Intelligent information input method and system, electronic equipment and storage medium

By deploying intelligent information entry methods on the client, using keyword retrieval and sorting of the historical sample library, the problems of high investment in existing AI technologies and insufficient personalized support are solved, efficient and accurate information entry is achieved, and hardware and labor costs and improve the real-time update capability of the model.

CN119066187APending Publication Date: 2024-12-03CHINA PING AN PROPERTY INSURANCE CO LTD
View PDF 0 Cites 2 Cited by

Patent Information

Application Number
CN202411206625.3
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-08-30
Publication Date
2024-12-03

AI Technical Summary

Technical Problem

The existing general-purpose big model generation AI technology has high investment and its effects are affected by input, making it difficult to achieve personalized support, and the hardware cost is high, which cannot effectively improve the efficiency and accuracy of information entry.

Method used

By deploying intelligent information entry methods on the client, similar replies are retrieved using keywords in the historical sample library, sorting them according to the number of matches and the most recent sending time, providing sorting results for customer service to choose, and updating the number of usages and time in real time to reduce hardware Cost, improve search accuracy.

Benefits of technology

It realizes that without increasing hardware costs, improves the efficiency and accuracy of information entry, reduces labor costs, provides personalized information services, and has high real-time model updates, and does not affect the model results of other users.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119066187A_ABST
    Figure CN119066187A_ABST
Patent Text Reader

Abstract

The invention provides an intelligent information input method and system, electronic equipment and a storage medium, and the method is characterized in that the method comprises the steps: obtaining the problem information of a customer, and confirming the category of a problem; similar reply recommendation entries containing the keywords are retrieved from a historical sample library through the keywords; the similar reply recommendation entries are sorted according to the weight information, the higher the weight is, the higher the sorting order is, and otherwise, the higher the sorting order is; and outputting a sorting result. According to the method, the hardware cost is saved, and the retrieval accuracy is improved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent information entry, and in particular to an intelligent information entry method, system, electronic device and storage medium. Background Art

[0002] Ping An Property & Casualty Insurance's main auto insurance business is divided into sales period and service period based on the customer's insurance expiration date, which has obvious periodicity. Instant messaging services and text form entry are used more frequently in daily work, and the content is similar. Based on the repetitive periodicity and similarity of historical text entry content, certain prompts are given to customer service staff when they enter text, and automatic filling can be used to effectively improve work efficiency and reduce repetitive work.

[0003] The overall input-output ratio of the currently popular general-purpose large-model generative AI technology on the market is very difficult to quantify, which hinders the implementation of the technology. The main reasons are as follows:

[0004] 1. Deep learning is required for different business scenarios, which has high thresholds for learning samples, hardware computing power, and model algorithms, and requires huge investments.

[0005] 2. The questioner’s question description greatly affects the AI’s answer, and the input is uncontrollable.

[0006] 3. Individual users have different subjective feelings about satisfaction with answers. Other people’s correct answers may not be applicable to themselves, and it is difficult to customize satisfactory answers for each user. Summary of the invention

[0007] The embodiments of the present invention provide a method, system, electronic device and storage medium for intelligent information entry, which reduce hardware costs and improve the efficiency and accuracy of information entry.

[0008] The technical solution of the present invention is:

[0009] First, as Figure 1 As shown, the present application provides an information intelligent entry method, comprising:

[0010] Obtain customer problem information and confirm the scope of the problem;

[0011] Retrieve similar reply recommendation entries containing the keyword from the historical sample library through the keyword;

[0012] Sort similar reply recommendation terms according to weight information, where the higher the weight, the higher the ranking, otherwise the lower the ranking;

[0013] Output the sorting results, and check the top-ranked recommended reply terms for customer service to choose when replying to customers.

[0014] Furthermore, the method for sorting similar reply recommendation terms according to weight information includes performing sorting calculations using the number of matching texts and the most recent sending time as weights, and the calculation formula is:

[0015]

[0016] Among them, F(t,n) represents the final score, t represents the difference between the sending time of the matching historical entry and the current time of the system. The smaller the value, the higher the matching degree of the output result; n represents the number of times the search term appears in the historical sent text, and e represents the exponential function.

[0017] Furthermore, the output ranking result includes each recommended term and its corresponding usage count and most recent usage time.

[0018] Furthermore, the method further comprises filtering the search results to filter out the results that match the error set, and for the matching results that match the star-marked search set (that belong to common terms), placing them at the top of the output ranking for recommendation.

[0019] After outputting the ranking result, the method updates the number of times and the time of use of the top reply term to the historical sample library after the reply term is used.

[0020] Furthermore, for the historical text recommendation finally given, the user will manually decide whether to use it, and the selection result (selected, discarded, starred) will be fed back to the engine search set as the marked content and participate in the next calculation.

[0021] The search engine of this method is deployed on the client computer, the server resources will not increase due to the increase in the number of users, and the hardware cost is greatly reduced.

[0022] Each new input by the customer service staff of the present invention will immediately generate new recommendations for their own models (for example, if only one word "claim" is input, there are too many matching entries and no differentiation. After adding "arrival", no recommendation is found, indicating that "arrival" is not suitable as a keyword. Then the recommended combination of "claim" + "personal injury" is changed to the most appropriate one). In this way, users can try multiple keyword combinations in a very short time and quickly find suitable recommendations in the current scenario. And by sending a confirmation, the weight of the selected entry in the next matching recommendation is strengthened.

[0023] The training and labeling of this method are integrated into the regular work process of customer service, without interfering with their normal work content. At the same time, it does not require the addition of additional professional data labeling personnel. Compared with most AI product solutions, the labor cost is greatly improved.

[0024] The data retrieved by this method are all contents sent by the users themselves, which is equivalent to having an exclusive knowledge base. Keywords are used based on memory and usage habits, reducing the cost of learning how to ask questions.

[0025] In a second aspect, the present application also provides an information intelligent entry system, comprising:

[0026] Customer problem acquisition unit, used to obtain customer problem information and confirm the problem scope;

[0027] A recommended term retrieval unit, used to retrieve similar reply recommended terms containing the keyword from a historical sample library by using the keyword;

[0028] A recommended term ranking processing unit, used to rank similar reply recommended terms according to weight information, wherein the higher the weight, the higher the ranking, otherwise the lower the ranking;

[0029] The result output unit is used to output the sorting result.

[0030] Furthermore, the system also includes a filtering unit, which is used to filter the search results and recommend the matching results to the top in the output ranking.

[0031] Furthermore, the system also includes a historical sample updating unit, which is used to update the number of times and the usage time of the top-ranked reply term to the historical sample library after outputting the sorting result and after the reply term is used.

[0032] In a third aspect, the present application also provides a computer storage medium, characterized in that: the computer storage medium stores a plurality of instructions, and the instructions are suitable for being loaded by a processor and executing the above-mentioned method steps.

[0033] In a fourth aspect, the present application further provides an electronic device, characterized in that it comprises a processor and a memory, wherein the memory stores a computer program, and the computer program is suitable for being loaded by the processor and executing the above-mentioned method steps.

[0034] The beneficial effects brought about by the technical solutions provided by some embodiments of the present application include at least:

[0035] The computing program that can perform real-time matching can be directly deployed on the customer service's work computer and run without adding any additional hardware costs. The decentralized design eliminates server resource bottlenecks, network bandwidth bottlenecks, and data storage bottlenecks in business concurrency.

[0036] By limiting the historical data entered by the user as samples and not introducing other data, users can maintain their usual expression habits to retrieve information and improve retrieval accuracy.

[0037] For high-quality / inaccurate retrieval results, you can directly mark them and apply them to your own retrieval model. The model is updated in real time and will not affect other people's model results. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] In order to more clearly illustrate the embodiments of the present invention, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0039] Figure 1 is a flow chart of a method according to an embodiment of the present invention;

[0040] Figure 2 This is a graph showing the output result of the “Example of Consultation Dialogue on Time Limitation for Injury Compensation without Person” according to an embodiment of the present invention;

[0041] Figure 3 is a system block diagram of an embodiment of the present invention;

[0042] Figure 4 It is a schematic diagram of the structure of the electronic device of the present invention.

[0043] The specific implementation modes of the present invention are further described in detail below with reference to the accompanying drawings. DETAILED DESCRIPTION

[0044] The technical solution provided by the present invention is described in more detail below in conjunction with the accompanying drawings. The description of the exemplary implementation force is merely illustrative and is by no means intended to limit the present disclosure and its application or use. The present disclosure can be implemented in many different forms and is not limited to the embodiments described herein. These embodiments are provided to make the present disclosure thorough and complete and to fully express the scope of the present disclosure to those skilled in the art. It should be noted that unless otherwise specifically stated, the relative arrangement of the parts and steps, and the composition of the materials described in these embodiments. Numerical expressions and numerical values ​​should be interpreted as being merely exemplary and not as limitations.

[0045] The serial numbers assigned to the components in the present disclosure, such as "first", "second", etc., are only used to distinguish the objects described and do not have any order or technical meaning. The "connection" and "coupling" mentioned in the present invention, unless otherwise specified, include direct and indirect. In the description of the present invention, it should be understood that the directions or positional relationships indicated by the terms "upper", "lower", "front", "rear", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside", "clockwise", "counterclockwise", etc. are based on the directions or positional relationships shown in the accompanying drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the referred device or element must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as a limitation to the invention.

[0046] In the present disclosure, when a specific device is described as being located between a first device and a second device, there may or may not be an intermediate device between the specific device and the first device or the second device. When a specific device is described as being connected to other devices, the specific device may be directly connected to the other devices without an intermediate device, or may not be directly connected to the other devices but have an intermediate device.

[0047] All terms (including technical terms or scientific terms) used in the present disclosure have the same meanings as those understood by ordinary technicians in the field to which the present disclosure belongs, unless otherwise specifically defined. It should also be understood that terms defined in general dictionaries should be interpreted as having meanings consistent with their meanings in the context of the relevant technology, and should not be interpreted in an idealized or extremely formal sense, unless explicitly defined herein.

[0048] Technologies, methods, and equipment known to ordinary technicians in the relevant art may not be discussed in detail, but where appropriate, the technologies, methods, and equipment should be considered part of the specification.

[0049] The dimensions of the various parts shown in the drawings are not drawn according to the actual proportional relationship. In the drawings, the same figure marks are attached to common structural elements or structural elements of the same type, and their repeated descriptions are appropriately omitted.

[0050] This application proposes an edge computing-based intelligent information entry solution to address the three problems of high investment, effect affected by input, and insufficient personalized support in the background technology. The purpose of this solution is to achieve:

[0051] The real-time matching calculation program is directly deployed on the customer service's work computer and runs without adding additional hardware costs. The decentralized design eliminates server resource bottlenecks, network bandwidth bottlenecks, and data storage bottlenecks in business concurrency.

[0052] The samples are limited to the historical data entered by the user, and no other data is introduced. Users can maintain their usual expression habits to search for information, thus improving the search accuracy.

[0053] For high-quality / inaccurate retrieval results, you can directly mark them and apply them to your own retrieval model. The model is updated in real time and will not affect other people's model results.

[0054] See also Figure 1-Figure 4 .

[0055] The present application provides an intelligent information entry method, comprising:

[0056] Obtain customer problem information and confirm the scope of the problem;

[0057] Retrieve similar reply recommendation entries containing the keyword from the historical sample library through the keyword;

[0058] Sort similar reply recommendation terms according to weight information, where the higher the weight, the higher the ranking, otherwise the lower the ranking;

[0059] Output the sorted results.

[0060] In some embodiments, the customer service can use his / her computer to extract his / her own form input information, instant messaging chat information, email information, etc. from the server as historical sample data. The customer service can manage and maintain this information in the following ways:

[0061] A local cache system can be established

[0062] Real-time synchronization: Develop a local cache system that automatically synchronizes with the server. The system can regularly pull the latest data to the customer service staff's local computer to ensure the timeliness and integrity of the data.

[0063] Fast access: Through local caching, customer service can quickly access the required information and reduce the time waiting for server responses, especially in the case of poor network connection, they can still maintain high work efficiency.

[0064] Data security: Ensure that local systems have data encryption and security protection measures in place to prevent sensitive information from being leaked or accessed by unauthorized third parties.

[0065] You can also use cloud storage services to manage:

[0066] Cloud backup: Store data in the cloud, such as Google Drive, Dropbox, etc., so that customer service can access historical sample data from any location.

[0067] Sharing and collaboration: Cloud storage services usually have good sharing and collaboration functions, which facilitate information sharing and collaborative work among team members and improve the overall work efficiency of the team.

[0068] Data recovery: In the event of data loss or system failure, cloud storage can serve as an effective means of data recovery to ensure business continuity.

[0069] In addition, the development of customized data management tools can be designed with the following functions:

[0070] Personalized interface: Based on the specific needs of customer service, develop data management tools with personalized user interfaces to make information retrieval and display more intuitive and easy to understand.

[0071] Function customization: Integrates data extraction, analysis, reporting and other functions into one, provides one-click operation, simplifies data processing flow and improves work efficiency.

[0072] Technical Support: Establish a technical support team to provide timely technical support and training to customer service personnel when using customized tools to ensure the effective use of the tools.

[0073] In addition, you can also use existing CRM software for management:

[0074] Customer information integration: Ready-made CRM (customer relationship management) software can integrate various types of customer information including communication records, email interactions, form filling, etc., to facilitate unified management.

[0075] Automation functions: CRM software usually includes automation functions, such as automatic recording of conversations and generation of reports, which reduces the workload of customer service staff.

[0076] Data analysis: By utilizing the powerful data analysis capabilities of CRM software, customer service can better understand customer needs and behavior patterns, and improve service quality and customer satisfaction.

[0077] Implementing data extraction scripts

[0078] Automated extraction: Write scripts to automatically extract required data from the server database, which can be run at set time intervals to ensure the freshness and accuracy of the data.

[0079] Customized output: According to customer service needs, the script can output data reports in different formats, such as CSV, Excel, etc., to facilitate subsequent processing and analysis.

[0080] Easy to deploy and maintain: Scripts are generally easy to deploy and maintain, do not require complex technical support, can be put into use quickly and adjusted and optimized according to actual conditions.

[0081] In some embodiments, existing technologies may be used to retrieve recommended terms from a historical sample library using keywords. Retrieving similar reply recommended terms containing the keywords from the historical sample library using keywords helps provide users with more relevant and accurate information.

[0082] For example, you can use efficient keyword indexing tools to search:

[0083] Inverted index construction: In order to quickly retrieve entries containing specific keywords, an inverted index can be constructed for the entries in the historical sample library. The inverted index records the list of entries where each keyword appears, enabling fast search.

[0084] N-gram index: In addition to indexing based on a single keyword, you can also consider using N-gram index, which can match terms containing exact phrases. This is especially effective for retrieving replies containing specific expressions or phrases.

[0085] Incremental index update: As new data is added, the index needs to be updated regularly. Designing an efficient incremental index update mechanism can ensure the timeliness and accuracy of the index while reducing the consumption of system resources.

[0086] You can also optimize the keyword matching algorithm.

[0087] Fuzzy matching and synonym expansion: The introduction of fuzzy matching technology can retrieve relevant entries even if the keywords entered by the user are not completely consistent with the keywords in the index. At the same time, the introduction of a synonym dictionary can expand the matching range of keywords.

[0088] Relevance scoring mechanism: When searching, we should not only consider whether the entry contains keywords, but also evaluate the relevance of the entry to the keywords. We can use factors such as word frequency, entry length, and the position of keywords in the entry to comprehensively score.

[0089] Context-aware matching: Considers the user's query context, such as the user's historical queries, geographic location, and other information, and weights the search results so that terms that are more relevant to the user context are recommended first.

[0090] Introducing natural language processing technology.

[0091] Semantic analysis: Use natural language processing techniques, such as word sense disambiguation and semantic role labeling, to understand the exact meaning of user queries and provide more accurate search results.

[0092] Topic modeling: Apply topic modeling techniques, such as LDA (Latent Dirichlet Allocation), to analyze the topic distribution of terms in the historical sample library. When users query, terms related to the query topic can be recommended.

[0093] Sentiment analysis: By performing sentiment analysis on the entries, we can understand their emotional tendencies.

[0094] You can also use machine learning to improve the quality of recommendations:

[0095] Collaborative filtering: Collaborative filtering technology is used to recommend relevant reply terms based on the user's historical behavior and the behavior patterns of similar users, further improving the personalization of recommendations.

[0096] Deep learning models: Use deep learning models, such as convolutional neural networks or recurrent neural networks, to learn feature representations of terms and use these representations to retrieve similar response terms.

[0097] Online learning: The system can conduct online learning based on real-time feedback from users and continuously adjust recommendation strategies to adapt to changing user needs.

[0098] In summary, the above existing methods can be used to retrieve similar reply recommendation entries containing specific keywords from the historical sample library to provide users with more relevant, accurate and personalized information services.

[0099] In some embodiments, after a customer asks a question, the customer service staff answers it in the form of text input. The method of sorting similar reply recommendation entries according to weight information includes sorting calculations based on the number of matching texts and the most recent sending time as weights, and the calculation formula is:

[0100]

[0101] Among them, F(t,n) represents the final score, t represents the difference between the sending time of the matching historical entry and the current time of the system. The smaller the value, the higher the matching degree of the output result; n represents the number of times the search term appears in the historical sent text.

[0102] For example, the following content is the content sent by a customer service in the recent period (serial number, content, time)

[0103] 1. Hello, I am Xiao Liu, a customer service specialist. I am happy to serve you. How can I help you?

[0104] Time: July 13, 2024, 12:00:00 noon

[0105] 2. OK, when was your last payment? I will check if the payment has been received. The payment time is: July 13, 2024, 12:01:00 p.m.

[0106] 3. Thank you for your support. Ping An customer service is committed to caring for customers. Thank you for your consultation. Please rate my service.

[0107] Time: July 13, 2024, 12:02:00 p.m.

[0108] 4. Hello, I am Xiao Liu, a customer service specialist. I am happy to serve you. How can I help you?

[0109] Time: July 13, 2024, 15:00:00

[0110] 5. Thank you for your support. Ping An customer service is committed to caring for customers. Thank you for your consultation. Please rate my service.

[0111] Time: July 14, 2024, 16:00:00

[0112] Assume that the current time is July 31, 2024, 13:00 p.m., and the customer service enters the word "you". The system needs to calculate and recommend the ranking of the above 5 sentences and bring them into the formula

[0113] For the first sentence, F(t,n)=2*2.71(-1558989) / [(1+2.71(-1558989))*(1+2.71(-1558989))]=0.83

[0114] / / This sentence appears twice, n=2, t is the difference from the current time (the value is converted to seconds), and the constant e is the base of the natural logarithm = 2.7182818459...

[0115] For the second sentence, F(t,n)=1*2.71(-1558989) / [(1+2.71(-1558900))*(1+2.71(-1558900))]=0.41 ... ...

[0118] For the fifth sentence, F(t,n)=1*2.71(-1472500) / [(1+2.71(-1472500))*(1+2.71(-1472500))]=0.85

[0119] And so on.

[0120] F(t,n) is the final score. The highest score has the highest match, so the fifth sentence is recommended first.

[0121] In some embodiments, in order to improve the efficiency of answering customer questions, when customer service answers questions raised by customers by entering text, a quick reply tool can be developed or integrated on the client side, allowing customer service to quickly enter common answers through preset shortcut phrases or macro commands, saving time.

[0122] Voice recognition input is also possible: with the introduction of voice recognition technology, customer service can answer questions by voice input, and the system will automatically convert it into text, which can reduce typing time and improve work efficiency.

[0123] Use of entry templates: Design customizable entry templates. Customer service staff can choose the corresponding template according to different question types and quickly generate answers.

[0124] Natural language processing techniques can also be applied.

[0125] Automatic question-and-answer system: Using natural language processing technology, we develop an automatic question-and-answer system that can automatically generate answers to simple and common questions, reducing the workload of customer service.

[0126] Intent recognition and sentiment analysis: Through intent recognition and sentiment analysis, the system can more accurately understand the needs and emotions behind customers' questions, and assist customer service to provide more precise answers.

[0127] Text generation assistance: Use text generation assistance tools, such as intelligent writing assistants, to help customer service staff quickly conceive and generate answer texts to ensure the coherence and professionalism of the answers.

[0128] Improve customer service training and support

[0129] Regular training: Provide regular product knowledge and communication skills training to the customer service team to ensure that they can answer customer questions accurately and professionally.

[0130] Real-time support: Provide real-time support during the customer service process, such as expert consultation, team collaboration, etc., so that you can get instant help when dealing with complex issues.

[0131] Feedback and improvement loop: Establish a monitoring and feedback mechanism for the quality of customer service responses, and continuously improve response quality and service processes through customer satisfaction surveys and internal reviews.

[0132] Personalized service: Provide personalized answers and services based on the customer's past interaction records and preferences to improve customer satisfaction.

[0133] Clear answer format: Use clear format and concise language to organize answers so that customers can easily understand the information and improve the overall service experience.

[0134] In some embodiments, the output ranking results include each recommended term and its corresponding usage count, most recent usage time, etc.

[0135] In some embodiments, the method further includes filtering the search results to filter out the results that match the error set, and for the matching results in the star-marked search set (which are common terms), placing them at the top of the output ranking for recommendation.

[0136] After the method outputs the ranking results, when the top-ranked reply term is used, its usage count and usage time are updated to the historical sample library. The following factors can be considered during the update process:

[0137] 1. Establish an automated update mechanism

[0138] Trigger mechanism: Set a trigger between the user interface operation and the database. Once the user selects a reply term, the update mechanism is automatically triggered.

[0139] Transactional operations: Ensure that update operations are transactional, that is, operations that update the number of usage times and usage times must either all succeed or all fail to ensure data consistency.

[0140] Real-time data synchronization: Real-time data synchronization technology is used to ensure that the usage count and usage time information between different systems and databases remain consistent.

[0141] 2. Optimize database design

[0142] Efficient indexing strategy: In order to quickly access and update entry data, index the usage count and most recent usage time fields in the database.

[0143] Data caching mechanism: Introduce a data caching mechanism, such as Redis, to reduce direct access to the database and improve data update efficiency.

[0144] Historical data archiving: Historical data is archived regularly to optimize database performance while retaining sufficient historical information for future analysis.

[0145] 3. Application data security and privacy protection

[0146] Access control: Implement strict data access control to ensure that only authorized systems and users can update the usage count and usage time in the database.

[0147] Data encryption: Sensitive data is encrypted. Even if the data is accessed, it can only be decrypted with a specific key to protect user privacy.

[0148] Audit log: Records all changes to usage count and usage time data to facilitate subsequent review and tracking of potential security issues.

[0149] In some embodiments, for the historical text recommendation finally given, the user ultimately manually determines whether to use it, and the selection result (selected, discarded, starred) is fed back to the engine search set as marked content and participates in the next calculation.

[0150] The following is an example of a consultation dialogue on the time limit for compensation for injury without a person to be injured to illustrate the main steps and processes of the method of the present invention:

[0151] Scenario description: The customer purchased Ping An auto insurance. After the accident, he completed a series of operations such as reporting the case, assessing the damage, and sending the car for repair according to the process. Now he is waiting for the claim payment to be received. The customer initiated a question about the time limit for payment on the Ping An Good Car Owner APP, and the Ping An claims specialist responded online.

[0152] User operation: After verifying through a simple online conversation that the client's case involved no personal injury, Ping An's claims specialist confirmed the scope of the issue and searched for historical information (QA column) that he had processed and sent by entering the keyword "personal injury compensation received", hoping to quickly find a similar response.

[0153] Retrieve, sort and output the results, as shown in the attached Figure 2 As shown: Through the Chinese word segmenter, the keywords are split into three words: "personal injury", "compensation", and "account received" for retrieval

[0154] a) Personal injury and compensation, the two keywords have successful matching historical message records (QA column, matching words are marked in red, matching entries are 1, 2, 4, 5)

[0155] b) According to the most recent sending time and number of sending corresponding to the content of the QA column, the sorting is calculated by the search algorithm formula. The sorting results are as follows:

[0156] i. Entry 1, "《No one injured, quick compensation...》", 10 recent times, most recent on May 6

[0157] ii. Entry 4, "We are... formulating compensation...", 5 recent times, most recent on May 4

[0158] iii. Entry 2, "《Injury Compensation Standards ...", 3 recent times, most recent on May 1

[0159] iv. Entry 5, "Mr. Wang, last week...compensation...", 1 recent time, the most recent time was 9:55 on the same day (that day was May 7)

[0160] In this case, terms 1, 2, 4, and 5 match the wrong search set, and the order remains unchanged.

[0161] There are star tags on entries 1 and 2, and the tags match, so the weight is increased and entry 2 rises to the second place.

[0162] The final ranking recommendation is shown in "Dialogue Example" for customer service to choose.

[0163] In addition, the customer service confirms that the content of entry 1 meets the current question based on the context, and after clicking, it covers the input box and sends it to the customer. The number of times the entry is used is increased by 1, and the most recent time of use is also updated, which will be used in the next recommendation calculation.

[0164] You can also do the following:

[0165] In the recommended terms, for content that is not universal (such as term 5, which is only applicable to Mr. Wang and not to other customers), you can mark it as abandoned, and the term will be added to the error set and will not appear next time. Specifically, you can use the following methods to optimize:

[0166] 1. Establish a management mechanism for abandoned entries

[0167] Deprecation standards: Clearly define what content is not universally applicable and needs to be deprecated, such as erroneous information, outdated information, or information that is not relevant to the user community.

[0168] Abandonment process design: Design a process for discarding entries, including the steps of identification, labeling, review, and final addition to the error set, to ensure the efficiency and accuracy of the process.

[0169] Automated processing: Develop automated tools to help identify and mark entries that need to be discarded, reduce manual intervention, and improve processing speed and accuracy.

[0170] 2. Optimize error set management

[0171] Error collection database: Establish a special error collection database to store entries marked as obsolete to facilitate subsequent management and query.

[0172] Regular review and update: Regularly review the entries in the error set to ensure that all obsolete entries are accurate and necessary, and promptly remove obsolete entries that are no longer needed.

[0173] Error set analysis: Statistical analysis is performed on the terms in the error set to identify the common characteristics of the discarded terms in order to optimize the recommendation algorithm and reduce the possibility of similar problems in the future.

[0174] 3. Strengthen user feedback mechanism

[0175] User reporting function: Provides user reporting function, allowing users to directly participate in the identification process of abandoned entries, improving the accuracy of the system and user participation.

[0176] User feedback loop: Encourage users to provide feedback on recommendation results, such as ratings or comments, and use this information to adjust recommendation strategies and reduce the appearance of content that is not universal.

[0177] Transparency: Clearly show users which entries have been abandoned and the reasons for abandonment, increasing system transparency and user trust.

[0178] 4. Apply machine learning techniques

[0179] Pattern recognition: Use machine learning technology to identify patterns and features that make content non-universal, and automatically mark and discard such content.

[0180] Self-learning ability: Through technologies such as deep learning, the system has the ability to self-learn, constantly learn from discarded entries, and improve the accuracy of future recommendations.

[0181] Multi-model fusion: Combine the judgments of multiple machine learning models to conduct a comprehensive assessment on whether to abandon a certain entry, thereby improving the accuracy of decision-making.

[0182] 5. Enhance system adaptability and flexibility

[0183] Dynamic update mechanism: Establish a dynamic update mechanism to ensure that the recommendation system can respond to entry abandonment operations in a timely manner and update the recommendation list in real time.

[0184] Flexible recovery strategy: For erroneously abandoned entries, design flexible recovery strategies to allow them to re-enter the recommendation process under certain conditions.

[0185] Adaptive algorithm optimization: Continuously optimize the recommendation algorithm to make it better adapt to user needs and content changes, and reduce the number of entries that need to be discarded in the future.

[0186] The above method can more effectively manage content that is not universal and improve recommendation quality and user experience.

[0187] In the recommended entries, for content that is ranked low but considered to be universal, you can click the star and configure tags to increase its weight.

[0188] In some embodiments, boosting weight is an important user interaction and system feedback mechanism. This mechanism can not only improve user engagement, but also provide valuable signals for the recommendation system to optimize and adjust the recommendation algorithm. The specific analysis is as follows:

[0189] User interaction design

[0190] Intuitive interface: Make sure the functionality of adding stars and configuring labels is obvious and easy to operate in the user interface so that users can easily annotate.

[0191] Instant feedback: After the user stars and tags the recommended entry, the system should provide immediate feedback, such as by changing the display style of the entry or popping up a prompt message to confirm that the operation has taken effect.

[0192] Diversified labels: Provide diverse labeling options, allowing users to segment content based on its attributes, categories, or personal preferences, thereby improving the accuracy of annotation.

[0193] Weight adjustment mechanism

[0194] Dynamic weight adjustment: The system should dynamically adjust the weight of related terms based on the user's star and tag configuration behavior to improve their ranking in the recommendation list.

[0195] Comprehensive consideration of multiple factors: When adjusting weights, multiple factors such as the user's annotation behavior, the initial ranking position of the entry, and the frequency with which the entry is annotated are considered to ensure the rationality of the weight adjustment.

[0196] Algorithm optimization and learning

[0197] Feature extraction and utilization: Extract features from the user's annotation behavior, such as tag type, annotation time, user's historical behavior, etc., and incorporate these features into the recommendation algorithm to improve the accuracy of the recommendation.

[0198] Model iterative update: Based on the user's labeled data, the recommendation model is regularly iteratively updated to make it better adapt to changes in user preferences.

[0199] Feedback loops and continuous improvement

[0200] User feedback collection: Encourage users to provide more feedback, such as through ratings and comments, to collect more user opinions on the quality of recommendations.

[0201] System self-evaluation: Regularly self-evaluate the recommendation system, analyze the relationship between user annotation behavior and recommendation effect, and identify room for improvement.

[0202] Personalization and diversity guarantee

[0203] Enhanced personalized recommendations: Through the user's star and tag configuration behavior, personalized recommendations are further enhanced to ensure that each user can obtain content that is more in line with their preferences.

[0204] Diversity of recommendation results: While increasing the weight, maintain the diversity of recommendation results to avoid over-concentration on a certain type of frequently annotated content.

[0205] In addition, after understanding the above content, there are a few points to note:

[0206] In-depth analysis of user behavior: Through in-depth analysis of user star and tag configuration behaviors, we can discover the deep structure and patterns of user preferences, and provide more accurate personalized recommendation basis for the recommendation system.

[0207] Enhanced explainability of recommendations: Through the labeling system, users can be provided with more information about why a certain content is recommended, which enhances the transparency and explainability of the recommendations and improves user trust.

[0208] Cultivating a community co-building atmosphere: Encouraging users to participate in the labeling and optimization of recommended content can cultivate a community co-building atmosphere and increase user stickiness and activity.

[0209] In general, comprehensive consideration is given to user interaction design, weight adjustment mechanism, algorithm optimization and learning, feedback loop and continuous improvement, and personalization and diversity assurance. Through such design and implementation, the recommendation system can not only more accurately capture and adapt to the personalized needs of users, but also improve user participation and promote continuous optimization and improvement of the system.

[0210] The search engine of this method is deployed on the client computer, the server resources will not increase due to the increase in the number of users, and the hardware cost is greatly reduced.

[0211] Each new input by the customer service staff of the present invention will immediately generate new recommendations for their own models (for example, if only one word "claim" is input, there are too many matching entries and no differentiation. After adding "arrival", no recommendation is found, indicating that "arrival" is not suitable as a keyword. Then the recommended combination of "claim" + "personal injury" is changed to the most appropriate one). In this way, users can try multiple keyword combinations in a very short time and quickly find suitable recommendations in the current scenario. And by sending a confirmation, the weight of the selected entry in the next matching recommendation is strengthened.

[0212] The training and labeling of this method are integrated into the regular work process of customer service, without interfering with their normal work content. At the same time, it does not require the addition of additional professional data labeling personnel. Compared with most AI product solutions, the labor cost is greatly improved.

[0213] The data retrieved by this method are all contents sent by the users themselves, which is equivalent to having an exclusive knowledge base. Keywords are used based on memory and usage habits, reducing the cost of learning how to ask questions.

[0214] This embodiment also provides an information intelligent entry system, including:

[0215] Customer problem acquisition unit, used to obtain customer problem information and confirm the problem scope;

[0216] A recommended term retrieval unit, used to retrieve similar reply recommended terms containing the keyword from a historical sample library by using the keyword;

[0217] A recommended term ranking processing unit, used to rank similar reply recommended terms according to weight information, wherein the higher the weight, the higher the ranking, otherwise the lower the ranking;

[0218] The result output unit is used to output the sorting result.

[0219] In some embodiments, the main function of the customer question acquisition unit is to receive and understand the customer's questions. It needs to be able to parse natural language, understand the meaning of the question, and classify the question into the corresponding category. Some natural language processing (NLP) technologies such as semantic analysis and entity recognition need to be used.

[0220] In some embodiments, the customer questions that the customer question acquisition unit may acquire include but are not limited to:

[0221] What does car insurance include?

[0222] Specifically, car insurance mainly includes compulsory traffic insurance and commercial insurance. Compulsory traffic insurance is a mandatory insurance purchased by the state, which is mainly used to compensate for personal injury and property damage caused to third parties. Commercial insurance is insurance purchased voluntarily by car owners, including vehicle loss insurance, third-party liability insurance, vehicle theft and robbery insurance, vehicle occupant liability insurance, glass breakage insurance, spontaneous combustion loss insurance, scratch loss insurance, etc.

[0223] How to buy car insurance?

[0224] The answers to specific questions may involve purchasing auto insurance through the insurance company's official website, mobile APP, phone or insurance company's business outlets. When purchasing, you need to provide relevant information about the vehicle, such as license plate number, vehicle model, purchase date, etc.

[0225] How is the price of car insurance calculated?

[0226] The answer to the specific question may involve that the price of auto insurance is mainly calculated based on factors such as the type of vehicle, purchase price, age, and driving experience of the driver. Different insurance companies may have different rate tables, and you can consult the insurance company for specific prices.

[0227] What is the claims process for car insurance?

[0228] The content that needs to be answered may involve that when a vehicle accident occurs, the owner needs to immediately call the police and contact the insurance company to report the accident. The insurance company will send a surveyor to investigate the accident site and make a claim based on the results of the investigation. The owner needs to provide relevant supporting documents, such as the accident identification certificate, repair invoice, etc.

[0229] What does the deductible mean in car insurance?

[0230] The answer to the specific question may involve the deductible, which refers to the portion that the insurance company does not bear when making a claim. For example, if the deductible is 500 yuan, then when making a claim, the insurance company will only bear the portion that exceeds 500 yuan. Setting a deductible can reduce the risk of the insurance company and also reduce the premium of the car owner.

[0231] Can car insurance be renewed early?

[0232] The answers to specific questions may include: Car insurance can usually be renewed one month in advance. Renewing in advance can avoid insurance gaps caused by forgetting to renew, and ensure that the vehicle is always covered by insurance.

[0233] Can car insurance be refunded?

[0234] The answers to specific questions may include: If no claim is made for car insurance, you can cancel the policy according to the insurance company's regulations. However, it should be noted that canceling the policy may result in loss of premiums, and the specific cancellation amount can be consulted with the insurance company.

[0235] The customer problem acquisition unit can be a system or other tool unit used to collect and analyze customer problems and needs. The main purpose of this unit is to help the company better understand customer needs and problems.

[0236] The task of the recommended term retrieval unit is to find relevant replies in the historical sample library based on the keywords of the customer's question. This involves text search and information retrieval technologies, such as inverted index, TF-IDF, etc.

[0237] In some other embodiments, extracting keywords from customer questions is the first step in retrieval of recommended terms. Natural language processing techniques, such as semantic analysis and entity recognition, are usually involved to ensure accurate understanding of the meaning of the question and extract core keywords. After extraction, these keywords are used to search in a historical database or sample library. This historical sample library may include previous conversation records, FAQs, and other knowledge base content.

[0238] The recommended term ranking processing unit needs to rank the retrieved terms according to certain weight information. The weight calculation may take into account multiple factors. In some embodiments, it may also be necessary to use machine learning or deep learning algorithms, such as ranking learning, deep learning ranking models, etc.

[0239] The result output unit displays the sorted entries to the user in some form.

[0240] In some embodiments, the result output unit is an important part of the system responsible for displaying the sorted entries to the user in some form.

[0241] In recommendation systems, computing platforms, or any software application that needs to give corresponding outputs based on inputs, the result output unit plays a vital role. It not only ensures that users can clearly see and understand the data processed by the system, but also directly affects the user experience and the usability of the system. The specific introduction is as follows:

[0242] Results display method

[0243] Graphical User Interface: Most modern software and systems use a graphical user interface (GUI) to display results. This method is intuitive and easy to understand, and users can easily view and manipulate results by clicking, sliding, etc. For example, in an e-commerce recommendation system, the user's question and the related products recommended by the system will be directly displayed on the web page, and users can easily browse and select.

[0244] Text output: In some command line tools or old systems, the results may be output to the console or text files in plain text. This method is simple and direct, and is suitable for scenarios where there is no need for a graphical interface, such as data analysis scripts running on a server.

[0245] Data formatting

[0246] Tabular format: In spreadsheet software such as Excel, results are often displayed in tabular format, with cells showing the calculation formula and the corresponding result directly. Users can enter the formula in one column and the calculation results are displayed in the adjacent column, which makes it easy to compare and analyze large amounts of data.

[0247] Chart form: For data that needs to be visualized, converting it into a chart can more intuitively show trends, distributions, and relationships. For example, in data analysis software, the user's query results can be dynamically generated and displayed in the form of bar charts, line charts, etc.

[0248] Interactive Operations

[0249] Filter and sort: Allow users to filter and sort the results again to meet personalized viewing needs. For example, in a customer support system, users can sort by relevance and posting time to quickly find the most suitable response.

[0250] Detailed View: When the results are displayed in a list format, each record can usually be further expanded to view the details. For example, when a user clicks on a recommended term in the search results, a detailed window can pop up to display the full content of the term and user feedback information.

[0251] User feedback mechanism

[0252] Feedback collection: The result output interface usually provides options for user feedback, such as rating, commenting, or reporting functions. These feedbacks will be sent directly back to the system to help improve the recommendation algorithm and user experience.

[0253] Real-time update: Based on user feedback and behavior, the result output interface can update the recommended content in real time. For example, if a user frequently ignores a certain type of recommendation, the system will gradually reduce the frequency of such recommendations.

[0254] Multi-terminal adaptability

[0255] Mobile optimization: With the popularity of mobile devices, the result output unit needs to consider the display effect on different devices. This includes adjusting the layout, font size and interaction method to ensure a good user experience on devices such as mobile phones and tablets.

[0256] Cross-platform support: The output of results should not only consider the Web environment, but also how to maintain consistency and fluency in multiple platforms such as desktop applications and mobile applications. This may require the use of cross-platform development frameworks and technologies.

[0257] Security and Privacy

[0258] Permission management: The display of results will be different according to the user's permissions. Users with high permissions can see more detailed information and advanced functions, while ordinary users can only see basic results.

[0259] Privacy desensitization: When outputting results involving sensitive information, the system needs to desensitize some of the information to ensure that the user's personal privacy is not leaked. For example, in medical and health applications, the patient's personal information will be hidden or replaced with non-sensitive identifiers.

[0260] Internationalization and Localization

[0261] Multi-language support: Global applications need to support multiple languages, and the result output unit should display the corresponding language interface according to the language preference set by the user. This means that the interface text, prompt information, etc. need to have corresponding translated versions.

[0262] Local cultural adaptation: In addition to language, regional cultural differences must also be considered, such as date format, currency unit, etc. The result output unit should be appropriately localized according to the user's region.

[0263] Performance and Optimization

[0264] Response speed: The loading and response speed of the result output unit directly affects the user experience. By optimizing the front-end code, using lazy loading and asynchronous request techniques, the response speed of the interface can be significantly improved.

[0265] Scalability: When designing the result output unit, it is also necessary to consider the possible increase in functions and data volume in the future to ensure that it has good scalability. This includes modular design, API abstraction, and the use of high-performance data processing technology.

[0266] In summary, as the key bridge between users and the system, the design and implementation of the result output unit requires careful consideration of many factors to ensure that users can obtain the required information clearly and conveniently. Through continuous optimization and upgrading, this unit can not only improve the user experience, but also effectively improve the value and satisfaction of the overall system.

[0267] In some embodiments, the system further includes a filtering unit, which is used to filter the search results and place the matching results at the top of the output ranking for recommendation.

[0268] In some embodiments, the system further comprises a historical sample updating unit, which is used to update the usage count and usage time of the top-ranked reply term to the historical sample library after outputting the sorting result and after the reply term is used.

[0269] The present application also provides a computer storage medium, which stores multiple instructions, which are suitable for being loaded by a processor and executing the above method steps. Computer storage medium is a physical device for storing data, such as a hard disk, an optical disk, a USB flash drive, etc. These storage media can store various types of data, including text, images, audio, video, and program code. The storage medium stores some programs or scripts, which contain a series of instructions suitable for being loaded by a processor and executing the above method steps, and these instructions can be read, parsed, and executed by the computer's processor.

[0270] For example, an executable file (such as an .exe file) stored on a USB flash drive contains a series of instructions. When this file is copied to a computer and run, the computer's processor will operate according to the instructions in the file to achieve a specific function or task. The specific execution process can be found in the detailed description of the system and method, which will not be repeated here.

[0271] The present application also provides an electronic device, including a processor and a memory, wherein the memory stores a computer program, and the computer program is suitable for being loaded by the processor and executing the above method steps.

[0272] See also Figure 4 , is a schematic diagram of the main hardware structure of the electronic device provided in the embodiment of the present application. Figure 3 As shown,

[0273] The customer voice demand classification device 300 may include: at least one processor 301 , at least one network interface 304 , a user interface 303 , a memory 305 , and at least one communication bus 302 .

[0274] The communication bus 302 is used to realize the connection and communication between these components.

[0275] The user interface 303 may include a touch screen and a camera.

[0276] The network interface 304 may optionally include but is not limited to a standard wired interface, a wireless interface (such as a WI-FI interface, Bluetooth, a traffic card, etc.).

[0277] Among them, the processor 301 may include one or more processing cores. The processor 301 uses various interfaces and lines to connect the various parts of the entire multi-tenant mode data acquisition system 300, and executes various functions and processes data of the multi-tenant mode data acquisition system 300 by running or executing instructions, programs, code sets or instruction sets stored in the memory 305, and calling data stored in the memory 305. Optionally, the processor 301 can be implemented in at least one hardware form of digital signal processing (DSP), field programmable gate array (FPGA), and programmable logic array (PLA). The processor 301 can integrate one or more combinations of a central processing unit (CPU), a graphics processing unit (GPU), and a modem. Among them, the CPU mainly processes the operating system, user interface, and application programs; the GPU is responsible for rendering and drawing the content to be displayed on the display screen; and the modem is used to handle wireless communication. It is understandable that the above-mentioned modem may not be integrated into the processor 301, but may be implemented by a separate chip.

[0278] Among them, the memory 305 may include a random access memory (Random Access Memory, RAM) and may also include a read-only memory (Read-Only Memory). Optionally, the memory 305 includes a non-transitory computer-readable storage medium. The memory 305 can be used to store instructions, programs, codes, code sets or instruction sets. The memory 305 may include a program storage area and a data storage area, wherein the program storage area may store instructions for implementing an operating system, instructions for at least one function (such as a touch function, a sound playback function, an image playback function, etc.), instructions for implementing the above-mentioned various method embodiments, etc.; the data storage area may store data involved in the above-mentioned various method embodiments, etc. The memory 305 may optionally be at least one storage device located away from the aforementioned processor 301. As a computer storage medium, the memory 305 may include an operating system, a network communication module, a user interface module and an application. In Figure 4 In the system 300 shown, the user interface 303 is mainly used to provide an input interface for the user and obtain the data input by the user; and the processor 301 can be used to call the application stored in the memory 305 and specifically execute the following steps: Figure 1 The specific process can be referred to Figure 1 As shown, no further description is given here.

[0279] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiments can be implemented by instructing the relevant hardware through a computer program, and the program can be stored in a computer-readable storage medium along with the main application program. When the program is executed, it can include the processes of the embodiments of the above-mentioned methods. The storage medium can be cloud storage, a disk, an optical disk, a read-only storage memory, or a random access memory, etc.

[0280] The features or feature combinations mentioned above in the specification, drawings and claims, as long as they are meaningful within the scope of the present invention and do not contradict each other, can be used in any combination or alone. The advantages and features described for the system provided by the present invention are applicable to the method and computer-readable storage medium provided by the present invention in a corresponding manner, and vice versa.

[0281] It should also be noted that the terms "comprises", "includes" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, product or apparatus that includes a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, product or apparatus. In the absence of further restrictions, the elements defined by the sentence "comprises one..." do not exclude the presence of other identical elements in the process, method, product or apparatus that includes the elements.

[0282] The above description of the disclosed embodiments enables one skilled in the art to implement or use the present invention. Various modifications to these embodiments will be apparent to one skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but rather to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. An intelligent information entry method, characterized in that: include: Obtain customer problem information and confirm the scope of the problem; Retrieve similar reply recommendation entries containing the keyword from the historical sample library through the keyword; Sort similar reply recommendation terms according to weight information, where the higher the weight, the higher the ranking, otherwise the lower the ranking; Output the sorted results.

2. The method for intelligent information entry according to claim 1, characterized in that: The method for sorting similar reply recommendation terms according to weight information includes sorting calculation based on the number of matching texts and the most recent sending time. The calculation formula is: Among them, F(t,n) represents the ranking output according to the matching degree, t represents the difference between the sending time of the matching historical entry and the current time of the system. The smaller the value, the higher the matching degree of the output result; n represents the number of times the search term appears in the historical sent text, and e represents the exponential function.

3. The intelligent information entry method according to claim 1, characterized in that: The output sorting results include each recommended term and its corresponding usage count and most recent usage time.

4. The method for intelligent information entry according to claim 2, characterized in that: The method further includes filtering the search results, and placing the matching results at the top of the output ranking for recommendation.

5. The method for intelligent information entry according to claim 1, characterized in that: After outputting the ranking result, the method updates the number of times and the time of use of the top reply term to the historical sample library after the reply term is used.

6. An intelligent information entry system, characterized in that: include: Customer problem acquisition unit, used to obtain customer problem information and confirm the problem scope; A recommended term retrieval unit, used to retrieve similar reply recommended terms containing the keyword from a historical sample library by using the keyword; A recommended term ranking processing unit, used to rank similar reply recommended terms according to weight information, wherein the higher the weight, the higher the ranking, otherwise the lower the ranking; The result output unit is used to output the sorting result.

7. The intelligent information entry system according to claim 6, characterized in that: The system further comprises a filtering unit, which is used to filter the search results and to recommend the matching results to the top in the output ranking.

8. The intelligent information entry system according to claim 6, characterized in that: The system further comprises a historical sample updating unit, which is used to update the usage count and usage time of the top-ranked reply term to the historical sample library after the sorting result is output.

9. A computer storage medium, characterized in that: The computer storage medium stores a plurality of instructions, which are suitable for being loaded by a processor and executing the method steps according to any one of claims 1-5.

10. An electronic device, characterized in that: The method comprises a processor and a memory, wherein the memory stores a computer program, and the computer program is suitable for being loaded by the processor and executing the method steps according to any one of claims 1 to 5.

Citation Information

Cited By

  • Data caching method and device, equipment, medium and product

    CN120179411A

  • A data caching method, device, equipment, medium and product

    CN120179411B