Data archiving method, data retrieval method, data archiving device, data retrieval device, medium, equipment and product

Abstract

The invention relates to a data archiving method, a data retrieval method, a data archiving device, a data retrieval device, a medium, equipment and a product, and relates to the technical field of artificial intelligence, the method comprises the following steps: under the condition that a current dialogue meets an automatic archiving condition, obtaining dialogue content of the current dialogue; and generating dialogue data information of the current dialogue according to the dialogue content. And filing the dialogue data information into a target database. Through the automatic filing process provided by the invention, the limitation of memory storage of the traditional general large model is effectively solved. Dialogue content meeting conditions is automatically obtained, active capture of information is achieved, and timeliness and integrity of memory storage are improved. And the generated dialogue data information is archived to the target database, so that the model is endowed with an automatic archiving capability, the information is orderly precipitated and managed, the information management efficiency and the depth and breadth of model knowledge precipitation are greatly enhanced, and the application value of a large model in an information management scene is expanded.

Description

Technical Field

[0001] The present disclosure relates to the field of artificial intelligence technology, and in particular to a data archiving method, retrieval method, apparatus, medium, device, and product. Background Art

[0002] With the rapid development of artificial intelligence (AI), large, general-purpose models have demonstrated powerful capabilities in fields such as natural language processing and are widely used in scenarios such as conversational interaction and text generation. However, these models currently have significant limitations in terms of memory storage, requiring users to actively trigger or fully memorize conversations. This significantly limits the models' usefulness in information management and knowledge accumulation, necessitating the urgent need for effective solutions to overcome these technical bottlenecks. Summary of the Invention

[0003] In order to solve the above technical problems, the present disclosure provides a data archiving method, a retrieval method, an apparatus, a medium, a device and a product.

[0004] In a first aspect, the present disclosure provides a data archiving method, the method comprising: When the current conversation meets the automatic archiving conditions, obtain the conversation content of the current conversation; Generating conversation data information of the current conversation according to the conversation content; Archiving the conversation data information into a target database; The automatic archiving conditions include any one of the following: The number of conversation turns in the current conversation is greater than or equal to a first preset number threshold; The total number of conversations initiated within a preset time period is greater than or equal to a second preset number threshold, where the preset time period includes the time corresponding to the current conversation; The conversation content of the current conversation contains preset keywords.

[0005] Optionally, generating the conversation data information of the current conversation according to the conversation content includes: extracting key conversation information from the conversation content; The conversation data information is generated according to the conversation key information.

[0006] Optionally, generating the conversation data information according to the conversation key information includes: The conversation key information is encapsulated, and the encapsulated conversation key information is used as the conversation data information.

[0007] Optionally, the method further includes: Determining the timestamp information and conversation scene information corresponding to the current conversation based on the conversation content; Generating the conversation data information according to the conversation key information includes: The conversation data information is generated according to the timestamp information, the conversation scene information and the conversation key information.

[0008] Optionally, generating the conversation data information according to the timestamp information, the conversation scene information, and the conversation key information includes: Associating the timestamp information and the conversation scene information with the conversation key information; The conversation data information is obtained according to the associated timestamp information, conversation scene information and conversation key information.

[0009] In a second aspect, the present disclosure provides a data retrieval method, the method comprising: Get the input information entered by the user; According to the input information, target data information is obtained from a plurality of conversation data information stored in a target database; the conversation data information in the target database is archived according to the data archiving method provided by the first aspect of the present disclosure; Output the target data information.

[0010] Optionally, acquiring target data information from a plurality of conversation data information stored in a target database according to the input information includes: The input information is matched with the plurality of conversation data information, and the successfully matched conversation data information is used as the target data information.

[0011] Optionally, the input information includes data retrieval text, the conversation data information includes conversation key information, and matching the input information with the plurality of conversation data information and using the successfully matched conversation data information as the target data information includes: The data retrieval text is matched with the plurality of conversation key information, and the successfully matched conversation key information is used as the target data information.

[0012] Optionally, the data retrieval text includes a conversation retrieval time and / or a conversation retrieval scenario, and the conversation data information further includes timestamp information and conversation scenario information corresponding to the conversation key information. When the data retrieval text includes a conversation retrieval time, matching the data retrieval text with the plurality of conversation key information includes: Matching the conversation retrieval time with timestamp information corresponding to the plurality of conversation key information; In a case where the data retrieval text includes a conversation retrieval scenario, matching the data retrieval text with the plurality of conversation key information includes: Matching the conversation retrieval scene with the conversation scene information corresponding to the plurality of conversation key information; In a case where the data retrieval text includes a conversation retrieval scene and a conversation retrieval time, matching the data retrieval text with the plurality of conversation key information includes: The conversation retrieval scene is matched with the conversation scene information corresponding to the multiple conversation key information, and the conversation retrieval time is matched with the timestamp information corresponding to the multiple conversation key information.

[0013] Optionally, the input information includes question text, the dialogue data information includes dialogue key information, and acquiring target data information from multiple dialogue data information stored in a target database according to the input information includes: Determining the user's question intention based on the question text; Match the question intention with the multiple conversation key information, and use the successfully matched conversation key information as the target data information.

[0014] Optionally, the method further includes: Generate a response text corresponding to the question text according to the question text; The outputting of the target data information includes: Generate a suggestion prompt text according to the response text and the target data information; The response text and the suggestion prompt text are output.

[0015] In a third aspect, the present disclosure provides a data archiving device, the device comprising: A first acquisition module is configured to acquire the conversation content of the current conversation when the current conversation meets the automatic archiving condition; A generating module configured to generate conversation data information of a current conversation based on the conversation content; an archiving module, configured to archive the conversation data information into a target database; The automatic archiving conditions include any one of the following: The number of conversation turns in the current conversation is greater than or equal to a first preset number threshold; The total number of conversations initiated within a preset time period is greater than or equal to a second preset number threshold, where the preset time period includes the time corresponding to the current conversation; The conversation content of the current conversation contains preset keywords.

[0016] Optionally, the generating module is configured to extract conversation key information from the conversation content; and generate the conversation data information according to the conversation key information.

[0017] Optionally, the generating module is configured to encapsulate the conversation key information and use the encapsulated conversation key information as the conversation data information.

[0018] Optionally, the device further comprises: a determination module configured to determine timestamp information and conversation scene information corresponding to the current conversation based on the conversation content; The generation module is configured to generate the dialogue data information according to the timestamp information, the dialogue scene information and the dialogue key information.

[0019] Optionally, the generating module is configured to associate the timestamp information and the conversation scene information with the conversation key information; and obtain the conversation data information based on the associated timestamp information, conversation scene information and conversation key information.

[0020] In a fourth aspect, the present disclosure provides a data retrieval device, the device comprising: A second acquisition module is configured to acquire input information input by a user; a third acquisition module configured to acquire target data information from a plurality of conversation data information stored in a target database according to the input information; the conversation data information in the target database is archived according to the data archiving method provided in the first aspect of the present disclosure; The output module is configured to output the target data information.

[0021] Optionally, the third acquisition module is configured to match the input information with the plurality of conversation data information, and use the successfully matched conversation data information as the target data information.

[0022] Optionally, the input information includes data retrieval text, the conversation data information includes conversation key information, and the third acquisition module is configured to match the data retrieval text with the multiple conversation key information, and use the successfully matched conversation key information as the target data information.

[0023] Optionally, the data retrieval text includes a conversation retrieval time and / or a conversation retrieval scenario, and the conversation data information further includes timestamp information and conversation scenario information corresponding to the conversation key information. In the case where the data retrieval text includes the conversation retrieval time, the third acquisition module is configured to match the conversation retrieval time with the timestamp information corresponding to the plurality of conversation key information. In the case where the data retrieval text includes a conversation retrieval scenario, the third acquisition module is configured to match the conversation retrieval scenario with the conversation scenario information corresponding to the plurality of conversation key information; In the case where the data retrieval text includes a conversation retrieval scene and a conversation retrieval time, the third acquisition module is configured to match the conversation retrieval scene with the conversation scene information corresponding to the multiple conversation key information, and to match the conversation retrieval time with the timestamp information corresponding to the multiple conversation key information.

[0024] Optionally, the input information includes question text, the conversation data information includes conversation key information, and the third acquisition module is configured to determine the user's question intention based on the question text; match the question intention with multiple conversation key information, and use the successfully matched conversation key information as the target data information.

[0025] Optionally, the output module is configured to generate a response text corresponding to the question text based on the question text; generate a suggestion prompt text based on the response text and the target data information; and output the response text and the suggestion prompt text.

[0026] In a fifth aspect, the present disclosure provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the data archiving method provided in the first aspect of the present disclosure, or implements the steps of the data retrieval method provided in the second aspect of the present disclosure.

[0027] In a sixth aspect, the present disclosure provides an electronic device, including: a memory having a computer program stored thereon; A processor is used to execute the computer program in the memory to implement the steps of the data archiving method provided by the first aspect of the present disclosure, or to implement the steps of the data retrieval method provided by the second aspect of the present disclosure.

[0028] In a seventh aspect, the present disclosure provides a computer program product, comprising a computer program, which, when executed by a processor, implements the steps of the data archiving method provided in the first aspect of the present disclosure, or implements the steps of the data retrieval method provided in the second aspect of the present disclosure.

[0029] Through the above technical solution, first, if the current conversation meets the automatic archiving conditions, the conversation content of the current conversation is obtained. Then, based on the conversation content, conversation data information for the current conversation is generated. Finally, the conversation data information is archived to a target database. The automatic archiving conditions include any of the following: the number of conversation turns in the current conversation is greater than or equal to a first preset threshold; the total number of conversation initiations within a preset time period is greater than or equal to a second preset threshold, the preset time period including the time corresponding to the current conversation; and the conversation content of the current conversation contains preset keywords. The automated archiving process provided by this disclosure effectively addresses the limitations of traditional general-purpose large-scale model memory storage. First, conversation content that meets the conditions is automatically acquired, enabling proactive information capture when the automatic archiving conditions are met, improving the timeliness and completeness of memory storage. Second, conversation data information is generated, converting the original conversation content into structured data for subsequent retrieval and analysis. Finally, the generated conversation data information is archived to a target database, empowering the model with automated archiving capabilities, enabling orderly information accumulation and management. This significantly enhances information management efficiency and the depth and breadth of model knowledge accumulation, expanding the application value of large-scale models in information management scenarios. In this way, deep conversations, high-frequency conversations and conversations with key content can be automatically archived, greatly simplifying user operations and improving user experience.

[0030] Other features and advantages of the present disclosure will be described in detail in the following detailed description. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] The accompanying drawings are used to provide a further understanding of the present disclosure and constitute a part of the specification. Together with the following detailed description, they are used to explain the present disclosure but do not constitute a limitation of the present disclosure. In the accompanying drawings: Figure 1 The figure is a flowchart of a data archiving method according to an exemplary embodiment.

[0032] Figure 2 The figure is a flowchart of a data retrieval method according to an exemplary embodiment.

[0033] Figure 3 The figure is a schematic diagram of a data processing flow according to an exemplary embodiment.

[0034] Figure 4 The figure is a block diagram of a data archiving device according to an exemplary embodiment.

[0035] Figure 5 The figure is a block diagram of another data archiving device according to an exemplary embodiment.

[0036] Figure 6The figure is a block diagram of a data retrieval device according to an exemplary embodiment.

[0037] Figure 7 It is a block diagram of an electronic device according to an exemplary embodiment. DETAILED DESCRIPTION

[0038] The following describes the specific embodiments of the present disclosure in detail with reference to the accompanying drawings. It should be understood that the specific embodiments described herein are only used to illustrate and explain the present disclosure and are not intended to limit the present disclosure.

[0039] The terms "first," "second," and the like in the specification and claims of this application and the accompanying drawings are used to distinguish similar objects and are not necessarily to be construed as implying a particular order or sequential sequence. In addition, in the description with reference to the accompanying drawings, the same reference numerals in different drawings represent the same elements.

[0040] As used herein, the term "including" and its variations are open-ended, i.e., "including but not limited to." The term "based on" means "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," and the term "some embodiments" means "at least some embodiments." Other terms are defined in the following description.

[0041] In the description of this disclosure, unless otherwise specified, "plurality" refers to two or more than two, and other quantifiers are similar; "at least one item", "one or more items" or similar expressions refer to any combination of these items, including any combination of single or plural items. For example, at least one item a can represent any number of a; for another example, one or more items among a, b, and c can represent: a, b, c, ab, ac, bc, or abc, where a, b, and c can be single or plural; "and / or" is a type of relationship that describes the association of related objects, indicating that three relationships can exist. For example, A and / or B can represent: A exists alone, A and B exist at the same time, and B exists alone, where A and B can be singular or plural. The character " / " indicates that the related objects are in an "or" relationship.

[0042] Although operations or steps are described in a particular order in the drawings in the embodiments of the present disclosure, this should not be understood as requiring that these operations or steps be performed in the particular order shown or in a serial order, or that all of the operations or steps shown be performed to obtain a desired result. In the embodiments of the present disclosure, these operations or steps may be performed serially; these operations or steps may also be performed in parallel; or some of these operations or steps may be performed.

[0043] Before introducing the data archiving method, retrieval method, apparatus, medium, equipment and product provided by the present disclosure, the application scenarios involved in each embodiment of the present disclosure are first introduced. With the vigorous development of artificial intelligence technology, general large models have demonstrated powerful capabilities in fields such as natural language processing and are widely used in scenarios such as dialogue interaction and text generation. However, current general large models of this type have significant limitations in memory storage functions. They require users to actively trigger memory or memorize all conversations, and the archiving strategy is rigid, which also leads to the possibility of key information loss or redundant storage in the model. For example, users must explicitly enter specific instructions such as "remember this content" before the model will memorize, and it only supports passive storage in text form and cannot actively capture and store key information. At the same time, it lacks automated archiving and structured indexing capabilities, making it difficult to efficiently retrieve and manage stored information, greatly limiting the application value of the model in information management and knowledge precipitation. It is urgent to find an effective solution to break through these technical bottlenecks.

[0044] In order to solve the above technical problems, the present invention provides a data archiving method, retrieval method, device, medium, equipment and product. Through the automated archiving process provided by the present invention, the limitations of traditional general large model memory storage are effectively solved. First, the conversation content that meets the conditions is automatically obtained, which changes the passive mode that requires the user to explicitly input instructions. When the automatic archiving conditions are met, the active capture of information is realized, which improves the timeliness and completeness of memory storage; secondly, the conversation data information is generated, which can convert the original conversation content into structured data for subsequent retrieval and analysis; finally, the conversation data information is archived to the target database, giving the system the ability to automatically archive, so that information can be deposited and managed in an orderly manner, greatly enhancing the efficiency of information management and the depth and breadth of model knowledge deposition, and expanding the application value of large models in information management scenarios. In this way, automatic archiving can be achieved for deep conversations, high-frequency conversations and conversations with key content, which greatly simplifies user operations and improves user experience.

[0045] The specific embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0046] Figure 1 FIG. 1 is a flow chart showing a data archiving method according to an exemplary embodiment. Figure 1 As shown, the method may include the following steps: In step S101, when the current conversation meets the automatic archiving conditions, the conversation content of the current conversation is obtained.

[0047] In some embodiments, the automatic archiving condition may include any one of the following three conditions: 1. The number of conversation turns in the current conversation is greater than or equal to a first preset threshold.

[0048] A conversation turn generally refers to the number of times or stages in which two (or more) parties take turns speaking during a conversational interaction. That is, in a complete conversation, each participant's speech (or single input) counts as a turn. For example, a user asks a question (turn 1) → the system answers (turn 2) → the user asks a follow-up question (turn 3) → the system answers again (turn 4)—this constitutes four conversation turns. If the conversation involves multiple parties (such as a multi-person chat), each party's single speech counts as a turn, and the turns are accumulated sequentially.

[0049] If the number of conversation turns in the current conversation is greater than or equal to a first preset threshold, it indicates that the conversation has progressed to a certain depth (the number of turns has been met). More conversation turns generally indicate a deeper level of discussion (e.g., problem analysis, detailed discussion, multiple rounds of deliberation, etc.). By setting a first preset threshold (e.g., 30 or 50 rounds), conversations with a certain depth of discussion can be automatically filtered out, avoiding the ineffective storage of fragmented, shallow conversations without requiring manual user intervention. For example, in a customer service conversation scenario, the conversation archiving process is automatically triggered only after a user and customer service representative have engaged in multiple rounds of communication on a particular issue, ensuring the timeliness and effectiveness of the content being memorized. This significantly reduces the user's operational burden, avoids the omission of key information due to user forgetfulness or failure to actively trigger commands, and improves the integrity of memory.

[0050] 2. The total number of conversations initiated within a preset time period is greater than or equal to a second preset threshold value, where the preset time period includes the time corresponding to the current conversation.

[0051] The preset time period may be a time period between the current moment corresponding to the current conversation and a first moment before the current moment, and / or a time period between the current moment corresponding to the current conversation and a second moment after the current moment. That is, the preset time period may be the first time period before the current conversation (including the current moment), the second time period after the current conversation (including the current moment), or both the first and second time periods. The specific setting may be based on actual needs and is not specifically limited in this disclosure.

[0052] If a user frequently initiates conversations over a period of time, this indicates a high-frequency conversation scenario, and the current conversation can be proactively archived. For example, if the total number of conversations initiated by a user within a certain period (e.g., one day, one week) exceeds a second preset threshold (e.g., 10 or 20 times), this typically indicates a high-frequency or ongoing need for interaction (e.g., frequent consultations, feedback on repetitive questions, etc.). Memorizing the current conversations at this time can accurately identify the user's most frequently discussed topics, avoiding information fragmentation caused by the limitations of single-conversation memory.

[0053] 3. The current conversation contains preset keywords.

[0054] The preset keyword may be a user-defined word pre-set by the user, and may include multiple keywords. When the preset keyword appears in the conversation content, real-time archiving may be triggered.

[0055] Furthermore, taking into account the randomness of a single occurrence of a preset keyword, in order to improve the accuracy, reliability and robustness of decision-making and reduce the interference of accidental factors, real-time archiving can also be triggered when the conversation content of the current conversation contains a third preset number threshold of preset keywords.

[0056] For different preset keywords, you can also set their corresponding weight values. Based on the preset keywords contained in the conversation content of the current conversation, determine the total weight value corresponding to the preset keywords. When the total weight value is greater than or equal to the preset weight threshold, real-time archiving is triggered.

[0057] In addition, if the user explicitly states in the conversation interaction that this conversation does not need to be remembered, even if the current conversation meets the automatic archiving conditions, there is no need to remember and archive the current conversation, that is, the interaction record of the current conversation is excluded from the model memory range.

[0058] In step S102, the conversation data information of the current conversation is generated according to the conversation content.

[0059] In this step, if the current conversation is determined to meet the automatic archiving conditions, the conversation data information of the current conversation can be extracted based on the conversation content. This conversation data information can be a complete record of the conversation content, meaning that in one possible implementation, the conversation data information includes the conversation content. To reduce storage requirements and facilitate subsequent data retrieval, the conversation content can be refined to extract key conversation information, meaning that in another possible implementation, the conversation data information includes key conversation information.

[0060] In step S103, the conversation data information is archived into the target database.

[0061] After generating the conversation data, it can be archived to a target database. The target database can be a local database. For example, storage can be performed on a daily basis. For example, starting with the first day of use, all conversation data recorded on that day will be stored in the memsav0 file. The conversation data for the second day will be stored in the memsav1 file, and so on, resulting in daily conversation data. Of course, storage can also be performed on a weekly, monthly, or other statistical period basis to facilitate subsequent retrieval based on time. The above description is merely illustrative and is not specifically limited in this disclosure.

[0062] By adopting the above method and the automated archiving process provided by the present disclosure, the limitations of traditional general-purpose large-scale model memory storage are effectively resolved. First, the conversation content that meets the conditions is automatically obtained, and the active capture of information is achieved when the automatic archiving conditions are met, thereby improving the timeliness and completeness of memory storage. Secondly, the conversation data information is generated, and the original conversation content can be converted into structured data to facilitate subsequent retrieval and analysis. Finally, the generated conversation data information is archived to the target database, giving the model the ability to automatically archive, allowing information to be deposited and managed in an orderly manner, greatly enhancing the efficiency of information management and the depth and breadth of model knowledge deposition, and expanding the application value of large models in information management scenarios. In this way, automatic archiving can be achieved for deep conversations, high-frequency conversations, and conversations with key content, greatly simplifying user operations and improving user experience.

[0063] The following specifically describes the implementation of generating the conversation data information of the current conversation based on the conversation content in step S102.

[0064] In a possible implementation, the conversation content may be directly used as the conversation data information of the current conversation.

[0065] The conversation content may include, for example, conversation text, conversation images, conversation voice, conversation files, generated code snippets, etc. In this implementation, the conversation text, conversation images, conversation voice, conversation files, generated code snippets, etc. may be packaged into a .memzip file in a unified format and archived in the target database.

[0066] In addition, the encapsulated conversation content can be encrypted and archived in the target database to further ensure data security.

[0067] In another possible implementation, conversation key information may be extracted from the conversation content, and the conversation data information may be generated based on the conversation key information.

[0068] The conversation content can include, for example, multimodal content such as text, images, audio, files, and generated code snippets. Different types of content can be parsed separately to extract key information. First, non-text content can be converted into corresponding text. For example, for image-based content, OCR (Optical Character Recognition) technology can be used to accurately identify text within images, converting visual content such as scanned documents, handwritten notes, and screenshots into editable text. For audio-based content, ASR (Automatic Speech Recognition) technology can be used to convert the vocabulary in human speech into computer-readable text. Second, for text-based content, NLP (Natural Language Processing) technology can be used to extract the core content of the text, generating an overview of the core conversation content and obtaining key information about the conversation.

[0069] In some embodiments, key conversation information can be directly used as conversation data. Furthermore, considering the storage pressure of the target database, key conversation information can be encapsulated and used as the conversation data. For example, hash deduplication (specifically, a unique hash value (e.g., MD5, SHA-256) can be generated for conversation content (e.g., text, images, file links, etc.). By comparing the hash values, duplicates between different conversation segments can be quickly determined. For example, if two text segments are completely identical, their hash values ​​will be identical, automatically filtering out redundant records) and block thumbnails (specifically, for text content, keywords or summaries can be extracted by semantic block (e.g., by sentence or paragraph) to generate lightweight information while retaining the conversation role identifiers (user / model). For multimedia content, thumbnails or feature vectors (e.g., audio MFCC features, video keyframes) are generated for images, audio, and video, and stored instead of the original files) can be used to dedupe and compress the key conversation information. The compressed key conversation information can then be encapsulated (e.g., into a unified .memzip file) and archived in the target database. This upgrades conversational information management from raw data accumulation to intelligent knowledge accumulation, preventing duplicate storage. This reduces storage and computing costs while improving data availability, providing the underlying technical support for large models' long-term memory capabilities and decision support. It also overcomes the limitation of existing technologies that can only store text-based content.

[0070] In addition, the encapsulated key information of the conversation can be encrypted and the encrypted conversation content can be archived in the target database to further ensure data security.

[0071] In other embodiments, in order to facilitate the efficiency of subsequent retrieval, the timestamp information and conversation scene information (such as the conclusion of the project meeting) corresponding to the current conversation can also be determined based on the conversation content, and the conversation data information can be generated based on the timestamp information, the conversation scene information and the conversation key information.

[0072] Specifically, the timestamp information and the conversation scene information can be associated with the conversation key information, and the conversation data information can be obtained based on the associated timestamp information, conversation scene information, and conversation key information. In other words, the timestamp information and conversation scene information are used as index tags for the current conversation, and are associated with the conversation key information, and together are stored as conversation data information in the target database.

[0073] In this way, each conversation can be judged in the above manner to determine whether it meets the automatic archiving conditions. If it is determined that the automatic archiving conditions are met, the corresponding conversation data information is generated and stored in the target database. The conversation data information corresponding to multiple conversations can be obtained, which changes the passive mode that requires users to explicitly input instructions, realizes active capture of information, and improves the timeliness and completeness of memory storage.

[0074] By adopting the above method and the automated archiving process provided by the present disclosure, the limitations of traditional general-purpose large-scale model memory storage are effectively resolved. First, the conversation content that meets the conditions is automatically acquired, changing the passive mode that requires the user to explicitly input instructions. When the automatic archiving conditions are met, the information is actively captured, which improves the timeliness and completeness of memory storage. Secondly, the conversation data information is generated, which can convert the original conversation content into structured data for subsequent retrieval and analysis. Finally, the conversation data information is archived to the target database, giving the system the ability to automatically archive, allowing information to be deposited and managed in an orderly manner, greatly enhancing the efficiency of information management and the depth and breadth of model knowledge deposition, and expanding the application value of large models in information management scenarios. In this way, automatic archiving can be achieved for deep conversations, high-frequency conversations, and conversations with key content, greatly simplifying user operations and improving user experience.

[0075] Figure 2 FIG. 1 is a flow chart showing a method for retrieving data according to an exemplary embodiment. Figure 2 As shown, the method may include the following steps: In step S201, input information input by the user is obtained.

[0076] The input information can be data retrieval text initiated by the user (for example, "Find last year's discussion on cost reduction and efficiency improvement"), question text when the user is talking to the model, or an input image input by the user. When the input information is an input image, the input image can be recognized first to obtain the image text information therein.

[0077] In step S202, target data information is acquired from a plurality of conversation data information stored in a target database according to the input information.

[0078] Among them, the conversation data information in the target database is based on Figure 1 The data provided is archived using the archiving method.

[0079] In this step, the input information may be matched with the plurality of conversation data information, and the successfully matched conversation data information may be used as the target data information.

[0080] For example, if the input information includes "find discussions about cost reduction and efficiency improvement last year", the model will return the associated target data information after retrieval, such as conversation records, uploaded file summaries and corresponding timelines.

[0081] In some embodiments, natural language input is supported (e.g., "Find all communication records with customer Zhang San"), lowering the user barrier to entry and eliminating the need to memorize complex search syntax. Furthermore, combined with large model capabilities, semantic understanding of input information can be performed (e.g., expanding "equipment failure" to synonyms such as "equipment anomaly, error, maintenance record"), improving search hit rates.

[0082] In step S203, the target data information is output.

[0083] Using this method, user input can be used to quickly filter matching target data from the massive amount of conversation data in the target database, rather than loading all conversation data. This reduces memory usage and computational overhead. This allows the right information to be accurately retrieved at the right time, reducing labor costs. This method meets user needs through natural interaction, embracing real-life work and life scenarios. It transforms conversation data into a productive factor that drives decision-making, empowering large models with enhanced memory and reasoning capabilities. This allows for automatic archiving of deep, high-frequency, and key conversations, significantly simplifying user operations and improving the user experience.

[0084] The above step S202 is described in detail below.

[0085] In a possible implementation, the input information includes data retrieval text, and the conversation data information includes conversation key information. The data retrieval text can be matched with the multiple conversation key information, and the successfully matched conversation key information is used as the target data information.

[0086] For example, the data retrieval text and the conversation key information may be matched for similarity, and the conversation key information with the highest similarity may be used as the target data information.

[0087] In some embodiments, the data retrieval text includes the conversation retrieval time and / or the conversation retrieval scenario. The conversation data information also includes timestamp information and conversation scenario information corresponding to the key information of the conversation. The conversation retrieval time can be a time point or a time interval, thereby implementing a bidirectional index of time points and time intervals. A two-layer index structure is constructed based on time points (accurate to the minute) and time intervals (such as "Q1 2023"), while also supporting fuzzy queries (such as "marketing plan discussed last week"). Specifically, semantic analysis of the data retrieval text can be performed to obtain the corresponding conversation retrieval time and / or conversation retrieval scenario.

[0088] In a case where the data retrieval text includes a conversation retrieval time, matching the data retrieval text with the plurality of conversation key information includes: matching the conversation retrieval time with timestamp information corresponding to the plurality of conversation key information.

[0089] If the conversation retrieval time is a point in time, the timestamp information corresponding to the point in time can be filtered from multiple timestamps based on the conversation retrieval time, and a match is considered successful. If the conversation retrieval time is a time interval, the timestamp information within the time interval can be filtered from multiple timestamps based on the conversation retrieval time, and a match is considered successful.

[0090] Traditional conversation models only store interaction content and lack structured annotations of the time dimension (such as timestamps and time interval indexes). In this embodiment, precise timestamp information (such as "2024-01-15 14:30:00") is added to each conversation data entry, and key time features (such as year, quarter, and hour) are extracted as metadata. Furthermore, archived .memzip files are indexed by time granularity (day, week, or month) to enable rapid retrieval of all interaction records within a specific time period. For example, if a user enters "project decision for Q1 2024," the model automatically retrieves all relevant conversations and their inference logic for that month.

[0091] In the case that the data retrieval text includes a conversation retrieval scenario, matching the data retrieval text with the plurality of conversation key information includes: matching the conversation retrieval scenario with conversation scenario information corresponding to the plurality of conversation key information.

[0092] Considering that the dialogue retrieval scenario input by the user is not standardized enough, resulting in a low matching success rate with the dialogue scenario information, in order to improve the possibility of successful matching, the semantics of the dialogue retrieval scenario and the dialogue scenario information can be analyzed. If the two have the same meaning, the match can be considered successful.

[0093] In the case where the data retrieval text includes a conversation retrieval scene and a conversation retrieval time, matching the data retrieval text with the multiple conversation key information includes: matching the conversation retrieval scene with the conversation scene information corresponding to the multiple conversation key information, and matching the conversation retrieval time with the timestamp information corresponding to the multiple conversation key information.

[0094] For example, the time interval indexing algorithm can be expressed as: def time_index_query(start_time, end_time, context_tags): #Joint retrieval based on time range and scene tags memories = TemporalIndex.search( timestamp>= start_time, timestamp<= end_time, tags=context_tags ) In another possible implementation, the input information includes a question text, and the conversation data includes conversation key information. The user's question intent can be determined based on the question text. The question intent is then matched with multiple pieces of conversation key information, and the successfully matched conversation key information is used as the target data information.

[0095] In this scenario, the user may not actively initiate the retrieval of historical conversation data, but in order to avoid the user from repeatedly explaining the background and improve the continuous interactive experience, the user's current question text can be analyzed to obtain the user's current question intention, and based on the question intention, the target data information can be matched from multiple conversation key information.

[0096] Similarly, if the analyzed question intent contains time information or scene information, a matching method similar to that of data retrieval text can be used to match the target data information, which will not be elaborated here.

[0097] Furthermore, based on the question text, an answer text corresponding to the question text is generated. The answer text is the answer content generated by the model based on the question text input by the current user. At the same time, when the model generates the answer, historical decision logic (i.e., generating suggestion prompt text) can also be automatically injected. Specifically, outputting the target data information includes: generating suggestion prompt text based on the answer text and the target data information, and outputting the answer text and the suggestion prompt text. For example, a suggestion prompt text can be generated based on the similarity between the answer text and the target data information. If the similarity between the answer text and the target data information is greater than or equal to a preset similarity threshold, a suggestion prompt text is generated based on the user's historical decision corresponding to the target data information. The user's historical decision is used to characterize the user's feedback decision on the target data information in the historical conversation. The feedback decision may, for example, include actions such as selection, evaluation, and judgment.

[0098] For example, conversation data can be stored in a "time-scenario" format. For example: Time: 2024-01-10 | Decision: Reject Plan A | Reason: Cost Exceedance (20% Budget Overrun) | Basis: Comparison with Historical Cost Data. Using the model's semantic understanding capabilities, the model calculates the similarity between the user's current input (e.g., "Do you accept Plan B?") and the historical decisions (e.g., "Reject Plan A") in the extracted target data. This calculation is then used to calculate the similarity between Plan B and Plan A. Based on this comparison, a suggestion prompt is generated. When the user proposes a similar plan again, the system prompts, "In January 2024, you rejected a similar plan due to cost overruns."

[0099] That is to say, when generating a response text, the reasoning logic in historical interactions can be automatically associated based on the user input information. Specifically, if the semantic similarity between the current conversation and the historical decision exceeds a preset threshold (such as 80%), the memory archive file (i.e., the target data information) of the corresponding time interval will be automatically retrieved, and the difference in the reasoning logic of the two will be compared. By deeply analyzing the differences between the current conversation and historical decisions, a suggestion prompt text will be generated and output together with the response text to optimize the user's continuous interaction experience. By incorporating the time dimension to reconstruct the reasoning chain and strengthening the model's COT (Chain-of-Thought) capability, the accuracy of the model's cross-session reasoning is significantly improved, allowing users to obtain more logical and reference-worthy feedback during the interaction process.

[0100] By adopting the above method and the automated archiving process provided by the present disclosure, the limitations of traditional general-purpose large-scale model memory storage are effectively resolved. First, the conversation content that meets the conditions is automatically acquired, changing the passive mode that requires the user to explicitly input instructions. When the automatic archiving conditions are met, the information is actively captured, which improves the timeliness and completeness of memory storage. Secondly, the conversation data information is generated, which can convert the original conversation content into structured data for subsequent retrieval and analysis. Finally, the conversation data information is archived to the target database, giving the system the ability to automatically archive, allowing information to be deposited and managed in an orderly manner, greatly enhancing the efficiency of information management and the depth and breadth of model knowledge deposition, and expanding the application value of large models in information management scenarios. In this way, automatic archiving can be achieved for deep conversations, high-frequency conversations, and conversations with key content, greatly simplifying user operations and improving user experience.

[0101] Figure 3 FIG. 1 is a schematic diagram of a data processing flow according to an exemplary embodiment. Figure 3 As shown, first, a user initiates a conversational interaction. The model's memory agent (intelligent agent) monitors the current conversation. It then determines whether the current conversation meets the automatic archiving criteria. If so, it retrieves the conversation content and extracts conversation data based on the content. This process supports processing conversation content in various data types, including text, audio, and video. The content is then compressed and encapsulated using the memory compression and encapsulation module and archived to the target database. If the automatic archiving criteria are not met, the process returns to determine whether the current conversation meets the automatic archiving criteria. Furthermore, in application scenarios, when a user initiates a search request, relevant memory searches can be performed using methods such as NLP to obtain the target data. This target data is then used to enhance the ability to process historical events through thought chains. By automatically incorporating historical decision logic into the generated response, the system outputs a response text plus suggested prompts. This avoids the user having to repeatedly explain the context and enhances the continuous interaction experience.

[0102] Figure 4 FIG. 1 is a block diagram of a data archiving device according to an exemplary embodiment. Figure 4 As shown, the device 300 includes: The first acquisition module 301 is configured to acquire the conversation content of the current conversation if the current conversation meets the automatic archiving conditions; A generating module 302 is configured to generate conversation data information of the current conversation based on the conversation content; The archiving module 303 is configured to archive the conversation data information into a target database; The automatic archiving conditions include any of the following: The number of conversation turns in the current conversation is greater than or equal to a first preset number threshold; The total number of conversations initiated within a preset time period is greater than or equal to a second preset number threshold, the preset time period including the time corresponding to the current conversation; The conversation content of the current conversation contains preset keywords.

[0103] Optionally, the generating module 302 is configured to extract conversation key information from the conversation content; and generate the conversation data information according to the conversation key information.

[0104] Optionally, the generating module 302 is configured to encapsulate the conversation key information and use the encapsulated conversation key information as the conversation data information.

[0105] Alternatively, as Figure 5 As shown, the device 300 further includes: Determination module 304, configured to determine timestamp information and conversation scene information corresponding to the current conversation based on the conversation content; The generating module 302 is configured to generate the conversation data information according to the timestamp information, the conversation scene information and the conversation key information.

[0106] Optionally, the generating module 302 is configured to associate the timestamp information and the conversation scene information with the conversation key information; and obtain the conversation data information according to the associated timestamp information, conversation scene information and conversation key information.

[0107] By adopting the above-mentioned device and the automated archiving process provided by the present disclosure, the limitations of traditional general-purpose large-scale model memory storage are effectively resolved. First, the conversation content that meets the conditions is automatically acquired, changing the passive mode that requires the user to explicitly input instructions. When the automatic archiving conditions are met, the active capture of information is realized, which improves the timeliness and completeness of memory storage. Secondly, the conversation data information is generated, which can convert the original conversation content into structured data for subsequent retrieval and analysis. Finally, the conversation data information is archived to the target database, giving the system the ability to automatically archive, so that information can be deposited and managed in an orderly manner, greatly enhancing the efficiency of information management and the depth and breadth of model knowledge deposition, and expanding the application value of large models in information management scenarios. In this way, automatic archiving can be achieved for deep conversations, high-frequency conversations, and conversations with key content, greatly simplifying user operations and improving user experience.

[0108] Figure 6 is a block diagram of a data retrieval device according to an exemplary embodiment. Figure 6 As shown, the apparatus 400 includes: The second acquisition module 401 is configured to acquire input information input by the user; The third acquisition module 402 is configured to acquire target data information from multiple conversation data information stored in the target database according to the input information; the conversation data information in the target database is based on Figure 1 The archiving method of the data provided is obtained by archiving; The output module 403 is configured to output the target data information.

[0109] Optionally, the third acquisition module 402 is configured to match the input information with the plurality of conversation data information, and use the successfully matched conversation data information as the target data information.

[0110] Optionally, the input information includes data retrieval text, the conversation data information includes conversation key information, and the third acquisition module 402 is configured to match the data retrieval text with the multiple conversation key information and use the successfully matched conversation key information as the target data information.

[0111] Optionally, the data retrieval text includes a conversation retrieval time and / or a conversation retrieval scenario, and the conversation data information further includes timestamp information corresponding to the conversation key information and conversation scenario information. If the data retrieval text includes the conversation retrieval time, the third acquisition module 402 is configured to match the conversation retrieval time with timestamp information corresponding to the plurality of conversation key information. In the case where the data retrieval text includes a conversation retrieval scenario, the third acquisition module 402 is configured to match the conversation retrieval scenario with conversation scenario information corresponding to a plurality of conversation key information; In the case where the data retrieval text includes a conversation retrieval scene and a conversation retrieval time, the third acquisition module 402 is configured to match the conversation retrieval scene with the conversation scene information corresponding to multiple conversation key information, and match the conversation retrieval time with the timestamp information corresponding to multiple conversation key information.

[0112] Optionally, the input information includes question text, the conversation data information includes conversation key information, and the third acquisition module 402 is configured to determine the user's question intention based on the question text; match the question intention with multiple conversation key information, and use the successfully matched conversation key information as the target data information.

[0113] Optionally, the output module 403 is configured to generate a response text corresponding to the question text according to the question text; generate a suggestion prompt text according to the response text and the target data information; and output the response text and the suggestion prompt text.

[0114] The above-mentioned device, through the automated archiving process provided by this disclosure, effectively addresses the limitations of traditional general-purpose large-scale model memory storage. First, it automatically captures conversation content that meets certain conditions, changing the passive model that requires explicit user input. It now actively captures information when the automatic archiving conditions are met, improving the timeliness and completeness of memory storage. Second, it generates conversation data information, converting raw conversation content into structured data for easy subsequent retrieval and analysis. Finally, it archives the conversation data information to the target database, empowering the system with automated archiving capabilities, enabling orderly information accumulation and management. This significantly enhances information management efficiency and the depth and breadth of model knowledge accumulation, expanding the application value of large-scale models in information management scenarios. Using user-entered input, matching target data can be quickly filtered from the massive amount of conversation data in the target database, rather than loading all conversation data, reducing memory usage and computational overhead. This allows the right information to be accurately retrieved at the right time, reducing labor costs. This approach meets user needs through natural interaction, embracing real-world work and life scenarios. It transforms conversation data into a productive factor that drives decision-making, empowering large-scale models with stronger memory and reasoning capabilities. In this way, deep conversations, high-frequency conversations and conversations with key content can be automatically archived, greatly simplifying user operations and improving user experience.

[0115] Regarding the apparatus in the above embodiment, the specific manner in which each module performs operations has been described in detail in the embodiment of the method, and will not be elaborated here.

[0116] Figure 7 FIG. 5 is a block diagram of an electronic device 500 according to an exemplary embodiment. Figure 7 As shown, the electronic device 500 may include: a processor 501 , a memory 502 , and may further include one or more of a multimedia component 503 , an input / output (I / O) interface 504 , and a communication component 505 .

[0117] The processor 501 is used to control the overall operation of the electronic device 500 to complete all or part of the steps in the aforementioned data archiving method or data retrieval method. The memory 502 is used to store various types of data to support the operation of the electronic device 500. This data may include, for example, instructions for any application or method operating on the electronic device 500, as well as application-related data such as contact information, sent and received messages, images, audio, video, etc. The memory 502 may be implemented by any type of volatile or non-volatile storage device, or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk. The multimedia component 503 may include a screen and an audio component. The screen may be, for example, a touch screen, and the audio component is used to output and / or input audio signals. For example, the audio component may include a microphone for receiving external audio signals. The received audio signal may be further stored in the memory 502 or sent through the communication component 505. The audio component also includes at least one speaker for outputting audio signals. The I / O interface 504 provides an interface between the processor 501 and other interface modules. The above-mentioned other interface modules may be a keyboard, a mouse, buttons, etc. These buttons may be virtual buttons or physical buttons. The communication component 505 is used for wired or wireless communication between the electronic device 500 and other devices. Wireless communication, such as Wi-Fi, Bluetooth, Near Field Communication (NFC), 2G, 3G or 4G, or a combination of one or more of them, so the corresponding communication component 505 may include: a Wi-Fi module, a Bluetooth module, an NFC module.

[0118] In an exemplary embodiment, the electronic device 500 may be implemented by one or more application specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field programmable gate arrays (FPGAs), controllers, microcontrollers, microprocessors, or other electronic components to execute the above-mentioned data archiving method or data retrieval method.

[0119] In another exemplary embodiment, a computer-readable storage medium including program instructions is also provided. When executed by a processor, the program instructions implement the steps of the aforementioned data archiving method or data retrieval method. For example, the computer-readable storage medium may be the aforementioned memory 502 including the program instructions. The program instructions may be executed by the processor 501 of the electronic device 500 to implement the aforementioned data archiving method or data retrieval method.

[0120] In another exemplary embodiment, a computer program product is further provided. The computer program product includes a computer program that can be executed by a processor. When the computer program is executed by the processor, the steps of the above-mentioned data archiving method or data retrieval method are implemented.

[0121] In another exemplary embodiment, a computer program product is further provided. The computer program product includes a computer program that can be executed by a processor. When the computer program is executed by the processor, the steps of the above-mentioned data archiving method or data retrieval method are implemented.

[0122] The preferred embodiments of the present disclosure are described in detail above in conjunction with the accompanying drawings. However, the present disclosure is not limited to the specific details of the above embodiments. Within the technical concept of the present disclosure, various simple modifications can be made to the technical solutions of the present disclosure, and these simple modifications all fall within the scope of protection of the present disclosure.

[0123] It should also be noted that the various specific technical features described in the above specific embodiments can be combined in any appropriate manner without contradiction. In order to avoid unnecessary repetition, the present disclosure will not further describe various possible combinations.

[0124] In addition, the various embodiments of the present disclosure may be arbitrarily combined, and as long as they do not violate the concept of the present disclosure, they should also be regarded as the contents disclosed by the present disclosure.

Claims

1. A data archiving method, characterized in that: The method comprises: When the current conversation meets the automatic archiving conditions, obtain the conversation content of the current conversation; Generating conversation data information of the current conversation according to the conversation content; Archiving the conversation data information into a target database; The automatic archiving conditions include any one of the following: The number of conversation turns in the current conversation is greater than or equal to a first preset number threshold; The total number of conversations initiated within a preset time period is greater than or equal to a second preset number threshold, where the preset time period includes the time corresponding to the current conversation; The conversation content of the current conversation contains preset keywords.

2. The method according to claim 1, characterized in that Generating the conversation data information of the current conversation according to the conversation content includes: extracting key conversation information from the conversation content; The conversation data information is generated according to the conversation key information.

3. The method according to claim 2, characterized in that Generating the conversation data information according to the conversation key information includes: The conversation key information is encapsulated, and the encapsulated conversation key information is used as the conversation data information.

4. The method according to claim 2, characterized in that The method further comprises: Determining the timestamp information and conversation scene information corresponding to the current conversation based on the conversation content; Generating the conversation data information according to the conversation key information includes: The conversation data information is generated according to the timestamp information, the conversation scene information and the conversation key information.

5. The method according to claim 4, characterized in that The generating the conversation data information according to the timestamp information, the conversation scene information, and the conversation key information includes: Associating the timestamp information and the conversation scene information with the conversation key information; The conversation data information is obtained according to the associated timestamp information, conversation scene information and conversation key information.

6. A data retrieval method, characterized in that: The method comprises: Get the input information entered by the user; According to the input information, target data information is obtained from a plurality of conversation data information stored in a target database; the conversation data information in the target database is archived according to the data archiving method according to any one of claims 1 to 5; Output the target data information.

7. The method according to claim 6, characterized in that The acquiring target data information from the plurality of conversation data information stored in the target database according to the input information includes: The input information is matched with the plurality of conversation data information, and the successfully matched conversation data information is used as the target data information.

8. The method according to claim 7, characterized in that The input information includes data retrieval text, the conversation data information includes conversation key information, and the matching of the input information with the plurality of conversation data information and using the successfully matched conversation data information as the target data information includes: The data retrieval text is matched with the plurality of conversation key information, and the successfully matched conversation key information is used as the target data information.

9. The method according to claim 8, characterized in that The data retrieval text includes a conversation retrieval time and / or a conversation retrieval scene, and the conversation data information further includes timestamp information and conversation scene information corresponding to the conversation key information. When the data retrieval text includes the conversation retrieval time, matching the data retrieval text with the plurality of conversation key information includes: Matching the conversation retrieval time with timestamp information corresponding to the plurality of conversation key information; In a case where the data retrieval text includes a conversation retrieval scenario, matching the data retrieval text with the plurality of conversation key information includes: Matching the conversation retrieval scene with the conversation scene information corresponding to the plurality of conversation key information; In a case where the data retrieval text includes a conversation retrieval scene and a conversation retrieval time, matching the data retrieval text with the plurality of conversation key information includes: The conversation retrieval scene is matched with the conversation scene information corresponding to the multiple conversation key information, and the conversation retrieval time is matched with the timestamp information corresponding to the multiple conversation key information.

10. The method according to claim 6, characterized in that The input information includes question text, the dialogue data information includes dialogue key information, and acquiring target data information from a plurality of dialogue data information stored in a target database according to the input information includes: Determining the user's question intention based on the question text; Match the question intention with the multiple conversation key information, and use the successfully matched conversation key information as the target data information.

11. The method according to claim 10, characterized in that The method further comprises: Generate a response text corresponding to the question text according to the question text; The outputting of the target data information includes: Generate a suggestion prompt text according to the response text and the target data information; The response text and the suggestion prompt text are output.

12. A data archiving device, characterized in that: The device comprises: A first acquisition module is configured to acquire the conversation content of the current conversation when the current conversation meets the automatic archiving condition; A generating module configured to generate conversation data information of a current conversation based on the conversation content; an archiving module, configured to archive the conversation data information into a target database; The automatic archiving conditions include any one of the following: The number of conversation turns in the current conversation is greater than or equal to a first preset number threshold; The total number of conversations initiated within a preset time period is greater than or equal to a second preset number threshold, where the preset time period includes the time corresponding to the current conversation; The conversation content of the current conversation contains preset keywords.

13. A data retrieval device, characterized in that: The device comprises: A second acquisition module is configured to acquire input information input by a user; a third acquisition module configured to acquire target data information from a plurality of conversation data information stored in a target database according to the input information; the conversation data information in the target database is archived according to the data archiving method according to any one of claims 1 to 5; The output module is configured to output the target data information.

14. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the steps of the data archiving method according to any one of claims 1 to 5 are implemented, or the steps of the data retrieval method according to any one of claims 6 to 11 are implemented.

15. An electronic device, characterized in that: include: a memory having a computer program stored thereon; A processor, configured to execute the computer program in the memory to implement the steps of the data archiving method according to any one of claims 1 to 5, or to implement the steps of the data retrieval method according to any one of claims 6 to 11.

16. A computer program product comprising a computer program, characterized in that When the computer program is executed by a processor, the computer program implements the steps of the data archiving method according to any one of claims 1 to 5, or implements the steps of the data retrieval method according to any one of claims 6 to 11.

Images