Prompt word verbal skill hot update processing method for intelligent customer service question and answer scene

Abstract

The invention relates to the technical field of artificial intelligence and natural language processing, and discloses a prompt word and verbal skill hot update processing method for an intelligent customer service question and answer scene, and the method comprises the following steps: constructing a prompt word and verbal skill storage module which is used for storing prompt words and verbal skill for different business scenes; hot updating of prompt words and verbal skills is achieved, and real-time effectiveness of updated content is achieved by dynamically modifying the prompt words and verbal skill data in the storage module; performing semantic analysis and synonym replacement, performing semantic analysis on keywords of the cue words, and generating an expanded cue word set; performing linkage updating on cue word storage of the large model, and triggering fine adjustment of the large model by synchronizing updating contents of cue words; and constructing and dynamically expanding a question-answer knowledge base, and adding newly added question-answer pair data into the knowledge base. According to the method, real-time adaptation, efficient question answering and continuous optimization capabilities of the intelligent customer service system are realized through linkage updating of the cue word and the large model, semantic extension and dynamic knowledge base construction.

Description

Technical Field

[0001] The present invention relates to the technical field of artificial intelligence and natural language processing, and specifically to a method for hot updating of prompt words and speech for intelligent customer service question and answer scenarios. Background Art

[0002] Intelligent customer service question-and-answer systems are widely used in business scenarios in multiple fields. By automatically answering user questions, they reduce labor costs and improve service efficiency. In this process, prompt words and scripts, as core components of the system to understand user intent and provide standardized responses, need to have dynamic update capabilities to adapt to changes in business needs.

[0003] The existing intelligent customer service Q&A system prompt word hot update method mainly relies on manual update and preset static templates, combined with a simple keyword matching mechanism for Q&A processing. The advantages of this type of method are low implementation cost, suitable for systems with relatively simple scenarios, and can achieve relatively accurate answers in small-scale business scenarios. At the same time, it is easy to maintain and suitable for the needs of initial deployment and operation.

[0004] Existing technologies have obvious limitations in complex business environments with multiple scenarios and highly dynamic demands. The static settings of prompt words and scripts are difficult to adapt to business changes in real time, and the updating efficiency is low; the keyword matching method is unable to understand the diverse expressions of users, which can easily lead to inaccurate questions and answers; the expansion of the question and answer knowledge base lacks an efficient dynamic mechanism, and the quality and adaptability of the newly added question and answer data cannot be guaranteed, resulting in the gradual degradation of the system's question and answer effect, affecting the user experience. Summary of the invention

[0005] In view of the shortcomings of the prior art, the present invention provides a method for hot updating of prompt words and speech for intelligent customer service question and answer scenarios, which solves the problems in the prior art of low efficiency in updating prompt words and speech, insufficient understanding of users' diverse expressions, and lack of efficient dynamic mechanism for expanding the question and answer knowledge base.

[0006] To achieve the above objectives, the present invention is implemented through the following technical solutions: a method for hot updating of prompt words and phrases for intelligent customer service question and answer scenarios, comprising the following steps: Build a prompt word and speech storage module to store prompt words and speech for different business scenarios; Realize hot update of prompt words and speech, and make the updated content effective in real time by dynamically modifying the prompt words and speech data in the storage module; Perform semantic parsing and synonym replacement, perform semantic analysis on the keywords of the prompt words, and generate an expanded prompt word set; link and update the prompt word storage of the large model, and trigger the fine-tuning of the large model by synchronizing the updated content of the prompt words; Build and dynamically expand the question-and-answer knowledge base, add new question-and-answer pair data to the knowledge base, and link prompt words and large model updates.

[0007] Preferably, the step of constructing the prompt words and speech storage module includes: Predefine prompt words and scripts for each business scenario; Group and store prompt words and scripts according to business scenarios; Extract the keywords of the prompt word and generate a synonym set for each keyword; Store prompt words, phrases, keywords and their synonyms in a structured format in a file or database.

[0008] Preferably, the step of implementing hot update of prompt words and speech includes: Receive requests to add, modify or delete prompt words or scripts; Dynamically edit the prompt words and speech data in the storage module according to the request content; Call the update interface to synchronize the changes of prompt words and dialogues to the large model prompt word storage.

[0009] Preferably, the step of performing semantic parsing and synonym replacement includes: Segment the prompt words and extract the core keywords; Generate a set of synonyms for keywords based on a pre-trained word vector model; Replace the keywords in the prompt words with synonyms to generate prompt words with multiple semantic variants; The generated prompt word set is stored in a prompt word storage module.

[0010] Preferably, the step of linking and updating the prompt word storage of the large model includes: Extracting an updated prompt word set from a prompt word storage module; Synchronize the updated prompt word set to the large model prompt word storage file; Trigger fine-tuning training of the large model to ensure that the new prompt words can adapt to the generated question and answer content.

[0011] Preferably, the fine-tuning training of the large model includes: Extract training data from the question-answering knowledge base; Use transfer learning methods to fine-tune large models; Deploy the fine-tuned large model to the intelligent customer service system.

[0012] Preferably, the large model prompt word storage file is associated with the prompt word storage module through an interface, and the prompt word storage file is updated synchronously when the prompt word is updated.

[0013] Preferably, the question-answering knowledge base includes: An initial knowledge base built based on historical manual business question and answer pairs; New question-answer pair data collected in real time during the operation of intelligent customer service; New question and answer samples that have been manually reviewed and approved.

[0014] Preferably, after the prompt word is updated, the following linkage operations are automatically triggered: Extract the prompt word set and update it to the large model prompt word storage; Use prompt words to update content to fine-tune the large model; Deploy the fine-tuned large model to the intelligent customer service system.

[0015] Preferably, the step of dynamically expanding the question-answer knowledge base includes: Monitor the user interaction logs of intelligent customer service and extract unmatched user questions; Mark as potential question and answer samples and submit for manual review; After the review is passed, the new question and answer pairs will be stored in the knowledge base, and the prompt words and large model update process will be triggered.

[0016] The present invention provides a method for hot updating of prompt words and speech for intelligent customer service question and answer scenarios. It has the following beneficial effects: 1. The present invention realizes the dynamic update and synchronization of prompt word data through the linkage mechanism between the prompt word storage module and the large model prompt word storage file, and combines the fine-tuning training and rapid deployment of the large model to ensure that the system can adapt to new prompt words in real time and improve the response efficiency of intelligent customer service.

[0017] 2. The present invention adopts semantic analysis and synonym replacement technology to dynamically generate multiple semantic variants of prompt words. Combined with the expanded question-and-answer knowledge base, it enhances the system's ability to understand diverse user expressions and significantly improves the accuracy and scenario coverage of questions and answers.

[0018] 3. The present invention builds a dynamically expanding question-and-answer knowledge base by real-time monitoring of user interaction logs and manually reviewing the process of newly added question-and-answer pairs, while ensuring the high quality of newly added data and updated models, and cooperating with verification and monitoring mechanisms to improve the stability and long-term optimization capabilities of the system. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 The figure is a flow chart of the method of the present invention. DETAILED DESCRIPTION

[0020] The following will be combined with the drawings in the specification of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0021] Please refer to the attached Figure 1 The embodiment of the present invention provides a method for hot updating of prompt words and phrases for intelligent customer service question and answer scenarios, comprising the following steps: Build a prompt word and speech storage module to store prompt words and speech for different business scenarios. By building a unified storage module, prompt words and speech can be efficiently managed and retrieved, providing a basis for hot updates and linkage large model operations; Realize hot update of prompt words and speech, and make the updated content effective in real time by dynamically modifying the prompt words and speech data in the storage module. Hot update triggers the corresponding processing logic by monitoring the change events of the prompt words and speech storage module. The update operation uses the transaction processing mechanism to ensure data consistency and avoid data confusion caused by midway failure. Perform semantic parsing and synonym replacement, perform semantic analysis on the keywords of the prompt words, and generate an expanded prompt word set. Through semantic expansion, it can better match the diverse expressions of user input, reduce question-and-answer failures caused by keyword mismatch, and thus improve the system's response accuracy and user experience; Linked update of the prompt word storage of the large model, triggering fine-tuning of the large model by synchronizing the updated content of the prompt word. Linked update uses the interface mechanism to ensure that the prompt word changes can be transmitted to the large model storage file in real time, and optimize the model based on this. Linked update ensures that the content of the prompt word changes can be quickly reflected in the generation capability of the large model to avoid disconnection between data and model; Build and dynamically expand the question-and-answer knowledge base, add new question-and-answer data to the knowledge base, and link prompt words and large model updates. Dynamic expansion collects new user questions in real time and compares them with existing questions and answers to identify uncovered areas and supplement data in a timely manner. The expansion of the knowledge base, prompt words, and large models form a closed loop, thereby improving the intelligent customer service's ability to respond to complex user questions and optimizing the overall service effect.

[0022] Please refer to the attached Figure 1 In a preferred embodiment of the present invention, the step of constructing a prompt word and speech storage module includes: Predefine prompt words and scripts for each business scenario. By predefine prompt words and scripts, a standardized description of business requirements is formed to facilitate subsequent storage, management, and dynamic call. Standardized scripts ensure that users can get consistent answers and reduce misunderstandings caused by differences in expression. The prompt words and speech scripts are grouped and stored according to business scenarios. The prompt words and speech scripts are grouped and managed by business scenarios to facilitate independent maintenance and quick query of different scenarios. The grouped storage of data reduces redundancy and improves storage and call efficiency. Extract the keywords of the prompt word and generate a synonym set for each keyword. Keyword extraction is based on word segmentation technology. The prompt word content is semantically analyzed to identify the core description. Synonym generation uses word vector semantic matching technology to expand the expression range of keywords and enhance the coverage of prompt words for diverse user inputs. Store prompt words, sales pitches, keywords and their synonyms in a structured format in files or databases. Structured storage manages prompt words, sales pitches, keywords and their synonyms in a unified manner through clear field definitions, making them easy to query and dynamically update. Data structuring enables the system to efficiently handle complex multi-scenario, multi-level business logic.

[0023] Please refer to the attached Figure 1 In a preferred embodiment of the present invention, the step of implementing the hot update of prompt words and speech includes: Receive requests for adding, modifying or deleting prompt words or phrases. The management system monitors the adding, modifying or deleting operations through the interface and passes the request content to the storage module for processing. Each request type (adding, modifying, deleting) triggers a specific update operation to ensure the dynamic and accurate nature of the data, while improving the system's response speed and maintenance efficiency. Dynamically edit the prompt words and speech data in the storage module according to the request content. Data editing is implemented through a transaction processing mechanism to ensure that the addition, modification or deletion operations are atomic, that is, all operations are either successful or rolled back to avoid incomplete updates, thereby improving the management efficiency of prompt words and speech. Call the update interface to synchronize the changes of prompt words and dialogue to the large model prompt word storage. The update interface is implemented through REST API or RPC (remote procedure call) to ensure seamless synchronization of data from the storage module to the large model prompt word storage, and to ensure that the changes of prompt words and dialogue can be reflected in the large model in real time to avoid disconnection between data and model.

[0024] Please refer to the attached Figure 1 In a preferred embodiment of the present invention, the step of performing semantic parsing and synonym replacement includes: Segment the prompt words and extract the core keywords. The segmentation operation is based on statistical language models or rule methods to divide the prompt words into analyzable vocabulary units. Stop word filtering removes high-frequency meaningless words through word frequency statistics to ensure that the extracted keywords can accurately express the core semantics of the prompt words. Through segmentation and keyword extraction, the semantic parsability of the prompt words is improved, laying the foundation for synonym generation and expansion. Generate a set of synonyms for keywords based on a pre-trained word vector model. The word vector model represents the semantic relationship of words through a high-dimensional vector space. Synonym generation measures semantic similarity based on the distance between vectors. The dedicated vocabulary combines domain knowledge to optimize synonym generation and improve the coverage of industry-specific expressions, thereby enhancing the adaptability of the system. The keywords in the prompt words are replaced with synonyms to generate multiple semantic variant prompt words. The implementation method is as follows: according to the synonym set, the keywords in the prompt words are replaced one by one to generate multiple semantic variants: Replacement logic: traverse the keyword list of prompt words; for each keyword, replace it with each word in the synonym set to generate a new prompt word. By generating multiple semantic variant prompt words, the probability of question and answer failure due to different user expressions is reduced; The generated prompt word set is stored in the prompt word storage module. The prompt word storage module manages the semantic variant set in an efficient and structured manner, supports fast calling and updating operations, and the indexing mechanism improves retrieval efficiency. Version control ensures the traceability of the prompt word set. The structured storage of the prompt word set facilitates management and retrieval, ensuring that the system can call the adapted semantic variant in real time.

[0025] Please refer to the attached Figure 1 In a preferred embodiment of the present invention, the step of updating the prompt word storage of the large model in a linked manner includes: Extract the updated prompt word set from the prompt word storage module, and realize incremental extraction by recording the update time field of the prompt word to reduce system resource consumption. The full extraction mechanism is used as an alternative to ensure that no prompt word data is missed in special scenarios. Incremental extraction reduces unnecessary data reading operations and improves system update efficiency. The updated prompt word set is synchronized to the large model prompt word storage file. The prompt word storage module and the large model prompt word storage are linked through the interface to ensure the real-time and consistency of data synchronization. The format conversion and verification ensure that the prompt word set can be directly parsed and used by the large model. The updated prompt word set can be quickly transferred to the large model storage file, shortening the time from prompt word change to question and answer adaptation; data consistency verification reduces the risk of training failure due to format errors during the update process; triggers fine-tuning training of the large model to ensure that the new prompt word can adapt to the generated question and answer content. The fine-tuning training is based on transfer learning. The model weights are updated through small-scale data to adapt to the semantics and question and answer requirements of the new prompt word, thereby verifying the test by simulating user query scenarios to evaluate the question and answer generation effect of the new model on the prompt word. The verification test improves system reliability and avoids the decline in user experience due to model adaptation failure.

[0026] Please refer to the attached Figure 1 In a preferred embodiment of the present invention, the fine-tuning training of the large model includes: Extract training data from the question-answering knowledge base, and analyze the correlation between the prompt words and the question-answering knowledge base to ensure that the training set covers the semantic range of the prompt words. The formatting operation converts the question-answering data into input-output pairs that are suitable for the model to ensure that the data can be effectively used, providing high-quality training data and ensuring that the large model can accurately adapt to the semantic features of the prompt words. Expand the coverage of training data through semantic variants to further improve the model's question-answering capabilities; Use transfer learning to fine-tune the large model. Transfer learning can reduce the computing resource consumption of large-scale retraining by fine-tuning weight parameters based on the existing pre-trained model. Transfer learning can quickly adapt to the semantic needs of updating prompt words without training the large model from scratch. Dynamic adjustment of training strategies optimizes the balance between training efficiency and model performance. The fine-tuned model can accurately answer user queries based on new prompt words, significantly improving the accuracy of question and answer. The fine-tuned large model is deployed to the intelligent customer service system. Model verification simulates user interaction scenarios through test data to ensure that the model reaches the expected performance before going online and that the fine-tuned model can run efficiently and stably in the intelligent customer service system to provide accurate question-and-answer services. The deployment uses containerization technology to improve the scalability and operational stability of the model. Through real-time monitoring and user feedback, a continuous optimization mechanism is formed to continuously improve the intelligence level of the system.

[0027] Please refer to the attached Figure 1In a preferred embodiment of the present invention, the large model prompt word storage file is associated with the prompt word storage module through an interface. When the prompt word is updated, the prompt word storage file is updated synchronously. The interface association realizes the data linkage between the prompt word storage module and the large model prompt word storage file to avoid inconsistency between the two contents. The automated interface association mechanism improves the efficiency of the prompt word update process and shortens the data transmission time. The synchronization task is automatically triggered by monitoring the update event of the storage module, without manual intervention, ensuring real-time performance.

[0028] Please refer to the attached Figure 1 In a preferred embodiment of the present invention, the question-answering knowledge base includes: The initial knowledge base is built based on historical manual business question and answer pairs. The knowledge base is built using historical manual business question and answer pairs, laying the foundation for the initial question and answer of the intelligent customer service system. It provides high-quality initial question and answer data to ensure that the intelligent customer service system can provide accurate and comprehensive question and answer services in the early stage of going online. The newly added question-answer pairs data collected in real time during the operation of intelligent customer service is used to dynamically collect new question-answer pairs by using the real-time interaction records between users and intelligent customer service, expand the coverage of the knowledge base, dynamically expand the knowledge base, and adapt to the ever-changing question needs of users, thereby improving the updating efficiency of the knowledge base and enhancing the question-answer adaptability of intelligent customer service; The review process for the newly added Q&A samples that have been manually reviewed is as follows: Manually review the potential question-answer pairs collected in real time: verify the accuracy and semantic integrity of the questions; edit or supplement the standard answers to the questions according to business needs; mark the question-answer pairs that have passed the review as "verified"; Storage update: Add the approved question and answer pairs to the knowledge base, using a structured storage method consistent with the initial knowledge base; automatically trigger the prompt word update process, generate prompt words for the newly added questions and synchronize them to the prompt word storage module; Regular review: Regularly review the quality of new question-answer pairs added to the knowledge base and remove outdated or inapplicable data; The accuracy and reliability of knowledge base expansion are improved through manual review. At the same time, the newly added structured storage and annotation of question and answer samples support fast retrieval and model adaptation.

[0029] Please refer to the attached Figure 1 In a preferred embodiment of the present invention, after the prompt word is updated, the following linkage operations are automatically triggered: Extract the prompt word set and update it to the large model prompt word storage. The prompt word update trigger mechanism: when the prompt word storage module detects that the prompt word addition, modification or deletion operation is completed, the listener automatically triggers the data synchronization process; the listener filters out the changed prompt word set according to the update record and outputs it in a structured format; Data synchronization to large model prompt word storage: The extracted prompt word set is pushed to the large model prompt word storage file through the API interface or data stream; data synchronization adopts batch update mode to avoid performance overhead caused by frequent calls; The automatic trigger mechanism captures update events in real time by monitoring the operation log of the prompt word storage module. Data push is seamlessly connected through standardized interfaces to ensure that the prompt words remain consistent between the storage module and the large model, thereby ensuring that prompt word updates can be quickly synchronized to the large model prompt word storage, shortening data transmission time. Use prompt words to update content to fine-tune the large model. Fine-tune the model through transfer learning technology to optimize the model weights based on the pre-trained model to adapt it to the question-answering needs of the new prompt words. Training verification is simulated through actual scenarios to ensure that the model's question-answering capabilities for the newly added prompt words meet expectations, thereby improving the accuracy of the answers generated by the model. At the same time, verification testing reduces the risk of question-answering errors after model deployment and improves system stability. The fine-tuned large model is deployed to the intelligent customer service system. Model verification simulates real user query scenarios through test input to ensure the model's adaptability to prompt words. The fine-tuned model can be quickly launched to support new prompt words, shortening the time from data update to system question and answer adaptation.

[0030] Please refer to the attached Figure 1 In a preferred embodiment of the present invention, the step of dynamically expanding the question-answer knowledge base includes: monitoring the user interaction log of the intelligent customer service, extracting unmatched user questions, and the matching algorithm determines whether there is a corresponding question-answer pair for the user's question through keyword similarity, semantic analysis and other technologies, and the marking and storage of unmatched questions provide basic data for knowledge base expansion, extracting unmatched questions in real time, ensuring that the knowledge base can quickly respond to changes in user needs, while reducing the possibility of missing user questions and improving user experience; Mark as potential Q&A samples and submit for manual review. Reviewers process potential samples through the management platform or batch review tools: Verify the completeness and applicability of the question content; write standard answers based on business rules and existing knowledge; after the review is completed, mark the status as "approved"; During the review process, unqualified samples (such as those with unclear semantics) are directly marked as "failed the review" to prevent low-quality data from entering the knowledge base; Manual review provides a data quality control mechanism to ensure that the content of newly added question and answer pairs is accurate and meets business requirements, thereby improving the quality of newly added question and answer pairs and preventing the accuracy of the knowledge base from being affected by low-quality data; After the review is passed, the new question and answer pairs will be stored in the knowledge base, and the prompt word and large model update process will be triggered. The knowledge base update ensures that the knowledge base can quickly respond to user queries through structured storage and index optimization of the new question and answer pairs. The dynamic expansion of the knowledge base improves the system's coverage of new questions and optimizes the user experience. At the same time, the linkage update of prompt words and large models shortens the cycle from the collection of new data to the generation of question and answer, and realizes efficient dynamic adaptation.

[0031] In order to better understand the present invention, the above contents are described in detail below in conjunction with specific embodiments.

[0032] Example 1: Implementation method of a static intelligent customer service system without adding a dynamic update mechanism: prompt words and scripts are stored in a static manner and cannot be updated dynamically. The stored content needs to be updated through manual operation. The system only relies on keyword matching to answer user questions, and does not introduce semantic parsing and dynamic knowledge base expansion functions.

[0033] Beneficial effects: Suitable for systems with a single business scenario and low update frequency; low implementation cost, and relatively simple initial deployment and maintenance.

[0034] Example 2: Implementation method of an intelligent customer service system with a dynamic update mechanism for adding prompt words and lines of speech: The prompt words and lines of speech implement a hot update mechanism, which supports dynamic adjustment of content through adding, modifying or deleting operations and takes effect in real time, but does not introduce semantic extension functions and linkage updates with large models.

[0035] Beneficial effects: Dynamic adjustment of prompt words and scripts improves maintenance efficiency and adapts to the needs of some business changes; the system can quickly update prompt words and script content, improving the flexibility of data management.

[0036] Example 3: Implementation method of intelligent customer service system with semantic expansion and prompt word linkage update mechanism added: Based on Example 2, semantic analysis and prompt word expansion functions are added to extract keywords and generate synonyms for prompt words, expanding the coverage of prompt words. At the same time, after the prompt words are updated, they are synchronized with the large model storage file, triggering the large model fine-tuning training to adapt to the new prompt words.

[0037] Beneficial effects: The system's ability to understand users' diverse expressions has been improved, and the accuracy of question and answer has been significantly improved; the large model and prompt words are updated in conjunction, achieving real-time adaptation of question and answer capabilities and enhancing the system's dynamic response capabilities.

[0038] Example 4: Implementation method of dynamic intelligent customer service system based on full-process linkage mechanism: Based on Example 3, a dynamic knowledge base expansion function is introduced to extract unmatched questions by monitoring user interaction logs, add new question and answer pairs after manual review and store them in the knowledge base, and trigger the prompt word and large model update process at the same time. The full-process automated linkage mechanism forms a complete closed-loop adaptation system.

[0039] Beneficial effects: The full-process dynamic adaptation of prompt words, knowledge base and large models is realized, and the question-answering system can quickly respond to changes in business needs; the dynamic knowledge base expansion function enhances the system's coverage and self-optimization capabilities, and improves the reliability and scalability of the question-answering effect; the automated linkage mechanism reduces the need for manual intervention and further improves system efficiency.

[0040] Comparative experiment 1: Comparative experiment on real-time updating of prompt words and speech scripts Experimental purpose: To verify the real-time advantage of the hot update mechanism of prompt words and speech of the present invention.

[0041] Experimental setup Experimental Group Use the prompt words and speech hot update mechanism of the present invention.

[0042] After the prompt word is updated, it is automatically synchronized to the large model prompt word storage file and takes effect in the question and answer system in real time.

[0043] Control group 1: static update system After the prompt words and scripts are manually updated, the system needs to be restarted. The time when the update takes effect depends on the manual operation and the system restart time.

[0044] Control group 2: Dynamic update system without synchronization Prompt words and scripts can be dynamically updated to the storage module, but they are not synchronized with the big model and are only partially effective in the question-and-answer system.

[0045] Experimental procedures Prepare prompt words to update the scene Prepare 5 prompt word update tasks, including adding, modifying, and deleting operations. Each task updates the prompt word content in different scenarios.

[0046] Record update and effective time Complete the prompt word update task in each group and record: The time from submitting the update task to completing the storage module update.

[0047] The total time from when the storage module is updated to when the system Q&A takes effect.

[0048] Statistics and Analysis The average time for the system to take effect after completing the prompt word update task was compared among the groups to analyze the real-time differences.

[0049] The comparative experimental data are shown in Table 1: Table 1 Statistics of prompt word update and effective time From the data in Table 1, we can get: Experimental group (the present invention): through the hot update mechanism of prompt words and speech, after the system completes the update of the storage module, the prompt words are immediately synchronized to the large model prompt word storage file, and take effect in the question-answering system in real time. The total time is 8.4 seconds on average, which is significantly better than the control group.

[0050] Control group 1 (static update system): prompt word updates require manual restart of the system, and the total time is much higher than that of the experimental group, reaching an average of 102 seconds. The update efficiency is low and the system adaptability is poor.

[0051] Control group 2 (dynamic update system without synchronization): After the prompt words are updated to the storage module, the question-answering system takes a long time to take effect (21.8 seconds on average) due to the lack of a large model synchronization mechanism and insufficient question-answering adaptability.

[0052] Comparative Experiment 2: Comparative Experiment on Question and Answer Coverage Experiment Purpose: To verify the advantages of the present invention in the coverage of the question and answer system through the semantic expansion of prompt words and the synonym replacement mechanism, and to compare the question and answer coverage of user questions in the newly added scenarios.

[0053] Experimental setup Experimental Group Using the system of the present invention, the prompt word supports semantic expansion and synonym generation. After the prompt word is updated, the linkage update of the large model is triggered to adapt to the diverse user expressions.

[0054] Control group 1: static prompt word system The prompt words are fixed keywords without dynamic expansion or semantic adaptation. Users need to accurately match the prompt word expression to get the correct answer.

[0055] Control group 2: Dynamic update system without synonym expansion Prompt words can be updated dynamically, but synonym expansion is not supported, and the semantic coverage capability is limited.

[0056] Experimental procedures Building a test suite A question set containing 10 new business scenarios was constructed, and each question had three variant expressions (keyword replacement, sentence structure change, and colloquial description).

[0057] Tested on various systems Use the 30 questions in the test set to test each group of systems one by one, and record the question and answer coverage of each question (whether the answer is correct).

[0058] Statistics and Analysis The question-answer coverage (proportion of correct answers) of each group of systems was calculated, and the differences in coverage were analyzed.

[0059] The comparative experimental data are shown in Table 2: Table 2 Question and answer coverage statistics From the data in Table 2, we can get: Experimental group (the present invention): Through semantic analysis and synonym replacement technology, the system achieves high adaptability to diverse expressions, showing a high coverage rate (93.3% on average) in all test scenarios, especially in scenarios with complex expressions.

[0060] Control group 1 (static prompt word system): Since the prompt words are fixed keywords, users need to completely match the prompt word expression, resulting in extremely low question and answer coverage (13.3% on average), showing obvious limitations.

[0061] Control group 2 (dynamic update system without synonym expansion): The dynamic update prompt words improved the coverage to a certain extent (46.7% on average), but lacked a semantic expansion mechanism and was not adaptable enough to complex expressions.

[0062] Comparative Experiment 3: Comparative Experiment on Dynamic Knowledge Base Expansion Capability Experiment Purpose: To verify the advantages of the dynamic knowledge base expansion mechanism of the present invention in expansion efficiency and the quality of newly added question and answer pairs.

[0063] Experimental setup Experimental Group By using the dynamic knowledge base expansion mechanism of the present invention, the user interaction log is monitored in real time to extract unmatched questions, new question and answer pairs are added through manual review, and prompt words are automatically triggered to be linked to the large model for update.

[0064] Control group 1: Manually expanding the knowledge base Unmatched questions are manually screened, sorted and added to the knowledge base. There is no real-time monitoring or linkage mechanism, and efficiency depends entirely on manual processing.

[0065] Control group 2: Automatic scaling system without auditing The system can automatically collect unmatched questions and add them to the knowledge base, but due to the lack of manual review, the newly added question and answer pairs may be of low quality or inaccurate.

[0066] Experimental steps: Simulate interaction to generate unmatched problems Simulate the real interaction between users and the system and generate 30 unmatched questions distributed in different business scenarios.

[0067] Expand and add new question and answer pairs: record the completion time of each group processing 30 unmatched questions, including the entire process from question collection and review to the storage of new question and answer pairs.

[0068] Test the accuracy and adaptation rate of the newly added question and answer pairs.

[0069] Statistics and analysis: Compare the differences among the groups in expansion efficiency (time) and expansion quality (accuracy, adaptation rate).

[0070] The comparative experimental data are shown in Table 3: Table 3 Comparison of dynamic knowledge base expansion efficiency and quality From the data in Table 3, we can get: Experimental group (the present invention): realized real-time monitoring and collection of unmatched questions, ensured the high quality of new question-answer pairs through manual review, and automatically linked prompt words and large model updates. The average expansion completion time was 45 minutes, the accuracy of new question-answer pairs reached 97%, and the adaptation rate reached 95%.

[0071] Control group 1 (manual expansion of knowledge base): completely dependent on manual screening and addition, with low efficiency, an average completion time of up to 180 minutes, and an accuracy rate of 100% for newly added question and answer pairs, but the adaptation rate was only 80% due to delayed updates.

[0072] Control group 2 (automatic expansion system without review): relies on automatic collection to achieve rapid expansion, with an average completion time of only 30 minutes. However, due to the lack of manual review, the accuracy of new question and answer pairs is only 65%, the adaptation rate is 70%, and the quality of new data is low.

[0073] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A method for hot updating prompt words and phrases for intelligent customer service question and answer scenarios, characterized in that: The following steps are involved: Build a prompt word and speech storage module to store prompt words and speech for different business scenarios; Realize hot update of prompt words and speech, and make the updated content effective in real time by dynamically modifying the prompt words and speech data in the storage module; Perform semantic parsing and synonym replacement, perform semantic analysis on the keywords of the prompt words, and generate an expanded prompt word set; Linked update of the prompt word storage of the large model, triggering fine-tuning of the large model by synchronizing the updated content of the prompt words; Build and dynamically expand the question-and-answer knowledge base, add new question-and-answer pair data to the knowledge base, and link prompt words and large model updates.

2. The method for hot updating of prompt words and phrases for intelligent customer service question and answer scenarios according to claim 1 is characterized in that: The steps of constructing the prompt words and speech storage module include: Predefine prompt words and scripts for each business scenario; Group and store prompt words and scripts according to business scenarios; Extract the keywords of the prompt word and generate a synonym set for each keyword; Store prompt words, phrases, keywords and their synonyms in a structured format in a file or database.

3. The method for hot updating of prompt words and speech for intelligent customer service question and answer scenarios according to claim 1 is characterized in that: The steps of implementing the hot update of prompt words and speech scripts include: Receive requests to add, modify or delete prompt words or scripts; Dynamically edit the prompt words and speech data in the storage module according to the request content; Call the update interface to synchronize the changes of prompt words and dialogues to the large model prompt word storage.

4. The method for hot updating of prompt words and speech for intelligent customer service question and answer scenarios according to claim 3 is characterized in that: The steps of performing semantic parsing and synonym replacement include: Segment the prompt words and extract the core keywords; Generate a set of synonyms for keywords based on a pre-trained word vector model; Replace the keywords in the prompt words with synonyms to generate prompt words with multiple semantic variants; The generated prompt word set is stored in a prompt word storage module.

5. The method for hot updating of prompt words and speech for intelligent customer service question and answer scenarios according to claim 1 is characterized in that: The step of linking and updating the prompt word storage of the large model includes: Extracting an updated prompt word set from a prompt word storage module; Synchronize the updated prompt word set to the large model prompt word storage file; Trigger fine-tuning training of the large model to ensure that the new prompt words can adapt to the generated question and answer content.

6. The method for hot updating of prompt words and speech for intelligent customer service question and answer scenarios according to claim 5 is characterized in that: The fine-tuning training of the large model includes: Extract training data from the question-answering knowledge base; Use transfer learning methods to fine-tune large models; Deploy the fine-tuned large model to the intelligent customer service system.

7. The method for hot updating of prompt words and speech for intelligent customer service question and answer scenarios according to claim 1 is characterized in that: The large model prompt word storage file is associated with the prompt word storage module through an interface, and the prompt word storage file is updated synchronously when the prompt word is updated.

8. The method for hot updating of prompt words and speech for intelligent customer service question and answer scenarios according to claim 1 is characterized in that: The question-answering knowledge base includes: An initial knowledge base built based on historical manual business question and answer pairs; New question-answer pair data collected in real time during the operation of intelligent customer service; New question and answer samples that have been manually reviewed and approved.

9. The method for hot updating of prompt words and speech for intelligent customer service question and answer scenarios according to claim 1, characterized in that: After the prompt word is updated, the following linkage operations are automatically triggered: Extract the prompt word set and update it to the large model prompt word storage; Use prompt words to update content to fine-tune the large model; Deploy the fine-tuned large model to the intelligent customer service system.

10. The method for hot updating of prompt words and speech for intelligent customer service question and answer scenarios according to claim 1, characterized in that: The step of dynamically expanding the question-answering knowledge base includes: Monitor the user interaction logs of intelligent customer service and extract unmatched user questions; Mark as potential question and answer samples and submit for manual review; After the review is passed, the new question and answer pairs will be stored in the knowledge base, and the prompt words and large model update process will be triggered.

Images