A 5G message conversation method and a 5G message conversation system

Abstract

The present invention relates to the field of 5G communication technology, and in particular to a 5G message conversation method and a 5G message conversation system. The 5G message conversation method includes: creating a new conversation according to a 5G message uploaded by a user; using the historical conversation record of the user as a basic conversation sample for the new conversation; processing the 5G message through natural language processing (NLP); identifying the conversation intention of the user in combination with the processing result of step S3 and the basic conversation sample; and pushing a conversation message to the user according to the conversation intention. The 5G message conversation system is used to implement the 5G message conversation method. It can improve the matching ability and accuracy of intelligent responses in the process of human-computer interaction, avoid inaccurate content of reply messages or deviation from the direction of the problem, and more accurately control the true intention of the user, thereby enhancing user stickiness, making intelligent replies more accurate, flexible, and personalized, promoting users to participate deeply in conversations, and improving the conversion rate of 5G messages in business applications.

Description

Technical Field

[0001] The present invention relates to the field of 5G communication technology, and in particular to a 5G message conversation method and a 5G message conversation system. Background Art

[0002] As operators' traditional SMS services can no longer meet the diverse needs of users due to their simple functions and limited experience, China's three major operators, China Telecom, China Mobile, and China Unicom, jointly released the "5G Message White Paper" on April 8, 2020, proposing the concept of "5G Message" for the first time, and calling on industry partners to work together to fully launch 5G message services for users as soon as possible.

[0003] As an upgrade of traditional SMS services, it breaks through the limitations of traditional SMS on message length and content format and builds a new information service portal.

[0004] In the existing technology, 5G message terminal users use chatbots provided by industry customers to interact with 5G messages. Currently, there are two main ways to implement chatbot interaction capabilities:

[0005] One is to set fixed keywords in advance, and realize the interaction of 5G messages through simple keyword regular matching after the message arrives. This method is simple and crude, and only establishes the mapping relationship between 5G messages and keywords in advance. Unmatched keywords are uniformly replied with default messages. The direct disadvantage is that the reply message content obtained by the user's interaction with the chatbot is inaccurate or deviates from the direction of the problem, resulting in poor experience of the end user's human-computer interaction capabilities of 5G messages.

[0006] Another method is to achieve intelligent response by introducing an external intelligent chatbot knowledge base. This method can only achieve a question-and-answer human-computer interaction mode. In addition, many messages of intelligent replies are relatively general, the reply content is relatively simple and lacks stickiness, and the interaction content is not humane enough.

[0007] These problems directly affect the market transformation of 5G messages. In view of this, this application is specially proposed. Summary of the invention

[0008] The first purpose of the present invention is to provide a 5G message conversation method, which can improve the matching ability and accuracy of intelligent responses in the process of human-computer interaction, avoid inaccurate content of reply messages or deviation from the direction of the problem, more accurately control the true intentions of users, enhance user stickiness, and make intelligent replies more accurate, flexible, and personalized, thereby promoting users to further participate in conversational communication, which helps to improve the conversion rate of 5G messages in business applications.

[0009] The second purpose of the present invention is to provide a 5G message conversation system, which can improve the matching ability and accuracy of intelligent responses in the process of human-computer interaction, avoid inaccurate content of reply messages or deviation from the direction of the problem, more accurately control the true intentions of users, enhance user stickiness, and make intelligent replies more accurate, flexible, and personalized, which promotes users to further participate in conversational communication and helps to improve the conversion rate of 5G messages in business applications.

[0010] The embodiment of the present invention is achieved as follows:

[0011] A 5G message session method, comprising the following steps:

[0012] S1: Create a new session based on the user's uplink 5G message;

[0013] S2: The user's historical session records are used as the basic session samples for the new session;

[0014] S3: Process 5G messages through natural language processing (NLP);

[0015] S4: identifying the user's conversation intention by combining the processing result of step S3 and the basic conversation sample;

[0016] S5: Pushing a conversation message to the user according to the conversation intent.

[0017] Furthermore, 5G messages include message content and behavior data.

[0018] Further, step S3 includes the following steps:

[0019] S31: performing word segmentation, stem extraction and word form restoration, part-of-speech tagging, named entity recognition and block standardization on the message content of the 5G message;

[0020] S32: Output the processing result.

[0021] Further, step S4 includes the following steps:

[0022] S41: Use the basic conversation samples as context samples of the neural network model;

[0023] S42: Using the processing result of step S3 as input sample data of the neural network model;

[0024] S43: Model evaluation using neural network models;

[0025] S44: output the evaluation result;

[0026] S45: Identify the user's session intention according to the evaluation result.

[0027] Furthermore, the neural network model includes:

[0028]

[0029] Among them: PP is an indicator for evaluating the effect of language model, which mainly estimates the probability of occurrence of a sentence based on each word, |V| represents the number of words in the corpus, t represents the tth word in the sample: T is the number of words in the sample, y(j,t) is the probability of the word in the corpus, and y^(j,t) is the predicted probability of the word.

[0030] Furthermore, the 5G message session method further includes the following steps:

[0031] S6: Save the user's historical conversation records as conversation material samples, one user corresponds to one conversation material sample;

[0032] S7: Setting a first relevance judgment rule, matching relevant conversation material samples according to the user's conversation intention and adding them to the basic conversation samples.

[0033] Furthermore, step S6 also includes: establishing classification rules to classify the conversation material samples;

[0034] Step S7 includes: setting a relevance judgment rule, matching a relevant conversation material sample category according to the user's conversation intention, and adding the conversation material samples under the category to the basic conversation sample.

[0035] Furthermore, the 5G message session method further includes the following steps:

[0036] S8: Setting a second relevance judgment rule, according to which related conversation material samples are associated with each other to form a material chain, matching related conversation material samples according to the user's conversation intention, and adding the material chain where the related conversation material samples are located to the basic conversation sample.

[0037] Furthermore, in step S6, classifying the conversation material samples includes:

[0038] Set a waiting time, and output the conversation material samples that cannot be accurately classified after the waiting time has passed, prompting manual classification;

[0039] For a conversation material sample that cannot be accurately classified, if a material chain is formed within the waiting time, the conversation material sample will no longer prompt manual classification.

[0040] A 5G message conversation system, comprising: a first conversation end, a communication platform and a second conversation end; the communication platform is used to establish a communication relationship between the first conversation end and the second conversation end; the first conversation end is used for uplinking 5G messages; the second conversation end is used to execute the 5G message conversation method as claimed in any one of claims 1 to 9 to complete the conversation with the first conversation end.

[0041] The beneficial effects of the technical solution of the embodiment of the present invention include:

[0042] The 5G message conversation method provided by the embodiment of the present invention uses the user's historical conversation records as the basic conversation samples for the new conversation in the 5G message human-computer interaction scenario application. After analyzing, processing, summarizing and reasoning the 5G messages through natural language processing (NLP), the contextual scene context of the user's historical conversation content is deeply mined and subjectively analyzed to identify the user's true conversation intention, and generate corresponding conversation messages according to the user's conversation intention and push them to the user.

[0043] Through this conversation management method, more user information can be extracted, effectively improving the matching ability and accuracy of intelligent responses during human-computer interaction, significantly enhancing user stickiness, and promoting users to further participate in conversational communication.

[0044] It can realize intelligent response of 5G messages based on session management, and support the intelligent response context environment of chatbot human-computer interaction through this conversation mode, so as to achieve the purpose of improving the matching ability and accuracy of intelligent response in the process of human-computer interaction. It enhances user stickiness and promotes further user participation, thus giving vertical industry customers powerful interaction capabilities and promoting the application and promotion of 5G messages in various industries.

[0045] This technical solution gives vertical industry customers powerful interactive capabilities, encourages corporate customers to explore their own marketing and service needs, and quickly, flexibly and conveniently integrates the powerful intelligent interactive capabilities of 5G messages with the original management system, thereby helping to complete complex tasks such as company brand display, product display, product sales, and service display, forming efficient 5G message reach, promoting multiple transactions from one customer acquisition, and thus improving the conversion rate of 5G messages in business applications.

[0046] In general, the 5G message conversation method provided by the embodiment of the present invention can improve the matching ability and accuracy of intelligent responses in the process of human-computer interaction, avoid inaccurate content of the reply message or deviation from the direction of the problem, more accurately control the user's true intentions, enhance user stickiness, and make intelligent replies more accurate, flexible, and personalized, thereby promoting users to further participate in conversational communication, and helping to improve the conversion rate of 5G messages in business applications.

[0047] The 5G message conversation system provided by the embodiment of the present invention can improve the matching ability and accuracy of intelligent responses in the process of human-computer interaction, avoid inaccurate content of reply messages or deviation from the direction of the problem, more accurately grasp the true intentions of users, enhance user stickiness, and make intelligent replies more accurate, flexible, and personalized, thereby promoting users to further participate in conversational communication, which helps to improve the conversion rate of 5G messages in business applications. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for use in the embodiments are briefly introduced below. It should be understood that the following drawings only show certain embodiments of the present invention and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without creative work.

[0049] Figure 1 A schematic diagram of a flow chart of a 5G message session method provided in Embodiment 1 of the present invention;

[0050] Figure 2 A schematic diagram of the structure of the 5G message conversation system provided in Example 3 of the present invention.

[0051] Description of reference numerals:

[0052] 5G message conversation system 1000; first conversation terminal 100; communication platform 200; second conversation terminal 300. DETAILED DESCRIPTION

[0053] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Generally, the components of the embodiments of the present invention described and shown in the drawings here can be arranged and designed in various different configurations.

[0054] Therefore, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the invention claimed for protection, but merely represents selected embodiments of the present invention. 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.

[0055] It should be noted that similar reference numerals and letters denote similar items in the following drawings, and therefore, once an item is defined in one drawing, further definition and explanation thereof is not required in subsequent drawings.

[0056] In the description of the present invention, it is also necessary to explain that, unless otherwise clearly specified and limited, the terms "set", "install", "connect", and "connect" should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a direct connection, or an indirect connection through an intermediate medium, or it can be the internal communication of two elements. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0057] In addition, in the description of this application, the labels "S1, S2, S3..." of each step are only used to distinguish each step and should not be regarded as limiting the sequence of each step. The sequence of each step can be flexibly adjusted according to actual conditions and needs.

[0058] Example 1

[0059] Please refer to Figure 1 This embodiment provides a 5G message session method, and the 5G message session method includes the following steps:

[0060] S1: Create a new session based on the user's uplink 5G message;

[0061] S2: The user's historical session records are used as the basic session samples for the new session;

[0062] S3: Process 5G messages through natural language processing (NLP);

[0063] S4: identifying the user's conversation intention by combining the processing result of step S3 and the basic conversation sample;

[0064] S5: Pushing a conversation message to the user according to the conversation intent.

[0065] Each user has a unique identifier, including but not limited to a mobile phone number, and each user corresponds to his or her own historical session record.

[0066] In the 5G message human-computer interaction scenario application, the user's historical conversation records are used as the basic conversation samples for the new conversation. After analyzing, processing, summarizing and reasoning the 5G messages through natural language processing (NLP), the context of the user's historical conversation content is deeply mined and subjectively analyzed to identify the user's true conversation intention, and generate corresponding conversation messages based on the user's conversation intention and push them to the user.

[0067] Through this conversation management method, more user information can be extracted, effectively improving the matching ability and accuracy of intelligent responses during human-computer interaction, significantly enhancing user stickiness, and promoting users to further participate in conversational communication.

[0068] In general, the 5G message conversation method provided by the embodiment of the present invention can improve the matching ability and accuracy of intelligent responses in the process of human-computer interaction, avoid inaccurate content of the reply message or deviation from the direction of the problem, more accurately control the user's true intentions, enhance user stickiness, and make intelligent replies more accurate, flexible, and personalized, thereby promoting users to further participate in conversational communication, and helping to improve the conversion rate of 5G messages in business applications.

[0069] In this embodiment, the 5G message includes message content and behavior data.

[0070] The uplink 5G message triggered by the user actually contains two parts: one is the message content, and the other is the behavior data of the user triggering the message. The user's behavior data contains a lot of important information, such as opening the URL, launching the mini program, launching the app, obtaining location information, replying to the content of the previous message, etc.

[0071] After receiving the user's 5G message, the message is parsed according to the 5G message standard protocol to decompose the message content and the behavioral data of the message content.

[0072] Using message content and behavioral data together as the basis for determining the user's conversational intent can more effectively ensure the accuracy of the conversation process and avoid going off topic.

[0073] Further, step S3 includes the following steps:

[0074] S31: performing word segmentation, stem extraction and word form restoration, part-of-speech tagging, named entity recognition and block standardization on the message content of the 5G message;

[0075] S32: Output the processing result.

[0076] The message content of 5G messages is processed in sequence: word segmentation, stem extraction and word form restoration, part-of-speech tagging, named entity recognition and block standardization. Word segmentation is performed using an understanding-based word segmentation algorithm, and word form restoration is based on a dictionary to transform the complex form of a word into the most basic form.

[0077] Through the above processing, intelligent response can more accurately grasp the user's intention and reduce the possibility of deviation.

[0078] Step S4 includes the following steps:

[0079] S41: Use the basic conversation samples as context samples of the neural network model;

[0080] S42: Using the processing result of step S3 as input sample data of the neural network model;

[0081] S43: Model evaluation using neural network models;

[0082] S44: output the evaluation result;

[0083] S45: Identify the user's session intention according to the evaluation result.

[0084] Specifically, the neural network model includes:

[0085]

[0086] Among them: PP is an indicator for evaluating the effect of language model, which mainly estimates the probability of occurrence of a sentence based on each word, |V| represents the number of words in the corpus, t represents the tth word in the sample: T is the number of words in the sample, y(j,t) is the probability of the word in the corpus, and y^(j,t) is the predicted probability of the word.

[0087] In this process, the basic idea is to assign a language model with a higher probability value to the sentences in the test set. After the language model is trained, the sentences in the test set are all normal sentences, so the trained model has a higher probability on the test set, the better.

[0088] The smaller the confusion, the greater the sentence probability and the better the language model. After the message content of 5G messages is evaluated by the model, the user's true intention can be extracted.

[0089] This design further improves the accuracy of controlling the user's conversational intention.

[0090] In this way, intelligent responses to 5G messages based on session management can be realized, and this session mode can be used to support the intelligent response context environment of chatbot human-computer interaction, thereby achieving the purpose of improving the matching ability and accuracy of intelligent responses in the human-computer interaction process. It enhances user stickiness and promotes further user participation, thereby giving vertical industry customers powerful interactive capabilities and promoting the application and promotion of 5G messages in various industries.

[0091] This technical solution gives vertical industry customers powerful interactive capabilities, encourages corporate customers to explore their own marketing and service needs, and quickly, flexibly and conveniently integrates the powerful intelligent interactive capabilities of 5G messages with the original management system, thereby helping to complete complex tasks such as company brand display, product display, product sales, and service display, forming efficient 5G message reach, promoting multiple transactions from one customer acquisition, and thus improving the conversion rate of 5G messages in business applications.

[0092] Example 2

[0093] This embodiment provides a 5G message session method. Compared with the 5G message session method provided in Embodiment 1, the difference is that the 5G message session method provided in this embodiment further includes the following steps:

[0094] S6: Save the user's historical conversation records as conversation material samples, one user corresponds to one conversation material sample;

[0095] S7: Setting a first relevance judgment rule, matching relevant conversation material samples according to the user's conversation intention and adding them to the basic conversation samples.

[0096] Each user's historical conversation records are saved separately to form conversation material samples. One user's historical conversation record corresponds to one conversation material sample, and these conversation material samples will serve as a supplement to the basic conversation samples, thereby further improving the restoration and enrichment of the conversation scene, making it easier to control the user's conversation intentions.

[0097] Among them, the first relevance judgment rule can be flexibly set according to the actual situation. According to the set first relevance judgment rule, through the preliminary judgment of the user's conversation intention, the relevant conversation material samples can be matched according to the user's conversation intention and added to the basic conversation sample. In the subsequent conversation process, the user will send a new 5G message. Under the processing of the 5G message conversation method, the user's conversation intention will be more accurate. By matching the relevant conversation material samples for more accurate conversation intentions and adding them to the basic conversation samples, a more accurate conversation scenario can be constructed to improve the accuracy of conversation interaction.

[0098] It should be noted that in the process of matching relevant conversation material samples for more accurate conversation intent, the newly matched conversation material samples can be used to replace the originally matched conversation material samples, or can be used in combination with the original conversation material samples. The choice can be flexibly made according to actual conditions.

[0099] Through the above design, closed-loop optimization of basic conversation samples is achieved, which has positive significance for continuously improving and correcting the accuracy of content during conversation interaction.

[0100] Specifically, step S6 also includes: establishing classification rules to classify the conversation material samples;

[0101] Step S7 includes: setting a relevance judgment rule, matching a relevant conversation material sample category according to the user's conversation intention, and adding the conversation material samples under the category to the basic conversation sample.

[0102] Through this design, by pre-establishing a classification table of conversation material samples, the matching efficiency when matching relevant conversation material samples for conversation intent can be greatly improved, the background load can be reduced, the conversation response efficiency can be improved, and users can be avoided from waiting too long during the conversation, thereby optimizing the user's conversation experience.

[0103] Furthermore, the 5G message session method further includes the following steps:

[0104] S8: Setting a second relevance judgment rule, according to which related conversation material samples are associated with each other to form a material chain, matching related conversation material samples according to the user's conversation intention, and adding the material chain where the related conversation material samples are located to the basic conversation sample.

[0105] Among them, the second relevance judgment rule can be flexibly set according to actual conditions. The second relevance judgment rule is used to judge the relevance between conversation material samples, rather than the relevance between categories. On the basis of the second relevance judgment rule, conversation material samples that are judged to be related will be associated with each other to form a material chain. When matching related conversation material samples for the user's conversation intention, if one or more conversation material samples of a certain material chain are matched as related, then the conversation material samples of the entire material chain are all identified as related, and the conversation material samples of the entire material chain will be added to the basic conversation samples.

[0106] Through this design, directly related and indirectly related conversation material samples are added to the basic conversation samples, which makes it easier to provide a more comprehensive logical chain and a more realistic conversation environment for the conversation. This makes the conversation messages you want to reply to users closer to the real scene, and it is easier to dig out the real needs of users. At the same time, it can also guide users to improve and expand their needs, effectively improving the comprehensiveness and accuracy of the conversation process.

[0107] In particular, for sellers, it helps to complete complex tasks such as brand display, product display, product sales, and service display, and promote multiple transactions from one customer acquisition, thereby improving the conversion rate of 5G messages in real business applications.

[0108] Furthermore, in step S6, classifying the conversation material samples includes:

[0109] Set a waiting time, and output the conversation material samples that cannot be accurately classified after the waiting time has passed, prompting manual classification;

[0110] For a conversation material sample that cannot be accurately classified, if a material chain is formed within the waiting time, the conversation material sample will no longer prompt manual classification.

[0111] When classifying the conversation material samples based on the first relevance judgment rule, it is inevitable that some conversation material samples cannot be accurately classified, for example, automatic classification cannot find the corresponding classification category, but it is not limited thereto.

[0112] In order to ensure that the conversation material sample can be used normally, it needs to be classified into the corresponding category. If the waiting time has passed and the automatic classification cannot be completed, it will prompt manual classification of the conversation material sample. When manually classifying, it can be classified into the corresponding category by human judgment, or a new more suitable category can be created.

[0113] Among them, the length of the waiting time can be flexibly set according to actual needs.

[0114] There is a special case, that is, although the automatic classification cannot find the corresponding classification category at the beginning of classification, during the waiting time, one or some new conversation material samples appear. When these new conversation material samples appear, the conversation material samples that cannot be successfully classified can form a material chain with other conversation material samples through the second relevance judgment rule, which is equivalent to the new conversation material sample allowing this conversation material sample to establish an indirect correlation relationship with other conversation material samples. At this time, the conversation material sample no longer prompts manual classification. The original conversation material sample that cannot be classified can be called through the material chain.

[0115] Through this design, the automatic processing capability of conversation material samples is effectively reduced, which facilitates the automatic establishment of an orderly association network between conversation material samples, reduces dependence on manual intervention, and reduces the labor intensity of the manual intervention process.

[0116] It should be noted that during the conversation process, different conversations have different conversation material samples added to the basic conversation sample, which are determined based on the user's conversation intention. Therefore, the 5G message conversation method can accurately realize personalized interaction and break the deadlock of the traditional conversation mode, which has general messages, single content, lack of stickiness, and lack of humanization.

[0117] When the session ends, the session material samples added to the basic session samples are not used as the content of the user's historical session record. On the one hand, each user's drawing content is personalized and generally not universally applicable. On the other hand, it can avoid the high duplication of the content of the historical session record and the session material samples, which can reduce the background burden.

[0118] It is worth noting that although the conversation material samples added to the basic conversation samples are not included in the historical conversation records of the user, the association between these conversation material samples and the historical conversation records can be temporarily retained, and the retention time can be flexibly set according to actual needs. If the association between similar historical conversation records and these conversation material samples is repeated frequently within the retention time, it is likely to represent a new trend. At this time, this association can be retained, and while prompting the human, the associated historical conversation records and conversation material samples can also be merged to form new conversation material samples to help build a more comprehensive material library.

[0119] Example 3

[0120] Please refer to Figure 2 This embodiment provides a 5G message conversation system 1000, and the 5G message conversation system 1000 includes: a first conversation terminal 100, a communication platform 200 and a second conversation terminal 300.

[0121] The first session end 100 is for user use, and the first session end 100 is used for uplink 5G messages.

[0122] The communication platform 200 is used to establish a communication relationship between the first session end 100 and the second session end 300 .

[0123] The second session end 300 is used to execute the 5G message session method provided in Example 1 and / or Example 2 to complete the session with the first session end 100.

[0124] The first conversation terminal 100 may be a mobile phone, and the second conversation terminal 300 may be an enterprise chatbot, but is not limited thereto.

[0125] To sum up, the 5G message conversation method provided in the embodiment of the present invention can improve the matching ability and accuracy of intelligent responses in the process of human-computer interaction, avoid inaccurate content of the reply message or deviation from the direction of the problem, more accurately control the user's true intentions, enhance user stickiness, and make intelligent replies more accurate, flexible, and personalized, which promotes users to further participate in conversational communication and helps to improve the conversion rate of 5G messages in business applications.

[0126] The 5G message conversation system 1000 provided by the embodiment of the present invention can improve the matching ability and accuracy of intelligent responses in the process of human-computer interaction, avoid inaccurate content of the reply message or deviation from the direction of the question, more accurately control the user's true intentions, enhance user stickiness, and make intelligent replies more accurate, flexible, and personalized, thereby promoting users to further participate in conversational communication, which helps to improve the conversion rate of 5G messages in business applications.

[0127] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A 5G message session method, characterized in that: The following steps are involved: S1: Create a new session based on the user's uplink 5G message; S2: The user's historical session records are used as the basic session samples for the new session; S3: Processing the 5G message through natural language processing (NLP); S4: identifying the user's conversation intention by combining the processing result of step S3 and the basic conversation sample; S5: Pushing a conversation message to the user according to the conversation intention; S6: Save the user's historical conversation records as conversation material samples, one user corresponds to one conversation material sample; establish classification rules to classify the conversation material samples; S7: Setting a first relevance judgment rule, matching relevant conversation material samples according to the user's conversation intention and adding them to the basic conversation sample; Setting relevance judgment rules, matching relevant conversation material sample categories according to the user's conversation intention and adding the conversation material samples under the categories to the basic conversation samples; S8: Setting a second relevance judgment rule, according to which related conversation material samples are associated to form a material chain, matching related conversation material samples according to the conversation intention of the user, and adding the material chain where the related conversation material samples are located to the basic conversation sample; the second relevance judgment rule is used to judge the relevance between the conversation material samples; when matching related conversation material samples for the conversation intention of the user, if one or more conversation material samples of a certain material chain are matched as related, then the conversation material samples of the entire material chain are all identified as related, and the conversation material samples of the entire material chain are all added to the basic conversation sample; In step S6, classifying the conversation material samples includes: setting a waiting time, outputting the conversation material samples that cannot be accurately classified after the waiting time, and prompting manual classification; for the conversation material samples that cannot be accurately classified, if the material chain is formed within the waiting time, the conversation material samples will no longer prompt manual classification; Among them, the 5G message includes message content and behavior data, and the behavior data includes: opening the URL, launching the mini program, launching the app, obtaining location information and replying to the content of the previous message.

2. The 5G message conversation method according to claim 1, characterized in that: The step S3 comprises the following steps: S31: performing word segmentation processing, stem extraction and word form restoration processing, part-of-speech tagging processing, named entity recognition processing and block standardization processing on the message content of the 5G message; S32: Output the processing result.

3. The 5G message conversation method according to claim 1, characterized in that: The step S4 comprises the following steps: S41: Using the basic conversation sample as a context environment sample of the neural network model; S42: Using the processing result of step S3 as input sample data of the neural network model; S43: Model evaluation using neural network models; S44: output the evaluation result; S45: Identify the user's session intention according to the evaluation result.

4. The 5G message conversation method according to claim 3, characterized in that: The neural network model includes: ； Among them: PP is an indicator for evaluating the effect of language model, which mainly estimates the probability of occurrence of a sentence based on each word, |V| represents the number of words in the corpus, t represents the tth word in the sample: T is the number of words in the sample, y(j,t) is the probability of the word in the corpus, and y^(j,t) is the predicted probability of the word.

5. A 5G message conversation system, characterized in that: include: A first session end, a communication platform, and a second session end; The communication platform is used to establish a communication relationship between the first session end and the second session end; The first session end is used for uplink 5G messages; the second session end is used to execute the 5G message session method as described in any one of claims 1 to 4 to complete the session with the first session end.

Images