A question content conversion method, a terminal and a storage medium

Abstract

This invention discloses a method, terminal, and storage medium for converting question content. The method includes: analyzing user-submitted question content; if the question content involves a preset domain multiple times (reaching a threshold), assigning a role tag corresponding to the preset domain to the user; analyzing the conversation content between the user and the user; if the conversation content conforms to preset conversation rules, assigning a conversation tag corresponding to the preset conversation rules to the conversation content; if the conversation content involves periodic keywords, assigning a periodic tag corresponding to the periodic keywords to the conversation content; and fusing the role tag, the conversation tag, and the periodic tag into the user-submitted question content to obtain the converted question content. This invention, through a three-layer tagging system and intelligent question content optimization, achieves automatic and accurate extraction of user intent and context, significantly reducing the computational burden of large models and effectively improving response speed and answer accuracy under high concurrency.

Description

Technical Field

[0001] This invention relates to the fields of human-computer interaction and natural language processing technology, and in particular to a method, terminal, and storage medium for converting question content. Background Technology

[0002] With the rapid development of artificial intelligence technology, automated question-answering systems based on large language models have been widely used in scenarios such as intelligent customer service and virtual assistants. In these systems, in order to provide coherent and personalized answers, it is usually necessary to input the user's current question and its historical dialogue content into the large language model for processing.

[0003] However, this approach suffers from significant efficiency bottlenecks. Large language models need to process the entire dialogue history, which contains a large amount of irrelevant or redundant information, consuming enormous computing resources and time, leading to response latency. In high-concurrency scenarios, resource contention intensifies further, severely impacting user experience and system throughput. Furthermore, blindly relying on the entire historical context may interfere with the model's judgment of the current core intent.

[0004] Therefore, a method is needed to intelligently extract key contextual information without sacrificing the quality of responses, so as to reduce the computational burden of large models and improve the overall interaction efficiency. Summary of the Invention

[0005] The technical problem to be solved by the present invention is to provide a method, terminal and storage medium for converting question content, so as to solve the technical problems of existing intelligent question answering systems, such as high consumption of computing resources, high response latency and limited system performance in high-concurrency scenarios due to processing complete dialogue history.

[0006] To solve the above-mentioned technical problems, the technical solution adopted by the present invention is as follows: A method for transforming question content, including: Analyze the questions submitted by users. If the number of times the questions involve a preset field reaches a threshold, then assign a role tag corresponding to the preset field to the user. Analyze the conversation content with the user; if the conversation content conforms to the preset conversation rules, then assign the conversation content a conversation tag corresponding to the preset conversation rules. If the conversation content involves periodic keywords, then assign periodic tags corresponding to the periodic keywords to the conversation content; The user-submitted question content is combined with the role tag, the conversation tag, and the cycle tag to obtain the transformed question content.

[0007] To solve the above-mentioned technical problems, another technical solution adopted by the present invention is: a terminal for converting question content, including a memory, a processor, and a computer program stored in the memory and running on the processor, wherein the processor executes the computer program to implement the steps of the method described above.

[0008] To solve the above-mentioned technical problems, another technical solution adopted by the present invention is: a computer-readable storage medium storing a computer program thereon, wherein the computer program, when executed by a processor, implements the various steps of the method described above.

[0009] The beneficial effects of this invention are as follows: By constructing a three-level tagging system of role tags, conversation tags, and periodic tags, key contextual information is extracted from user questioning behavior and conversation history, and this information is used to intelligently optimize the user's original questions. This method avoids directly inputting lengthy original dialogue history into a large model. Instead, it integrates the extracted key tag information into the questions, enabling the large model to focus on the most relevant context for understanding and answering. This significantly reduces the amount of irrelevant data that the large model needs to process, lowers computational resource consumption and response latency, thereby effectively improving the system's question-answering efficiency and concurrent processing capabilities. Attached Figure Description

[0010] Figure 1 This is a flowchart illustrating a method for converting question content according to an embodiment of the present invention; Figure 2 This is a schematic diagram of a terminal for converting question content according to an embodiment of the present invention; Label Explanation: 1. A terminal for converting question content; 2. A processor; 3. A memory. Detailed Implementation

[0011] To explain in detail the technical content, objectives, and effects of the present invention, the following description is provided in conjunction with the embodiments and accompanying drawings.

[0012] Before detailing the embodiments of this application, some related concepts will first be explained: Role tags: These are tags assigned to users based on specific areas (such as "bank operations" or "healthcare") that they repeatedly ask over a period of time, and are used to identify their primary areas of interest or identity.

[0013] Conversation tags: These are tags assigned to conversation content based on discussions around a specific topic within a series of consecutive or related multi-turn conversations, and are used to associate the content with that topic.

[0014] Periodic tags: These are tags assigned to conversation content within a preset unit period (such as a day, a week, or a month) based on repeated questions from users about specific periodic events or keywords (such as "monthly report"). These tags are only valid within the corresponding period.

[0015] Large Language Model (LLM): refers to a pre-trained language model with a huge number of parameters, usually containing billions or even more parameters, capable of performing a variety of natural language processing tasks.

[0016] In existing technologies, intelligent question-answering systems typically concatenate the user's current question with historical dialogue records before inputting it into a large model to maintain the consistency of the response. However, this method has significant drawbacks: as the number of dialogue rounds increases, the model needs to repeatedly process the ever-growing amount of contextual data, consuming significant computational resources and causing response delays, which is particularly pronounced in high-concurrency scenarios. Furthermore, not all historical information is relevant to the current question; the inclusion of irrelevant content not only increases the processing burden but may also interfere with the model's accurate understanding of the user's true current intent, thus affecting the quality of the response.

[0017] To at least solve the above problems, please refer to Figure 1 This invention provides a method for converting question content, including: S1. Analyze the questions submitted by users. If the number of times the questions involve the preset domain reaches the threshold, assign the user a role tag corresponding to the preset domain. S2. Analyze the conversation content with the user. If the conversation content conforms to the preset conversation rules, assign a conversation tag corresponding to the preset conversation rules to the conversation content. S3. If the conversation content involves periodic keywords, then assign periodic tags corresponding to the periodic keywords to the conversation content; S4. Integrate the role tag, the conversation tag, and the period tag into the question content submitted by the user to obtain the converted question content.

[0018] As described above, the beneficial effects of this invention are as follows: Through steps S1 to S3, the system constructs a dynamic three-layer user tagging system (role tag, conversation tag, and periodic tag). These tags are not arbitrarily set, but are automatically generated through real-time analysis of user questioning patterns and conversation content; they represent the most relevant contextual features to the user. Step S4 integrates these highly refined feature tags into the user's new questions in a structured manner, forming a richer and more accurate enhanced prompt. After receiving this enhanced prompt, the large model no longer needs to painstakingly parse the complex original dialogue records, but directly obtains the key information of "who the user is (role), what they have recently discussed (conversation), and what periodic matters they are currently concerned about (period)." This is equivalent to pre-completing the key contextual summarization work for the large model, enabling it to quickly focus, thereby providing answers that fit the user's background and current needs while reducing the amount of data processing, ultimately achieving the goal of improving the system's question-and-answer efficiency.

[0019] Furthermore, the procedure prior to step S1 includes: S01. If a user logs in, the user's registration number at the time of login shall be used as the user's unique identifier. S02. If the user is not logged in, a unique device identifier is assigned to the electronic device used by the user, and the device identifier is used as the user's unique identifier. S03. Generate and maintain the corresponding role tag based on the unique identifier.

[0020] As described above, steps S01 to S03 enable the system to uniquely and continuously track both logged-in users and anonymous visitors using user ID or device ID, which forms the basis for dynamically generating and maintaining user-specific tags.

[0021] Further, step S1 includes: Collect and analyze the questions submitted by users consecutively within a preset time period; Identify the matching status between the continuously submitted questions and keywords in a preset domain; When the number of matches reaches the threshold, a role tag corresponding to the preset domain is assigned to the user.

[0022] As described above, the generation of role tags is based on the user's continuous behavioral patterns over a period of time, rather than a single question. This avoids misclassification due to a single accidental question, and thus can more accurately depict the user's long-term or main areas of interest and identity characteristics.

[0023] Further, step S2 includes: Analyze the content of multi-turn conversations with users. When the proportion of the content of the multi-turn conversations involving the same topic exceeds a preset proportion threshold, the frequency of the content of the multi-turn conversations involving the same topic exceeds a preset first frequency threshold, and the time span of the multi-turn conversations is less than a preset time span, assign corresponding conversation tags to the content of the multi-turn conversations.

[0024] As described above, the rules for generating conversation tags comprehensively consider factors such as the consistency of topics in multi-turn conversations, conversation frequency, and time span. These rules aim to extract the core topics from continuous dialogue, achieving accurate extraction of the conversation context.

[0025] Furthermore, step S2 also includes: Based on the frequency with which the content of the multi-round conversations revolves around the same topic, if the frequency exceeds a preset second frequency threshold, the conversation tag is determined to be a high-frequency conversation tag. If the frequency does not exceed the preset second frequency threshold, the session tag is determined to be a low-frequency session tag; When entering the next preset unit cycle, the high-frequency session tag is migrated to the next preset unit cycle, and the status of the high-frequency session tag is set to valid; At the same time, the status of the low-frequency session tag is set to invalid.

[0026] As described above, this step enables dynamic lifecycle management of conversation tags. By distinguishing between high-frequency and low-frequency tags and performing periodic migration operations on high-frequency tags, the system can automatically eliminate low-relevance topics and extend the lifecycle of high-relevance topics, thereby ensuring that the conversation tag system is always synchronized with the user's latest topics of interest.

[0027] Further, step S3 includes: Within a preset unit period, monitor the number of times users ask questions about the periodic keywords; When the number of questions reaches a preset question threshold, a periodic tag related to the periodic keyword is assigned to the user.

[0028] As described above, the periodic tags aim to capture users' regular needs within a specific time period. Their generation is also based on statistical analysis of question frequency, used to identify the key content that users focus on within the period.

[0029] Furthermore, step S3 also includes: Within the period corresponding to the period tag, the state of the period tag is set to valid; At the end of the cycle, the status of the cycle tag is set to invalid.

[0030] As described above, this step assigns a clear time-sensitive attribute to the periodic label. This mechanism ensures that the label is only valid within the relevant period, preventing expired information from interfering with the subsequent question-and-answer process, thereby guaranteeing the timeliness and accuracy of the question-and-answer content.

[0031] Further, step S4 includes: Obtain valid role tags, session tags, and cycle tags for the user's current state; The question content is combined with the content of the role tags, conversation tags, and cycle tags in a structured format to obtain the transformed question content.

[0032] As described above, step S4 is the core of the question content transformation process. The system obtains the most relevant set of tags through validity filtering and merges its content with the original question in a structured manner. The new question generated by this fusion directly provides the large language model with clear and concise key contextual information, greatly reducing the complexity of the model in parsing the dialogue background and guiding it to make efficient and accurate responses.

[0033] Please refer to Figure 2 Another embodiment of the present invention provides a terminal for converting question content, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the various steps of the above-described question content conversion method.

[0034] The method, terminal, and storage medium for converting question content described above are applicable to internet application scenarios such as intelligent customer service systems, personalized AI assistants, and online education Q&A platforms, which require handling multi-turn dialogues and have high requirements for response speed and personalized experience. The following detailed implementation methods illustrate these methods: Please refer to Figure 1 One embodiment of the present invention is as follows: A method for transforming question content, including: S01. If a user logs in, the user's registration number at the time of login shall be used as the user's unique identifier. S02. If the user is not logged in, a unique device identifier is assigned to the electronic device used by the user, and the device identifier is used as the user's unique identifier. S03. Generate and maintain the corresponding role tag based on the unique identifier.

[0035] Specifically, it is applied to the bank's internal intelligent data analysis assistant system, which assists bank operations staff in completing daily data query and analysis tasks.

[0036] Bank employee A logs into the system using their office computer, and the system uses their employee number "A001" as a unique identifier. If not logged in (e.g., for a temporary query), the system generates a unique device identifier for their device as the user identifier. Based on this unique identifier, the system initializes and maintains the user's tag system.

[0037] S1. Analyze the questions submitted by users. If the number of times the questions relate to a preset domain reaches a threshold, assign the user a role tag corresponding to the preset domain, including: Collect and analyze the questions submitted by users consecutively within a preset time period; Identify the matching status between the continuously submitted questions and keywords in a preset domain; When the number of matches reaches the threshold, a role tag corresponding to the preset domain is assigned to the user.

[0038] Specifically, A is a bank operations employee. The system analyzes all questions A submitted within a preset time period (e.g., the past month), identifying the match between the question content and preset keywords in the banking operations field (e.g., "deposit targets," "loan approval," "operating costs," etc.), and counting the number of matches. When the number of questions related to the "bank operations" field reaches a preset threshold (e.g., 30 times), the system automatically assigns A the "bank operations employee" role tag, which serves as a long-term valid identifier. This tag accurately reflects A's professional role and main work-related areas.

[0039] S2. Analyze the conversation content with the user. If the conversation content conforms to preset conversation rules, assign a conversation tag corresponding to the preset conversation rules to the conversation content, including: Analyze the content of multi-turn conversations with users. When the proportion of the content of the multi-turn conversations involving the same topic exceeds a preset proportion threshold, the frequency of the content of the multi-turn conversations involving the same topic exceeds a preset first frequency threshold, and the time span of the multi-turn conversations is less than a preset time span, assign corresponding conversation tags to the content of the multi-turn conversations.

[0040] Based on the frequency with which the content of the multi-round conversations revolves around the same topic, if the frequency exceeds a preset second frequency threshold, the conversation tag is determined to be a high-frequency conversation tag. If the frequency does not exceed the preset second frequency threshold, the session tag is determined to be a low-frequency session tag; When entering the next preset unit cycle, the high-frequency session tag is migrated to the next preset unit cycle, and the status of the high-frequency session tag is set to valid; At the same time, the status of the low-frequency session tag is set to invalid.

[0041] Specifically, the system analyzes the multiple rounds of dialogue initiated by A. In one session, A asked four questions within an hour, all revolving around the topic of "how to reduce the operating costs of a certain district branch in the third quarter." The system analysis revealed that the conversation topic was highly focused, the dialogue frequency was high, and the time span was short, conforming to the preset conversation rules. Therefore, the system generated a conversation tag "Optimization of Q3 Operating Costs of a Certain District Branch" for this session and marked it as a high-frequency conversation tag based on the intensity of the discussion.

[0042] S3. If the conversation content involves periodic keywords, then assign periodic tags corresponding to the periodic keywords to the conversation content, including: Within a preset unit period, monitor the number of times users ask questions about the periodic keywords; When the number of questions reaches a preset question threshold, a periodic tag related to the periodic keyword is assigned to the user.

[0043] Within the period corresponding to the period tag, the state of the period tag is set to valid; At the end of the cycle, the status of the cycle tag is set to invalid.

[0044] Specifically, the system monitors user questions within a preset period (e.g., monthly). This month, user A asked three questions related to the "Monthly Operations Report," including questions about "monthly deposit growth rate," "total loan disbursement statistics," and "operating expense details." Because the number of questions reached a preset threshold, the system assigned the user a "Monthly Operations Report" period tag and set its validity period to the current month.

[0045] S4. Integrate the role tag, the conversation tag, and the cycle tag into the user-submitted question content to obtain the converted question content, including: Obtain valid role tags, session tags, and cycle tags for the user's current state; The question content is combined with the content of the role tags, conversation tags, and cycle tags in a structured format to obtain the transformed question content.

[0046] Specifically, A submits the original question: "Please help me analyze this month's operational data." The system retrieves all of the user's current valid tags: Role tag: Bank operations personnel High-frequency conversation tags: Q3 operating cost optimization for a certain district branch Periodic tag: Monthly Operational Report The system integrates these tag information with the original question in a structured format to generate the transformed question content: I am a bank operations staff member, mainly concerned with the optimization of operating costs in the third quarter of a certain district branch. This month, the focus is on the monthly operating report. Please help me analyze the operating data for this month.

[0047] The system sends the converted question to the backend large language model. Based on the provided structured context information, the model immediately understands that the questioner is a bank operations professional who is currently focused on optimizing operating costs and needs the monthly operations report analysis.

[0048] Based on the structured data above, the model can directly generate a targeted response without needing to retrieve lengthy historical dialogues: "Okay, we have analyzed last month's operational data for you. Combined with your recent focus on cost optimization at a certain district branch, the data shows that operating costs in that district decreased by 3% month-on-month, but deposit growth slowed slightly. Below is a detailed monthly operational report, including a comparison of key indicators for each branch and optimization suggestions for balancing costs and business..." Over time, the system continuously updates its tagging system: at the beginning of the next month, the "Monthly Operations Report" periodic tag will automatically expire. If A no longer discusses cost issues in a certain area, that conversation tag may be reduced in frequency or expire. If A's long-term questioning areas change, the role tag may be re-evaluated. The system periodically recalculates the status of each tag from the latest questioning data.

[0049] Please refer to Figure 2 Embodiment two of the present invention is as follows: A query content conversion terminal 1 includes a memory 2 and a processor 3. The memory 2 stores a computer program that implements the method described in Embodiment 1.

[0050] The above description is merely an embodiment of the present invention and does not limit the patent scope of the present invention. Any equivalent modifications made based on the content of the present invention specification and drawings, or direct or indirect applications in related technical fields, are similarly included within the patent protection scope of the present invention.

Claims

1. A method for converting question content, characterized in that, include: Analyze the questions submitted by users. If the number of times the questions involve a preset field reaches a threshold, then assign a role tag corresponding to the preset field to the user. Analyze the conversation content with the user; if the conversation content conforms to the preset conversation rules, then assign the conversation content a conversation tag corresponding to the preset conversation rules. If the conversation content involves periodic keywords, then assign periodic tags corresponding to the periodic keywords to the conversation content; The user-submitted question content is combined with the role tag, the conversation tag, and the cycle tag to obtain the transformed question content.

2. The method according to claim 1, characterized in that, The analysis of user-submitted questions also includes: If a user logs in, the user's registration number used during login will be used as the user's unique identifier; If the user is not logged in, a unique device identifier is assigned to the electronic device used by the user, and the device identifier is used as the user's unique identifier; The corresponding role tag is generated and maintained based on the unique identifier.

3. The method according to claim 1, characterized in that, Analyze user-submitted questions. If the number of times a question relates to a preset domain reaches a threshold, assign the user a role tag corresponding to that preset domain, including: Collect and analyze the questions submitted by users consecutively within a preset time period; Identify the matching status between the continuously submitted questions and keywords in a preset domain; When the number of matches reaches the threshold, a role tag corresponding to the preset domain is assigned to the user.

4. The method according to claim 1, characterized in that, Analyze the conversation content with the user. If the conversation content conforms to preset conversation rules, assign a conversation tag corresponding to the preset conversation rules to the conversation content, including: Analyze the content of multi-turn conversations with users. When the proportion of the content of the multi-turn conversations involving the same topic exceeds a preset proportion threshold, the frequency of the content of the multi-turn conversations involving the same topic exceeds a preset first frequency threshold, and the time span of the multi-turn conversations is less than a preset time span, assign corresponding conversation tags to the content of the multi-turn conversations.

5. The method according to claim 4, characterized in that, Also includes: Based on the frequency with which the content of the multi-round conversations revolves around the same topic, if the frequency exceeds a preset second frequency threshold, the conversation tag is determined to be a high-frequency conversation tag. If the frequency does not exceed the preset second frequency threshold, the session tag is determined to be a low-frequency session tag; When entering the next preset unit cycle, the high-frequency session tag is migrated to the next preset unit cycle, and the status of the high-frequency session tag is set to valid; At the same time, the status of the low-frequency session tag is set to invalid.

6. The method according to claim 1, characterized in that, If the conversation content involves periodic keywords, then periodic tags corresponding to the periodic keywords are assigned to the conversation content, including: Within a preset unit period, monitor the number of times users ask questions about the periodic keywords; When the number of questions reaches a preset question threshold, a periodic tag related to the periodic keyword is assigned to the user.

7. The method according to claim 6, characterized in that, Also includes: Within the period corresponding to the period tag, the state of the period tag is set to valid; At the end of the cycle, the status of the cycle tag is set to invalid.

8. The method according to claim 1, characterized in that, The transformed question content is obtained by integrating the role tag, the conversation tag, and the cycle tag into the user-submitted question content, including: Obtain valid role tags, session tags, and cycle tags for the user's current state; The question content is combined with the content of the role tags, conversation tags, and cycle tags in a structured format to obtain the transformed question content.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 8.

10. A terminal for converting question content, characterized in that, It includes a memory, a processor, and a computer program stored in the memory and running on the processor, wherein the processor executes the computer program to implement the steps of the method for converting query content as described in any one of claims 1 to 8.

Images