be determined according to the data source of the retrievable data. That is, the retrievable data of the same data source can use the same data description information. The advantage of such setting is that it can quickly determine whether the retrieval scale is reasonable and accurately determine the problem solving state of the first query information and the retrieval feedback information in the problem solving state.

[0071] The technical solution integrates data description information and retrieval execution information, and the first intelligent agent can perform weighted evaluation on the recall result, thereby improving the accuracy of the answer obtained by the retrieval. Based on the real-time feedback of the retrieval execution information, the optimal retrieval strategy can be automatically switched to reduce the response time of complex queries. Moreover, when a multi-source data conflict is detected, a fusion answer with a traceable mark can be generated in combination with the data description information, and an incremental retrieval can be triggered to supplement missing elements. This dual mechanism of data perception and strategy adaptation significantly improves the one-time problem solving rate, reduces the need for manual intervention, and improves the information interaction experience. As an optional implementation of the embodiment of the present disclosure, the determining of the problem solving state of the first inquiry information and the retrieval feedback information in the problem solving state according to the first inquiry information, the plurality of first recall information and the first intelligent agent comprises: obtaining meta information of the plurality of first recall information, and splicing each first recall information and its meta information to obtain a plurality of second recall information; determining the problem solving state of the first inquiry information and the retrieval feedback information in the problem solving state according to the first inquiry information, the plurality of second recall information and the first intelligent agent.

[0073] As described above, the meta information of data refers to high-level information describing the attributes of data itself, that is, "information about data". The meta information of data usually does not directly contain the specific content of the data, but annotates or explains one or more information such as the source, structure, context, relationship and update time of the data, which is similar to the "label" or "instruction manual" of the data. Meta information is usually used to represent one or more attribute information of data such as source attribute, structure attribute, semantic attribute and usage attribute. Among them, the source attribute can include at least one attribute of the data creator, publication time, version number and reference file. The structure attribute can include at least one attribute of the organization form (such as chapter division, node relationship in knowledge graph) and logical framework (such as argument chain of premise and conclusion) of the data. The semantic attribute can include at least one attribute of the data term definition, field label and multilingual comparison label. The usage attribute can include at least one attribute of the data usage permission and applicable scenario.

[0074] By structurally splicing each first recall information and its meta information to form second recall information, the timeliness, reliability and applicable scenarios of the first recall information can be comprehensively considered by the first intelligent agent when evaluating the problem solving state; and the decision accuracy of complex problems is significantly improved by embedding the meta information, when multiple recall information conflicts, the intelligent agent can automatically select more reliable content according to the meta information, and mark low confidence information for manual review. Moreover, the mechanism greatly optimizes the interpretability of the search feedback information, which meets the compliance requirements and enhances the trust of the search results. This dynamic weighting decision mode based on meta information can improve the accuracy of answers in complex scenarios while ensuring response speed, and is especially suitable for vertical fields such as law and medicine that require strict information accuracy.

[0075] S130, in response to the problem solving state being an event that the problem has been solved, determining the problem reply information of the first inquiry information according to the search feedback information, and displaying the problem reply information.

[0076] The problem reply information can be understood as the final output information for replying to the first inquiry information. When the problem reply information is text information, the problem reply information can be output in a stream (word by word) form, for example.

[0077] In the embodiments of the present disclosure, optionally, the determining the problem reply information of the first inquiry information according to the search feedback information comprises: determining the search feedback information as the problem reply information of the first inquiry information; or inputting the first inquiry information and the search feedback information into a second intelligent agent to obtain the problem reply information of the first inquiry information.

[0078] In one embodiment, when the first intelligent agent has the function of sorting and editing multiple first recall information, such as trimming and / or splicing the first recall information, etc., the description data for answering the first inquiry information can be generated, at this time, the search feedback information can be determined as the problem reply information of the first inquiry information. The advantage of such setting is that the final problem reply information can be output by the first intelligent agent "one-stop", without the need for further processing of the search feedback information, reducing the information transmission and processing flow, and improving the information response efficiency.

[0079] In another embodiment, the first intelligent agent can be configured to feed back the first retrieval feedback information directly as the retrieval feedback information or filter out a part of the first retrieval feedback information highly related to the first query information as the retrieval feedback information, and then input the first query information and the retrieval feedback information into a second intelligent agent to further process the retrieval feedback information by the second intelligent agent to generate the problem reply information of the first query information. In this way, the feedback of the problem reply data and the decision of the retrieval operation are separated, so that the intelligent agent has higher concentration and accuracy, and is helpful for the targeted adjustment and maintenance of different intelligent agents, and improves the flexibility of the information interaction architecture.

[0080] As an optional implementation manner of the embodiments of the present disclosure, the problem reply information can be displayed in various manners, including but not limited to: displaying the problem reply information corresponding to the first query information on an interactive interface where the first query information is input; and / or generating audio data of the problem reply information and playing the audio data through an audio playing device.

[0081] The technical scheme of the embodiments of the present disclosure comprises: obtaining first query information, converting the first query information into second query information, searching in a target database according to the first query information and the second query information to obtain a plurality of first recall information, constructing a double search triggering mechanism by converting the original query (the first query information) into a more complete semantic search instruction (the second query information), so that the target database can match from two dimensions of expression form and potential demand at the same time, and the hit rate of high correlation information (the first recall information) is significantly improved. According to the first query information, the plurality of first recall information and the first intelligent agent, the problem solving state of the first query information and the search feedback information in the problem solving state are determined, which is different from the consumptive thinking of the traditional architecture that needs multiple rounds of model self-questioning and self-answering. The present scheme forms a closed-loop decision through the instant state evaluation (problem solving state) of the first intelligent agent on the recall information. This single decision mechanism makes the response speed of simple problems close to the level of real-time interaction. According to the search feedback information, the problem reply information of the first query information is determined and the problem reply information is displayed in response to the event that the problem solving state is that the problem has been solved. In the present technical scheme, the first intelligent agent compares the coverage of the original query and the search result. When it is judged that the recall information has contained a complete solution, the subsequent processing process is terminated immediately, and the reply information is generated directly. The present scheme essentially reconstructs the value transmission chain of the intelligent agent system: the "model thinking time" in the traditional architecture is converted into "effective information matching time", and through the double optimization of front semantic enhancement and back state verification, the processing capacity of complex problems is maintained while the response speed of simple problems is improved by an order of magnitude. For the scene of low customer service error rate and high response timeliness requirement, this end-to-end processing process makes the user not need to perceive the multi-system cooperation process behind, meets the dialogue expectation, and can balance the accuracy and efficiency, and significantly improves the information interaction experience.

[0082] Figure 2FIG. 1 shows a flowchart of another information interaction method provided by an embodiment of the present disclosure. The technical solution of the embodiment is further refined on the basis of the above-mentioned embodiment, and the implementation manner of the at least partially unresolved scenario is further refined. Optionally, the problem solving state includes at least partially unresolved; the retrieval feedback information includes at least one third inquiry information, target retrieval data in the target database, and a retrieval manner of the target retrieval data; and the retrieval manner at least includes preset word retrieval and / or semantic retrieval. Further, after determining the problem solving state of the first inquiry information and the retrieval feedback information in the problem solving state according to the first inquiry information, the plurality of first recall information, and the first agent, the method further includes: in response to an event that the problem solving state is at least partially unresolved, retrieving according to the at least one third inquiry information, the target retrieval data in the target database, and the retrieval manner of the target retrieval data to obtain a plurality of third recall information; determining problem reply information of the first inquiry information according to the plurality of third recall information, and displaying the problem reply information. The specific implementation can refer to the description of the embodiment. Wherein, the same or similar technical features as the foregoing embodiments are not described herein. As shown in FIG. 1, the method of the embodiment can specifically include: Figure 3

[0083] S210, obtaining the first inquiry information, converting the first inquiry information into second inquiry information, and retrieving in the target database according to the first inquiry information and the second inquiry information to obtain a plurality of first recall information.

[0084] S220, determining the problem solving state of the first inquiry information and the retrieval feedback information in the problem solving state according to the first inquiry information, the plurality of first recall information, and the first agent; wherein, in the case that the problem solving state is at least partially unresolved, the retrieval feedback information includes at least one third inquiry information, target retrieval data in the target database, and a retrieval manner of the target retrieval data.

[0085] Specifically, in the case that the problem solving state includes that the problem is solved, S230 is performed; and in the case that the problem solving state is at least partially unresolved, S240 is performed.

[0086] S230, in response to an event that the problem solving state is that the problem is solved, determining problem reply information of the first inquiry information according to the retrieval feedback information, and displaying the problem reply information.

[0087] ​S240, in response to the problem solving state being an event that at least part of the problem is not solved, retrieving, according to at least one third query information, target retrieval data in the target database, and a retrieval manner of the target retrieval data, to obtain a plurality of third recall information, performing S250.

[0088] Briefly speaking, the third query information is equivalent to the semantic complement of the first query information. As described above, the third query information can be understood as the problem in the first query information that has not been solved. That is, the third query information can be the problem to be further retrieved generated by the first intelligent agent. More specifically, the third query information can include an explicit query problem not covered by the first recall result, and can also include a potential demand found through dialogue state analysis, such as a deep retrieval triggered by continuously asking the same type of question for three times. The target retrieval data in the target database can be target retrieval data in the retrievable data that has not been retrieved, and / or the retrievable data that has not been retrieved in a preset manner, etc. That is, the target retrieval data can be at least part of the retrievable data in the target database. Exemplarily, the target retrieval data can include newly added data that is not indexed or historical low-frequency access data, and / or data that has only been retrieved by a preset word but not been analyzed semantically, etc. The retrieval manner of the target retrieval data can be one or more of the preset retrieval manners. Exemplarily, the retrieval manner at least includes at least one of a preset word retrieval, a semantic retrieval, and a knowledge graph reasoning, etc. That is, a preset word matching and / or a similarity matching of a semantic feature vector, etc.

[0089] Specifically, the retrieval manner of the target retrieval data analyzed by the first intelligent agent can be used to retrieve a plurality of third recall information matched with the third query information from the target retrieval data in the target database. Through the dynamically generated third query information, the information interaction method has the problem self-evolution ability to improve the problem solving rate. Precise definition of the target retrieval data can reduce a large amount of invalid calculation and reduce resource consumption. Moreover, the synergistic effect of the hybrid retrieval strategy can show significant advantages in the complex query scene. Through the deep coupling of semantic understanding and retrieval logic, an intelligent retrieval system with cognitive flexibility is constructed.

[0090] S250, determining the problem reply information of the first query information according to the plurality of third recall information, and displaying the problem reply information.

[0091] In the embodiments of the present disclosure, the problem solving state of the first inquiry information and the retrieval feedback information in the problem solving state can be determined according to the first inquiry information, the plurality of first recall information, the plurality of third recall information and the first agent. Further, in response to the problem solving state being an event that the problem has been solved, the problem reply information of the first inquiry information is determined according to the retrieval feedback information, and the problem reply information is displayed. In response to the problem solving state being an event that at least part of the problem has not been solved, the retrieval is continued according to at least one third inquiry information newly generated by the first agent, the target retrieval data in the target database and the retrieval manner of the target retrieval data, to obtain a plurality of fourth recall information, and the process is repeated to obtain the problem reply information of the first inquiry information. In actual application, the number of execution rounds of the process can be determined according to the practice needs, so as to ensure the accuracy and completeness of the problem reply while ensuring the response efficiency of the model.

[0092] In one embodiment, the plurality of first recall information and the plurality of third recall information can be spliced with the meta information thereof, and then the plurality of first recall information after splicing the meta information, the plurality of third recall information after splicing the meta information and the first inquiry information are input into the first agent to determine the problem solving state of the first inquiry information and the retrieval feedback information in the problem solving state.

[0093] In another embodiment, the plurality of first recall information and the plurality of third recall information can be spliced with the meta information thereof, and then the plurality of first recall information after splicing the meta information, the plurality of third recall information after splicing the meta information, the first inquiry information, the data description information of the plurality of retrievable data, the retrieval execution information and the first agent are used to determine the problem solving state of the first inquiry information and the retrieval feedback information in the problem solving state.

[0094] The technical solutions of the embodiments of the present disclosure are as follows: in response to an event in which the problem solving state is at least partially unresolved, third inquiry information, target retrieval data in the target database, and a retrieval manner of the target retrieval data are used for retrieval to obtain a plurality of third recall information, that is, instead of forcibly generating a reply based on a single retrieval result, a targeted supplementary retrieval is initiated by using the third inquiry information and in combination with the retrievable data and the retrieval manner. This problem atomization splitting capability can effectively deal with comprehensive consultation scenarios and avoid the "irrelevant answer" phenomenon caused by incomplete information in the traditional architecture. Moreover, by inheriting the context of the first inquiry information, the third inquiry information can automatically add filtering parameters. This context transmission mechanism enables the progressive retrieval to always revolve around the inquiry expectation, which can effectively prevent the "topic drift" commonly seen in traditional multi-round dialogues. By determining the problem reply information of the first inquiry information according to the plurality of third recall information and displaying the problem reply information, the third recall information and the first recall information can be cross-verified to ensure that the first inquiry information can be solved. Through the state-driven retrieval strategy, dynamic adjustment, context-aware retrievable data accurate positioning, and multi-dimensional fusion of the results, the agent system can quickly process simple queries and has the ability to "unravel the Gordian knot" to solve complex problems. Compared with the general architecture described in the background art, this mechanism is particularly suitable for application scenarios with high professional threshold and multi-condition coupling, which greatly improves the completeness of the reply while ensuring the response speed.

[0095] Figure 3 The flowchart of an optional example of another information interaction method provided by the embodiments of the present disclosure is shown in FIG. 6. As shown in FIG. 6, the method includes the following steps. Figure 3As shown, the method of the embodiment can specifically include: a user inputs query information A, and enters an agent A. The agent A converts the query information A into query information B, and then respectively according to the query information A and the query information B, finds and acquires relevant information from corresponding data storage or knowledge base, to obtain recall information A and recall information B, to provide basic data for subsequent processing. Then, an initialization operation is performed, including: acquiring a tool list to acquire available tools from the tool list, and performing meta information assembly, so as to provide more abundant data support for subsequent retrieval decision. Among them, the tool list can contain various retrieval tools such as document retrieval and question and answer pair retrieval, to provide a tool set for information retrieval. In this example, the meta information of the recall information A and the recall information B can be assembled. And the model dialogue history is stored in the memory, that is, the history record of the interaction between the agent and the user is stored, to provide a context reference for the agent to process the current query. After the recall operation is performed for the first time, the agent B can determine whether the recall information A and the recall information B (i.e. existing knowledge) are sufficient to solve the problem of the query information A according to the recall information A and the recall information B after assembling the meta information thereof. If yes, one or more operations such as cutting and splicing meta information are performed on the recall information A and the recall information B to generate a retrieval result. If not, a tool needed for retrieval and a question needed for subsequent supplementary retrieval are generated, and then a retrieval tool is executed to further perform a document retrieval or a question and answer pair retrieval operation on the question needed for supplementary retrieval, and after the retrieval operation is completed, the step of acquiring existing knowledge is returned to be processed in a cycle until the problem is solved, to obtain problem reply information. In this example, the problem reply information can be output by the agent A to display the problem reply information again. By using the technical solution, double-path fast and accurate data recall can be realized. For simple problems, the target correct answer can often be acquired by one-time recall. Therefore, one-time recall is performed before entering a planning model (agent B), and the agent B is sent as context knowledge, so that most simple problems can be directly returned with an answer after the first recall, without multiple planning, thereby greatly shortening the response time delay. Moreover, the original problem and the rewritten problem are used as two paths of recall, which can fully utilize the rewritten problem which is clearer, and can avoid a small number of cases of rewriting errors by introducing the original problem recall path, and fully play the advantages of the agent A in knowledge positioning and summary in question and answer scenarios.

[0096] In this example, the storage into the database can also be processed to present different knowledge granularity. Figure 4As shown, specifically, for document data, the original document data is first converted in format to meet the requirements of subsequent processing. On the one hand, the document data after conversion in format is summarized in full text to generate a document description block. On the other hand, the document data after conversion in format can be processed in rich text, and the content after processing in rich text is sliced in document. On the one hand, the meta information (article title and / or paragraph title, etc.) can be appended to the document slice to obtain a document slice block. Each document slice block can correspond to a document slice. On the other hand, the document slice can be summarized in content to obtain the summary information of the document slice to obtain a slice summary block. Similarly, each slice summary block can correspond to store the summary information of a document slice. The summary information can be understood as a summary of the content of the document slice, which is less than the original content of the document slice. For question and answer data, the processing flow can include converting the question and answer data in format to match the database requirements. Then, the question and answer data after conversion in format is processed in rich text to generate a question block and an answer block respectively. Each question block can correspond to a question. Each answer block can correspond to an answer. For the products of the above-mentioned flow, including the document description block, the document slice block, the slice summary block, the question block and the answer block are processed in word segmentation and coding features to obtain their corresponding knowledge index information, and the processed index information and the above-mentioned products are associated and stored in the database. By using the technical solution, the information interaction of the embodiment of the present disclosure can be prevented, the diversity and heterogeneous characteristics of the knowledge source (data source) are supported, the multi-granularity knowledge recall tool is combined to maximize the adaptation degree of the recall tool and the knowledge structure, and the target correct knowledge is covered as much as possible, so that the information interaction experience is improved.

[0097] Figure 5 A structural schematic diagram of an information interaction device provided by the embodiment of the present disclosure is shown as Figure 5 As shown, the device comprises an inquiry information acquisition module 510, a retrieval information feedback module 520 and a reply information display module 530. The inquiry information acquisition module 510 is used to acquire first inquiry information, convert the first inquiry information into second inquiry information, retrieve in a target database according to the first inquiry information and the second inquiry information to obtain a plurality of first recall information; the retrieval information feedback module 520 is used to determine the problem solving state of the first inquiry information and the retrieval feedback information in the problem solving state according to the first inquiry information, the plurality of first recall information and a first intelligent agent; the reply information display module 530 is used to respond to the event that the problem solving state is that the problem has been solved, determine the problem reply information of the first inquiry information according to the retrieval feedback information, and display the problem reply information.

[0098] The technical scheme of the embodiments of the present disclosure obtains first query information through the query information acquisition module 510, converts the first query information into second query information, performs retrieval in a target database according to the first query information and the second query information to obtain a plurality of first recall information, constructs a double-retrieval triggering mechanism by converting the original query (the first query information) into a more complete semantic retrieval instruction (the second query information), so that the target database can simultaneously match from two dimensions of expression form and potential demand, and the hit rate of high-relevance information (the first recall information) is significantly improved. The retrieval information feedback module 520 determines the problem solving state of the first query information and the retrieval feedback information in the problem solving state according to the first query information, the plurality of first recall information and the first intelligent agent. Unlike the consumptive thinking of the traditional architecture that needs multiple rounds of model self-questioning and self-answering, the present solution forms a closed-loop decision through the instant state evaluation (problem solving state) of the first intelligent agent on the recall information. This single decision mechanism makes the response speed of simple problems close to the level of real-time interaction. The reply information display module 530 determines the problem reply information of the first query information according to the retrieval feedback information in response to the event that the problem solving state is that the problem has been solved, and displays the problem reply information. In the present technical solution, the first intelligent agent compares the coverage of the original query and the retrieval result, and when it is judged that the recall information already contains a complete solution, the subsequent processing process is terminated immediately, and the reply information is generated directly. This solution essentially reconstructs the value transmission chain of the intelligent agent system: it converts the "model thinking time" in the traditional architecture into "effective information matching time", and through the double optimization of front-end semantic enhancement and back-end state verification, it realizes the order-of-magnitude improvement of the response speed of simple problems while maintaining the processing capability of complex problems. For scenes with low customer service error rate and high response timeliness requirement, this end-to-end processing process makes the user not need to perceive the multi-system cooperation process behind, meets the dialogue expectation, can balance the accuracy and efficiency, and significantly improves the information interaction experience.

[0099] On the basis of any optional technical scheme in the embodiments of the present disclosure, optionally, the target database includes a plurality of retrievable data of various data types. Further, the retrieval information feedback module 520 can be specifically used to obtain data description information and retrieval execution information of a plurality of retrievable data, and determine the problem solving state of the first query information and the retrieval feedback information in the problem solving state according to the first query information, the plurality of first recall information, the data description information of the plurality of retrievable data, the retrieval execution information and the first intelligent agent.

[0100] On the basis of any optional technical solution in the embodiments of the present disclosure, the retrieval information feedback module 520 may, optionally, specifically include a meta information splicing unit and a retrieval feedback unit. The meta information splicing unit may be configured to obtain meta information of a plurality of pieces of the first recall information, splice each piece of the first recall information with its meta information respectively, and obtain a plurality of pieces of second recall information. The retrieval feedback unit may be configured to determine, according to the first query information, the plurality of pieces of second recall information, and a first agent, a problem solving state of the first query information and retrieval feedback information in the problem solving state.

[0101] On the basis of any optional technical solution in the embodiments of the present disclosure, the problem solving state may include at least partial problems unsolved, and the retrieval feedback information may include at least one third query information, target retrieval data in the target database, and a retrieval manner for the target retrieval data. The retrieval manner may include at least preset word retrieval and / or semantic retrieval.

[0102] On the basis of any optional technical solution in the embodiments of the present disclosure, the information interaction apparatus may further include a supplementary retrieval module and a reply determination module. The supplementary retrieval module may be configured to, after determining, according to the first query information, a plurality of pieces of first recall information, and a first agent, a problem solving state of the first query information and retrieval feedback information in the problem solving state, in response to an event that the problem solving state is at least partial problems unsolved, perform retrieval according to at least one third query information, target retrieval data in the target database, and a retrieval manner for the target retrieval data, to obtain a plurality of pieces of third recall information. The reply determination module may be configured to determine, according to the plurality of pieces of third recall information, problem reply information of the first query information, and display the problem reply information.

[0103] On the basis of any optional technical solution in the embodiments of the present disclosure, the reply information display module 530 may specifically be configured to determine the retrieval feedback information as the problem reply information of the first query information, or input the first query information and the retrieval feedback information into a second agent to obtain the problem reply information of the first query information.

[0104] On the basis of any optional technical solution in the embodiments of the present disclosure, the query information acquisition module 510 may specifically be configured to input the first query information into a third agent to obtain second query information.

[0105] Optionally, based on any of the optional technical solutions in the embodiments of the present disclosure, the searchable data in the target database comprises document data. Further, the information interaction apparatus further comprises a document global processing module. The document global processing module is configured to summarize content of the document data according to a document processing model to obtain document description information of the document data before the searching in the target database according to the first inquiry information and the second inquiry information, and store the document data and the document description information corresponding thereto in the target database.

[0106] Optionally, based on any of the optional technical solutions in the embodiments of the present disclosure, the searchable data in the target database comprises document data. Further, the information interaction apparatus further comprises a document global processing module. The document global processing module is configured to summarize content of the document data according to a document processing model to obtain document description information of the document data before the searching in the target database according to the first inquiry information and the second inquiry information, and store the document data and the document description information corresponding thereto in the target database.

[0107] Optionally, based on any of the optional technical solutions in the embodiments of the present disclosure, the document global processing module is specifically configured to divide the document data into a plurality of initial document segments according to a data format of content data included in the document data, wherein the data format comprises at least one of an image, a table, a link, a preset word, and a text.

[0108] Optionally, based on any of the optional technical solutions in the embodiments of the present disclosure, the document global processing module comprises a document segment determination unit and / or a segment description generation unit. The document segment determination unit is specifically configured to obtain meta information of the document data, and add the meta information to the initial document segment to obtain a target document segment. The segment description generation unit is configured to summarize content of the initial document segment according to a document processing model to obtain segment description information of the initial document segment, and store the initial document segment and the segment description information corresponding thereto in the target database.

[0109] Optionally, based on any of the optional technical solutions in the embodiments of the present disclosure, the searchable data in the target database comprises question and answer data. Further, the information interaction apparatus further comprises a question and answer data processing module. The question and answer data processing module is configured to generate question data and reply data according to the question and answer data before the searching in the target database according to the first inquiry information and the second inquiry information, and store the question data and the reply data corresponding thereto in the target database.

[0110] The information interaction apparatus provided by the embodiments of the present disclosure can execute the information interaction method provided by any of the embodiments of the present disclosure, and has the corresponding function modules and beneficial effects of executing the information interaction method.

[0111] It is worth noting that each unit and module included in the above apparatus is only divided according to the function logic, but is not limited to the above division, as long as the corresponding function can be realized; in addition, the specific name of each functional unit is only for convenient mutual distinction, and does not serve to limit the protection scope of the embodiments of the present disclosure.

[0112] The following refers to Figure 6 , which shows a structural schematic diagram of an electronic device (such as a terminal device or a server) 600 suitable for implementing the embodiments of the present disclosure. The terminal device in the embodiments of the present disclosure can include, but is not limited to, mobile terminals such as mobile phones, notebook computers, digital broadcast receivers, PDAs (Personal Digital Assistants), PADs (Tablet Personal Computers), PMPs (Portable Multimedia Players), vehicle-mounted terminals (such as vehicle-mounted navigation terminals), and the like, as well as fixed terminals such as digital TVs, desktop computers, and the like. Figure 6 The electronic device shown is only an example, and should not bring any limitation to the functions and use range of the embodiments of the present disclosure.

[0113] As shown in Figure 6 , the electronic device 600 can include a processing apparatus (such as a central processor, a graphics processor, etc.) 601, which can perform various appropriate actions and processes according to programs stored in a read-only memory (ROM) 602 or programs loaded from a storage apparatus 608 into a random access memory (RAM) 603. In the RAM 603, various programs and data required for the operation of the electronic device 600 are also stored. The processing apparatus 601, the ROM 602, and the RAM 603 are connected to each other through a bus 604. An input / output (I / O) interface 605 is also connected to the bus 604.

[0114] Generally, the following apparatuses can be connected to the I / O interface 605: an input apparatus 606 including, for example, a touch screen, a touch pad, a keyboard, a mouse, a camera, a microphone, an accelerometer, a gyroscope, and the like; an output apparatus 607 including, for example, a liquid crystal display (LCD), a speaker, a vibrator, and the like; a storage apparatus 608 including, for example, a magnetic tape, a hard disk, and the like; and a communication apparatus 609. The communication apparatus 609 can allow the electronic device 600 to communicate with other devices wirelessly or by wire to exchange data. Although Figure 6 The electronic device 600 with various apparatuses is shown, but it should be understood that it is not required to implement or have all the apparatuses shown. More or fewer apparatuses can be alternatively implemented or provided.

[0115] In particular, according to embodiments of the present disclosure, the processes described above with reference to the flowcharts can be implemented as a computer software program. For example, embodiments of the present 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 executing the methods illustrated by the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network by the communication apparatus 609, or installed from the storage apparatus 608, or installed from the ROM 602. When the computer program is executed by the processing apparatus 601, the above-mentioned functions defined in the methods of the embodiments of the present disclosure are executed.

[0116] The names of the messages or information exchanged between the plurality of devices in the embodiments of the present disclosure are only for illustrative purposes, and are not intended to limit the scope of the messages or information.

[0117] The electronic device provided by the embodiments of the present disclosure and the information interaction method provided by the above-mentioned embodiments belong to the same inventive concept, and the technical details not described in detail in the embodiments of the present disclosure can be referred to the above-mentioned embodiments, and the present embodiments have the same beneficial effects as the above-mentioned embodiments.

[0118] The embodiments of the present disclosure provide a computer storage medium, which stores a computer program, and the program is executed by a processor to implement the information interaction method provided by the above-mentioned embodiments.

[0119] It should be noted that the computer readable medium in the foregoing of the present disclosure can be a computer readable signal medium or a computer readable storage medium or any combination of the two. The computer readable storage medium may, for example, but is not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or apparatus, or any combination of the above. More specific examples of the computer readable storage medium can include, but are not limited to, an electrical connection having one or more wires, a portable computer diskette, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In the present disclosure, the 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, device or apparatus. In the present disclosure, the computer readable signal medium can include a data signal carried in a baseband or as a part of a carrier wave, which carries computer readable program code. Such a propagated data signal can take many forms, including but not limited to an electromagnetic signal, an optical signal or any suitable combination of the above. The computer readable signal medium can also be any computer readable medium other than the computer readable storage medium, which can send, propagate or transmit a program for use by or in conjunction with an instruction execution system, device or apparatus. The program code contained in the computer readable medium can be transmitted by any suitable medium, including but not limited to a wire, a cable, an RF (radio frequency) or the like, or any suitable combination of the above.

[0120] According to one or more embodiments of the present disclosure,

Example One

[0121] According to one or more embodiments of the present disclosure, Example Two provides the method of Example One, further comprising: optionally, the searchable data in the target database comprises a plurality of data types; and the determining, by the first agent, the problem resolution state of the first query information and the retrieval feedback information in the problem resolution state based on the first query information, the plurality of first recall information, and the first agent comprises: obtaining data description information of the plurality of searchable data and retrieval execution information thereof, and determining, by the first agent, the problem resolution state of the first query information and the retrieval feedback information in the problem resolution state based on the first query information, the plurality of searchable data, the data description information thereof, the retrieval execution information, and the first agent.

[0122] According to one or more embodiments of the present disclosure, Example Three provides the method of Example One, further comprising: optionally, the determining, by the first agent, the problem resolution state of the first query information and the retrieval feedback information in the problem resolution state based on the first query information, the plurality of first recall information, and the first agent comprises: obtaining meta-information of the plurality of first recall information, and splicing each of the first recall information and the meta-information thereof to obtain a plurality of second recall information; and determining, by the first agent, the problem resolution state of the first query information and the retrieval feedback information in the problem resolution state based on the first query information, the plurality of second recall information, and the first agent.

[0123] According to one or more embodiments of the present disclosure, Example Four provides the method of Example One, further comprising: optionally, the problem resolution state comprises at least partial problem unresolved; the retrieval feedback information comprises at least one third query information, target retrieval data in the target database, and a retrieval manner for the target retrieval data; and the retrieval manner comprises at least preset word retrieval and / or semantic retrieval.

[0124] According to one or more embodiments of the present disclosure, Example Five provides the method of Example Four, further comprising: optionally, after the determining, by the first agent, the problem resolution state of the first query information and the retrieval feedback information in the problem resolution state based on the first query information, the plurality of first recall information, and the first agent, further comprising: in response to an event that the problem resolution state is at least partial problem unresolved, performing retrieval based on the at least one third query information, the target retrieval data in the target database, and the retrieval manner for the target retrieval data to obtain a plurality of third recall information; determining, by the first agent, problem reply information of the first query information based on the plurality of third recall information, and displaying the problem reply information.

[0125] According to one or more embodiments of the present disclosure, Example Six provides the method of Example One, further comprising: optionally, the determining the question answer information of the first query information according to the search feedback information comprises: determining the search feedback information as the question answer information of the first query information; or inputting the first query information and the search feedback information into a second intelligent agent to obtain the question answer information of the first query information.

[0126] According to one or more embodiments of the present disclosure, Example Seven provides the method of Example Four, further comprising: optionally, the converting the first query information into the second query information comprises: inputting the first query information into a third intelligent agent to obtain the second query information.

[0127] According to one or more embodiments of the present disclosure, Example Eight provides the method of Example One, further comprising: optionally, the searchable data in the target database comprises document data; and before the searching in the target database according to the first query information and the second query information, the method further comprises: summarizing content of the document data according to a document processing model to obtain document description information of the document data, and storing the document data and the document description information corresponding thereto into the target database.

[0128] According to one or more embodiments of the present disclosure, Example Nine provides the method of Example One, further comprising: optionally, the searchable data in the target database comprises document data; and before the searching in the target database according to the first query information and the second query information, the method further comprises: dividing the document data to obtain a plurality of initial document segments, determining a target document segment according to the initial document segments, and storing the target document segment into the target database.

[0129] According to one or more embodiments of the present disclosure, Example Ten provides the method of Example Nine, further comprising: optionally, the dividing the document data to obtain a plurality of initial document segments comprises: dividing the document data into a plurality of initial document segments according to a data format of content data included in the document data; and wherein the data format comprises at least one of an image, a table, a link, a preset word, and a text.

[0130] According to one or more embodiments of the present disclosure, Example Eleven provides the method of Example Nine, further comprising: optionally, the determining the target document segment according to the initial document segment comprises: obtaining meta information of the document data, adding the meta information into the initial document segment to obtain a target document segment; and / or summarizing content of the initial document segment according to a document processing model to obtain segment description information of the initial document segment, and storing the initial document segment and the segment description information thereof into the target database correspondingly.

[0131] According to one or more embodiments of the present disclosure, Example Twelve provides the method of Example One, further comprising: optionally, the retrievable data in the target database comprises question and answer data; and before the retrieving in the target database according to the first inquiry information and the second inquiry information, further comprising: generating question data and reply data according to the question and answer data, and storing the question data and the reply data thereof into the target database correspondingly.

[0132] According to one or more embodiments of the present disclosure, Example Thirteen provides an information interaction device, comprising: an inquiry information acquisition module, configured to acquire first inquiry information, convert the first inquiry information into second inquiry information, and retrieve in a target database according to the first inquiry information and the second inquiry information to obtain a plurality of first recall information; a retrieval information feedback module, configured to determine, according to the first inquiry information, the plurality of first recall information and a first intelligent agent, a problem solving state of the first inquiry information and retrieval feedback information in the problem solving state; and a reply information display module, configured to, in response to an event that the problem solving state is that the problem has been solved, determine, according to the retrieval feedback information, problem reply information of the first inquiry information, and display the problem reply information.

[0133] In some embodiments, the client, server can communicate using any currently known or future developed network protocols, such as HTTP (HyperText Transfer Protocol), and can be interconnected with any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include local area networks ("LANs"), wide area networks ("WANs"), inter-networks (e.g., the Internet), and peer-to-peer networks (e.g., ad hoc peer-to-peer networks), as well as any currently known or future developed networks.

[0134] The above computer readable medium can be included in the above electronic device; or can exist separately without being assembled into the electronic device.

[0135] The computer readable medium described above carries one or more programs, which when executed by the electronic device, cause the electronic device to: acquire first query information, convert the first query information into second query information, perform a search in a target database according to the first query information and the second query information to obtain a plurality of first recall information, determine a problem solving state of the first query information and search feedback information in the problem solving state according to the first query information, the plurality of first recall information and a first intelligent agent, and in response to an event that the problem solving state is that the problem has been solved, determine problem reply information of the first query information according to the search feedback information, and display the problem reply information.

[0136] Computer program code for carrying out operations of the present disclosure can be written in any combination of one or more programming languages, including an object oriented programming language such as Java, Smalltalk, C++ or the like and conventional procedural programming languages, such as the "C" programming language or similar programming languages. The program code can execute entirely on the user's computer, partly on the user's computer, as a stand-alone software package, partly on the user's computer and partly on a remote computer or entirely on the remote computer or server. In the latter scenario, the remote computer can be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or the connection can be made to an external computer (for example, through the Internet using an Internet Service Provider).

[0137] The computer program instructions can also be loaded onto a computer or other programmable information processing apparatus to cause a series of operations to be performed on the computer or other programmable information processing apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable information processing apparatus implement the functions / acts specified in the flowchart and / or block diagram block or blocks.

[0138] The units and modules described in the embodiments of the present disclosure can be implemented in the form of software, or can be implemented in the form of hardware. In some cases, the names of the units and modules do not constitute a limitation on the units themselves, for example, the inquiry information acquisition module can also be described as "a module that retrieves a plurality of pieces of first recall information according to the first inquiry information and the second inquiry information converted from the first inquiry information".

[0139] The functions described above in this document can be performed, at least in part, by one or more hardware logic components. For example, and without limitation, exemplary types of hardware logic components that can be used include: field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), system on a chips (SOCs), complex programmable logic devices (CPLDs), etc.

[0140] In the context of the present disclosure, a machine-readable medium can be a tangible medium that contains or stores a program for use by or in connection with an instruction execution system, apparatus, or device. The machine-readable medium can be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium can include, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any suitable combination of the foregoing. More specific examples of the machine-readable storage medium will include one or more lines of electrical connections, portable computer disks, hard disk drives, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or Flash memory), optical fibers, portable compact disc read-only memories (CD-ROMs), optical storage devices, magnetic storage devices, or any suitable combination of the foregoing.

[0141] The above description is merely the preferred embodiments of the present disclosure and the explanation of the principles of the applied technology. It should be understood by those skilled in the art that the disclosed scope of the present disclosure is not limited to the technical solutions formed by the specific combinations of the technical features described above, and also covers other technical solutions formed by any combinations of the technical features described above or their equivalent features without departing from the disclosed concept. For example, the technical solutions formed by the mutual replacement of the above-described features and the technical features disclosed in the present disclosure (but not limited to) having similar functions.

[0142] Moreover, while operations are depicted in a particular order, this should not be understood as requiring such an order nor infringing on the scope of the disclosure. Certain of the operations described in the discussion are combinable into a single operation, and certain operations can be separated into several operations. In some embodiments, the operations described in the discussion can be performed in an order different than presented in the discussion. In some embodiments, the operations described in the discussion can be performed concurrently. Also, while several specific implementation details are discussed in the discussion, these should not be interpreted as limiting the scope of the disclosure. Rather, certain features described in the context of separate embodiments can also be implemented in combination in a single embodiment. Conversely, various features described in the context of a single embodiment can also be implemented in multiple embodiments separately or in any suitable sub-combination.

[0143] Although the subject matter has been described in language specific to structural features and / or methodological acts, it is to be understood that the subject matter defined in the appended claims is not necessarily limited to the specific features or acts described above. Rather, the specific features and acts described above are disclosed as example forms of implementing the claims.

Claims

1. An information exchange method, characterized in that, include: Obtain first query information, convert the first query information into second query information, and search the target database based on the first query information and the second query information to obtain multiple first recall information; Based on the first query information, multiple first recall messages, and the first intelligent agent, determine the problem-solving status of the first query information and the retrieval feedback information under the problem-solving status; In response to an event that the problem resolution status is "problem resolved", the question answer information of the first query information is determined based on the retrieval feedback information, and the question answer information is displayed.

2. The information interaction method according to claim 1, characterized in that, The target database includes searchable data of various data types; the step of determining the problem-solving status of the first query information and the retrieval feedback information under the problem-solving status based on the first query information, multiple pieces of the first recall information, and the first agent includes: The system acquires data description information and retrieval execution information of various types of searchable data. Based on the first query information, multiple first recall information entries, the data description information of various types of searchable data, the retrieval execution information, and the first intelligent agent, it determines the problem-solving status of the first query information and the retrieval feedback information under the problem-solving status.

3. The information interaction method according to claim 1, characterized in that, The step of determining the problem-solving status of the first query information and the retrieval feedback information under the problem-solving status based on the first query information, multiple first recall information messages, and the first intelligent agent includes: Obtain the metadata of multiple first recall messages, and concatenate each first recall message with its metadata to obtain multiple second recall messages; Based on the first query information, multiple second recall messages, and the first intelligent agent, the problem-solving status of the first query information and the retrieval feedback information under the problem-solving status are determined.

4. The information interaction method according to claim 1, characterized in that, The problem resolution status includes at least some problems remaining unresolved; the retrieval feedback information includes at least one third query, target retrieval data in the target database, and retrieval methods for the target retrieval data; the retrieval methods include at least preset word retrieval and / or semantic retrieval.

5. The information interaction method according to claim 4, characterized in that, After determining the problem-solving status of the first query information and the retrieval feedback information under the problem-solving status based on the first query information, multiple first recall information, and the first agent, the method further includes: In response to an event that the problem resolution status is at least partially unresolved, a search is performed based on at least one third query message, target retrieval data in the target database, and a retrieval method for the target retrieval data to obtain multiple third recall messages; The question response information for the first inquiry information is determined based on multiple pieces of the third recall information, and the question response information is displayed.

6. The information interaction method according to claim 1, characterized in that, Determining the question answer information for the first query information based on the retrieval feedback information includes: The retrieval feedback information is determined as the answer to the question in the first query; or... The first query information and the search feedback information are input into the second intelligent agent to obtain the question answer information of the first query information.

7. The information interaction method according to claim 1, characterized in that, The step of converting the first query information into the second query information includes: The first query information is input into the third agent to obtain the second query information.

8. The information interaction method according to claim 1, characterized in that, The searchable data in the target database includes document data; prior to the search in the target database based on the first query information and the second query information, the method further includes: The document data content is summarized according to the document processing model to obtain the document description information of the document data, and the document data and its document description information are stored in the target database.

9. The information interaction method according to claim 1, characterized in that, The searchable data in the target database includes document data; prior to the search in the target database based on the first query information and the second query information, the method further includes: The document data is divided into multiple initial document fragments, a target document fragment is determined based on the initial document fragments, and the target document fragment is stored in the target database.

10. The information interaction method according to claim 9, characterized in that, The process of dividing the document data to obtain multiple initial document fragments includes: According to the data format of the content data included in the document data, the document data is divided into multiple initial document fragments; wherein, the data format includes at least one of images, tables, links, preset words, and text.

11. The information interaction method according to claim 9, characterized in that, The step of determining the target document fragment based on the initial document fragment includes: Obtain the metadata of the document data, add the metadata to the initial document fragment to obtain the target document fragment; and / or, The content of the initial document fragment is summarized according to the document processing model to obtain the fragment description information of the initial document fragment, and the initial document fragment and its fragment description information are stored in the target database.

12. The information interaction method according to claim 1, characterized in that, The searchable data in the target database includes question-and-answer data; before the search is performed in the target database based on the first query information and the second query information, the method further includes: Question data and answer data are generated based on the question and answer data, and the question data and its answer data are stored in the target database accordingly.

13. An information interaction device, characterized in that, include: The query information acquisition module is used to acquire first query information, convert the first query information into second query information, and search the target database based on the first query information and the second query information to obtain multiple first recall information; The retrieval information feedback module is used to determine the problem-solving status of the first query information and the retrieval feedback information under the problem-solving status based on the first query information, multiple first recall information, and the first intelligent agent. The response information display module is used to respond to an event where the problem resolution status is "problem resolved", determine the problem response information of the first query information based on the search feedback information, and display the problem response information.

14. An electronic device, characterized in that, The electronic device includes: One or more processors; Storage device for storing one or more programs. When the one or more programs are executed by the one or more processors, the one or more processors implement the information interaction method as described in any one of claims 1-12.

15. A storage medium containing computer-executable instructions, characterized in that, The computer-executable instructions, when executed by a computer processor, are used to perform the information interaction method as described in any one of claims 1-12.

16. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by the processor, it implements the information interaction method as described in any one of claims 1-12.

Images