Intelligent agent self-learning method and system based on corpus learning
Through the corpus learning device, the intent data of the agent is marked and trained, which solves the problem that existing agents require a lot of manual intervention when optimizing the quality of human-computer dialogue, and realizes the efficiency of the agent's self-learning and the improvement of dialogue quality.
Patent Information
- Application Number
- CN202411555895.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-11-04
- Publication Date
- 2025-02-25
AI Technical Summary
Existing agents require a lot of manual intervention when optimizing the quality of human-computer dialogue, and it is difficult to directly improve the quality of dialogue, resulting in inefficiency.
The corpus learning device generates dialogue data, extracts customer response sentence samples for intention judgment, marks and corrects the agent's intention data, uses the labeled data to train the corpus learning model, and the agent conducts self-learning to improve semantic understanding and intention recognition capabilities.
It greatly reduces the time and energy of business personnel when optimizing the quality of human-computer dialogue, improves the self-learning performance of the agent, and improves the quality of dialogue.
Smart Images
Figure CN119513249A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of human-computer dialogue, and in particular to a corpus-based learning agent self-learning method and system. Background Art
[0002] Intelligent agents in the prior art, such as voice robots, usually need to be optimized based on accumulated historical data. However, this optimization process usually requires a high degree of involvement of the backend personnel of the intelligent agent and takes a lot of time, because the backend personnel need to directly adjust the model parameters of the intelligent agent, and the optimization may not achieve the desired effect because the backend personnel cannot directly understand the business logic. Even if users have a way to perform preliminary optimization, such as quality inspection of conversation data, it will still remain at the optimization of the text information level due to format settings, limited intelligent agent capabilities, etc., and it is difficult to directly convert it into an improvement in the quality of subsequent conversations. Summary of the invention
[0003] Based on this, in order to solve the technical problems in the prior art, an intelligent agent self-learning method is proposed, including:
[0004] The system generates a dialogue; after the dialogue is generated, the dialogue is stored in the database;
[0005] The corpus learning device extracts samples based on customer response sentences for intention judgment verification;
[0006] Checking the intention obtained by the agent through the corpus learning device to recognize the customer's response sentence, marking and correcting the intention data with recognition errors, and sending the marked intention data to the agent as the corpus learning data of the agent;
[0007] The corpus learning device trains the corpus learning model using the corpus learning data, and the intelligent agent performs self-learning using the corpus learning model;
[0008] After receiving the labeled corpus learning data, the agent uses it to adjust the corpus learning model.
[0009] In one embodiment, the corpus learning device includes a learning workbench and a data learning platform; corpus learning is performed through the corpus learning device;
[0010] Among them, the learning workbench is used to create corpus learning batches and train the learning model through the learning workbench; the data learning platform is a working environment for corpus learning labeling, and the data learning platform includes a data learning page, which displays the corpus data learning results.
[0011] In one embodiment, performing corpus learning by a corpus learning device includes:
[0012] The learning workbench creates a corpus learning batch. After the corpus learning batch is created, the corpus is marked in the data learning platform.
[0013] Among them, corpus tagging includes determining the number and range of corpus data that need to be labeled and extracting corpus data samples, and marking the extracted samples in the data learning platform; marking the batch of corpus data as completed tagging after tagging is completed; using the marked corpus data for training the corpus learning model, and providing the corpus learning model to the intelligent agent for self-learning.
[0014] In one embodiment, the corpus tagging includes, in sequence, importing an intent list, setting data screening conditions and push methods, and intent annotation;
[0015] Importing the intent list includes first importing the intent list before performing intent annotation on the corpus data, and initializing the intent list after importing, wherein a preset intent list is selected to be imported from the flowchart, or a new intent list is created by the user;
[0016] Among them, setting data screening conditions and push methods includes: displaying qualified corpus data by filling in screening conditions, including time range, number of data words, keywords, and intent; marking the screened corpus data; setting advanced screening mode and refining screening conditions when the data volume is large;
[0017] Among them, intent labeling includes displaying relevant parameters for each customer response sentence, including sentence text, time, whether it has been learned, whether it has been labeled, and recommended intent judgment; when the intent judgment is correct, it is directly marked as completed; when the intent judgment is incorrect, it is corrected first and then marked as completed.
[0018] In one embodiment, the learning workbench aggregates the corpus data that have been learned to create a corpus learning model, trains the corpus learning model, and completes the intent learning of the batch of corpus data;
[0019] After the corpus learning model training is completed, the parameters of the corpus learning model are displayed on the data learning page. The corpus learning model parameters include the version number and version name, training status, creation time, the amount of data included in the model, and the list of intents involved in the training;
[0020] The learning workbench provides the trained corpus learning model to the intelligent agent for self-learning.
[0021] In addition, in order to solve the technical problems in the prior art, an intelligent agent self-learning system is proposed, including a database and a corpus learning device;
[0022] The database is used to store the conversation data; the system generates the conversation, and after the conversation is generated, the conversation is stored in the database;
[0023] The corpus learning device extracts samples based on customer response sentences for intention judgment verification;
[0024] Checking the intention obtained by the agent through the corpus learning device to recognize the customer's response sentence, marking and correcting the intention data with recognition errors, and sending the marked intention data to the agent as the corpus learning data of the agent;
[0025] The corpus learning device trains the corpus learning model using the corpus learning data, and the intelligent agent performs self-learning using the corpus learning model;
[0026] After receiving the labeled corpus learning data, the intelligent agent uses it to adjust the corpus learning model and uses the corpus learning model for self-learning.
[0027] In one embodiment, the corpus learning device includes a learning workbench and a data learning platform; corpus learning is performed through the corpus learning device;
[0028] Among them, the learning workbench is used to create corpus learning batches and train the learning model through the learning workbench; the data learning platform is a working environment for corpus learning labeling, and the data learning platform includes a data learning page, which displays the corpus data learning results.
[0029] In one embodiment, performing corpus learning by a corpus learning device includes:
[0030] The learning workbench creates a corpus learning batch. After the corpus learning batch is created, the corpus is marked in the data learning platform.
[0031] Among them, corpus tagging includes determining the number and range of corpus data that need to be labeled and extracting corpus data samples, and marking the extracted samples in the data learning platform; marking the batch of corpus data as completed tagging after tagging is completed; using the marked corpus data for training the corpus learning model, and providing the corpus learning model to the intelligent agent for self-learning.
[0032] In one embodiment, the corpus tagging includes, in sequence, importing an intent list, setting data screening conditions and push methods, and intent annotation;
[0033] Importing the intent list includes first importing the intent list before performing intent annotation on the corpus data, and initializing the intent list after importing, wherein a preset intent list is selected to be imported from the flowchart, or a new intent list is created by the user;
[0034] Among them, setting data screening conditions and push methods includes: displaying qualified corpus data by filling in screening conditions, including time range, number of data words, keywords, and intent; marking the screened corpus data; setting advanced screening mode and refining screening conditions when the data volume is large;
[0035] Among them, intent labeling includes displaying relevant parameters for each customer response sentence, including sentence text, time, whether it has been learned, whether it has been labeled, and recommended intent judgment; when the intent judgment is correct, it is directly marked as completed; when the intent judgment is incorrect, it is corrected first and then marked as completed.
[0036] In one embodiment, the learning workbench aggregates the corpus data that have been learned to create a corpus learning model, trains the corpus learning model, and completes the intent learning of the batch of corpus data;
[0037] After the corpus learning model training is completed, the parameters of the corpus learning model are displayed on the data learning page. The corpus learning model parameters include the version number and version name, training status, creation time, the amount of data included in the model, and the list of intents involved in the training;
[0038] The learning workbench provides the trained corpus learning model to the intelligent agent for self-learning.
[0039] Implementing the embodiments of the present invention will have the following beneficial effects:
[0040] In the present invention, the accumulated historical data is used to optimize the quality of human-computer dialogue in the human-computer collaborative dialogue system; the system combines the self-learning ability of the intelligent agent on the labeled data, and the business personnel directly connect to the intelligent agent for debugging through the corpus learning interface, which greatly saves the time and energy required by the business personnel in optimizing the quality of human-computer dialogue in the prior art, and greatly improves the overall performance of the intelligent agent's self-learning. BRIEF DESCRIPTION OF THE DRAWINGS
[0041] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, 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 creative work.
[0042] in:
[0043] Figure 1 Schematic diagram of the process of the intelligent agent self-learning method in the present invention;
[0044] Figure 2 Schematic diagram of the intelligent agent self-learning system in the present invention; DETAILED DESCRIPTION
[0045] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.
[0046] The present invention discloses an intelligent agent self-learning method based on corpus learning, comprising:
[0047] System generated dialogue;
[0048] The interaction process between the customer and the agent is a dialogue; the dialogue types include outbound calls and inbound calls; outbound calls are when the agent initiates a dialogue with the customer, and inbound calls are when the customer initiates a dialogue with the agent; outbound call types include connected and unconnected;
[0049] Among them, the intelligent agent is a computing entity that can function autonomously and sustainably; in the human-computer collaborative dialogue system, it is a voice robot or text robot that can communicate with customers;
[0050] After the conversation is generated, it is stored in the system database as historical data; system users can call the historical data stored in the database;
[0051] Among them, system users include quality inspectors and agent trainers;
[0052] The intelligent agent uses the corpus learning device to self-learn semantic understanding;
[0053] In the self-learning for semantic understanding, the corpus learning device extracts samples based on customer response sentences for intention judgment verification;
[0054] Checking the intention obtained by the agent through the corpus learning device to recognize the customer's response sentence, marking and correcting the intention data with recognition errors, and sending the marked intention data to the agent as the corpus learning data of the agent;
[0055] The corpus learning device trains the corpus learning model according to the marked corpus learning data, and the intelligent agent performs self-learning using the corpus learning model;
[0056] Among them, the intention is for the agent to recognize the key meaning in the customer's response sentence; the recognition engine of the agent reads and recognizes the customer's voice, associates the recognized customer's speech content with the intention, and uses the intention as the basis for the agent to jump to the branch;
[0057] After receiving the labeled corpus learning data, the agent uses it to adjust the corpus learning model;
[0058] In particular, the corpus learning device includes a learning workbench and a data learning platform; corpus learning is performed through the corpus learning device;
[0059] The learning workbench is used to create corpus learning batches and train learning models through the learning workbench. The data learning platform is a working environment for corpus learning labeling. The data learning platform includes a data learning page, which displays corpus data learning results.
[0060] In particular, the corpus learning is performed by the corpus learning device, including:
[0061] The learning workbench creates a corpus learning batch. After the corpus learning batch is created, the corpus is marked in the data learning platform.
[0062] Among them, corpus tagging includes determining the number and range of corpus data that need to be tagged and extracting corpus data samples, and marking the extracted samples in the data learning platform; marking the batch of corpus data as completed after tagging; using the tagged corpus data for training the corpus learning model, and providing the corpus learning model to the intelligent agent for self-learning, thereby improving the intelligent agent's semantic understanding and intention recognition capabilities;
[0063] In particular, corpus tagging includes importing intent lists, setting data filtering conditions and push methods, and intent annotations in sequence;
[0064] Importing the intent list includes first importing the intent list before performing intent annotation on the corpus data, and initializing the intent list after importing, wherein a preset intent list is selected to be imported from the flowchart, or a new intent list is created by the user;
[0065] Among them, setting data screening conditions and push methods includes: displaying eligible data subjects by filling in screening conditions, including time range, data word count, keywords, and intent; marking the listed data; setting advanced screening mode and refining screening conditions when the data volume is large;
[0066] Among them, the intention annotation includes displaying the relevant parameters for each customer response sentence, including sentence text, time, whether it has been learned, whether it has been annotated, and recommended intention judgment; when the intention judgment is correct, it is directly marked as completed; when the intention judgment is incorrect, it is corrected first and then marked as completed;
[0067] The learning workbench aggregates the corpus data that have been learned to create a corpus learning model, trains the corpus learning model, and completes the intent learning of the batch of corpus data;
[0068] After the corpus learning model training is completed, the parameters of the corpus learning model are displayed on the data learning page. The corpus learning model parameters include the version number and version name, training status, creation time, the amount of data included in the model, and the list of intents involved in the training;
[0069] The learning workbench provides the trained corpus learning model to the intelligent agent for self-learning; at the same time, system users can view various versions of the learning model.
[0070] The corpus learning model is provided to the intelligent agent for self-learning. The intelligent agent adjusts the corpus learning model according to the marked corpus learning data to improve the intelligent agent's ability of semantic understanding and intent recognition.
[0071] Specifically, for semantic understanding problems, we use corpus data screening strategies to focus on multiple types, including ununderstood sentences, sentence length, sentence word count, and multiple rounds of dialogue;
[0072] The agent's optimization method for semantic understanding problems is related to the semantic understanding method, including increasing matching keywords and expanding the keyword library, expanding regular matching conditions, increasing machine learning sample sentences, increasing semantic tags and expanding the semantic tag library.
[0073] The present invention discloses an intelligent agent self-learning system based on corpus learning, including a database and a corpus learning device;
[0074] The system generates a dialogue; the interaction process between the customer and the agent is a dialogue; the dialogue types include outbound calls and inbound calls; outbound calls are when the agent initiates a dialogue with the customer, and inbound calls are when the customer initiates a dialogue with the agent; outbound call types include connected and unconnected;
[0075] Among them, the intelligent agent is a computing entity that can function autonomously and sustainably; in the human-computer collaborative dialogue system, it is a voice robot or text robot that can communicate with customers;
[0076] The database is used to store conversation data; the system generates conversations, and after the conversations are generated, the conversations are stored in the system database as historical data; system users can call the historical data stored in the database;
[0077] Among them, system users include quality inspectors and agent trainers;
[0078] The intelligent agent uses the corpus learning device to self-learn semantic understanding;
[0079] In the self-learning for semantic understanding, the corpus learning device extracts samples based on the customer's response sentences for intention judgment verification; the corpus learning device checks the intention obtained by the intelligent agent through the recognition of the customer's response sentences, marks and corrects the intention data with recognition errors, and sends the marked intention data to the intelligent agent as the corpus learning data of the intelligent agent;
[0080] The corpus learning device trains the corpus learning model using the marked corpus learning data; the intelligent agent performs self-learning using the corpus learning model;
[0081] Among them, the intention is for the agent to recognize the key meaning in the customer's response sentence; the recognition engine of the agent reads and recognizes the customer's voice, associates the recognized customer's speech content with the intention, and uses the intention as the basis for the agent to jump to the branch;
[0082] After receiving the labeled corpus learning data, the agent uses it to adjust the corpus learning model;
[0083] In particular, the corpus learning device includes a learning workbench and a data learning platform; corpus learning is performed through the corpus learning device;
[0084] The learning workbench is used to create corpus learning batches and train learning models through the learning workbench. The data learning platform is a working environment for corpus learning labeling. The data learning platform includes a data learning page, which displays corpus data learning results.
[0085] In particular, the corpus learning is performed by the corpus learning device, including:
[0086] The learning workbench creates a corpus learning batch. After the corpus learning batch is created, the corpus is marked in the data learning platform.
[0087] Among them, corpus tagging includes determining the number and range of corpus data that need to be tagged and extracting corpus data samples, and marking the extracted samples in the data learning platform; marking the batch of corpus data as completed after tagging; using the tagged corpus data for training the corpus learning model, and providing the corpus learning model to the intelligent agent for self-learning, thereby improving the intelligent agent's ability to understand semantics and recognize intent;
[0088] In particular, corpus tagging includes importing intent lists, setting data filtering conditions and push methods, and intent annotations in sequence;
[0089] Importing the intent list includes first importing the intent list before performing intent annotation on the corpus data, and initializing the intent list after importing, wherein a preset intent list is selected to be imported from the flowchart, or a new intent list is created by the user;
[0090] Among them, setting data screening conditions and push methods includes: displaying qualified corpus data by filling in screening conditions, including time range, number of data words, keywords, and intent; marking the screened corpus data; setting advanced screening mode and refining screening conditions when the data volume is large;
[0091] Among them, the intention annotation includes displaying the relevant parameters for each customer response sentence, including sentence text, time, whether it has been learned, whether it has been annotated, and recommended intention judgment; when the intention judgment is correct, it is directly marked as completed; when the intention judgment is incorrect, it is corrected first and then marked as completed;
[0092] The learning workbench aggregates the corpus data that have been learned to create a corpus learning model, trains the corpus learning model, and completes the intent learning of the batch of corpus data;
[0093] After the corpus learning model training is completed, the parameters of the corpus learning model are displayed on the data learning page. The corpus learning model parameters include the version number and version name, training status, creation time, the amount of data included in the model, and the list of intents involved in the training;
[0094] The learning workbench provides the trained corpus learning model to the intelligent agent for self-learning; at the same time, system users can view various versions of the learning model.
[0095] The corpus learning model is provided to the intelligent agent for self-learning. The intelligent agent adjusts the corpus learning model according to the marked corpus learning data to improve the intelligent agent's ability of semantic understanding and intent recognition.
[0096] Specifically, for semantic understanding problems, we use corpus data screening strategies to focus on multiple types, including ununderstood sentences, sentence length, sentence word count, and multiple rounds of dialogue;
[0097] The agent's optimization method for semantic understanding problems is related to the semantic understanding method, including increasing matching keywords and expanding the keyword library, expanding regular matching conditions, increasing machine learning sample sentences, increasing semantic tags and expanding the semantic tag library.
[0098] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features may be replaced by equivalents. Such modifications or replacements will not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.
Claims
1. A method for self-learning of an intelligent agent based on corpus learning, characterized in that: include: System generated dialogue; After the dialogue is generated, the dialogue is stored in the database; The corpus learning device extracts samples based on customer response sentences for intention judgment verification; Checking the intention obtained by the agent through the corpus learning device to recognize the customer's response sentence, marking and correcting the intention data with recognition errors, and sending the marked intention data to the agent as the corpus learning data of the agent; The corpus learning device trains the corpus learning model using the corpus learning data, and the intelligent agent performs self-learning using the corpus learning model; After receiving the labeled corpus learning data, the agent uses it to adjust the corpus learning model.
2. The agent self-learning method based on corpus learning according to claim 1 is characterized in that: in, The corpus learning device includes a learning workbench and a data learning platform; corpus learning is performed through the corpus learning device; Among them, the learning workbench is used to create corpus learning batches and train the learning model through the learning workbench; the data learning platform is a working environment for corpus learning labeling, and the data learning platform includes a data learning page, which displays the corpus data learning results.
3. According to claim 2, the intelligent agent self-learning method based on corpus learning, It is characterized in that in, Corpus learning is performed through a corpus learning device, including: The learning workbench creates a corpus learning batch. After the corpus learning batch is created, the corpus is marked in the data learning platform. Among them, corpus tagging includes determining the number and range of corpus data that need to be labeled and extracting corpus data samples, and marking the extracted samples in the data learning platform; marking the batch of corpus data as completed tagging after tagging is completed; using the marked corpus data for training the corpus learning model, and providing the corpus learning model to the intelligent agent for self-learning.
4. According to claim 3, the intelligent agent self-learning method based on corpus learning, It is characterized in that in, Corpus tagging includes importing intent list, setting data filtering conditions and push methods, and intent annotation. Importing the intent list includes first importing the intent list before performing intent annotation on the corpus data, and initializing the intent list after importing, wherein a preset intent list is selected to be imported from the flowchart, or a new intent list is created by the user; Among them, setting data screening conditions and push methods includes: displaying qualified corpus data by filling in screening conditions, including time range, number of data words, keywords, and intent; marking the screened corpus data; setting advanced screening mode and refining screening conditions when the data volume is large; Among them, intent labeling includes displaying relevant parameters for each customer response sentence, including sentence text, time, whether it has been learned, whether it has been labeled, and recommended intent judgment; when the intent judgment is correct, it is directly marked as completed; when the intent judgment is incorrect, it is corrected first and then marked as completed.
5. The agent self-learning method based on corpus learning according to claim 2, It is characterized in that The learning workbench aggregates the corpus data that have been learned to create a corpus learning model, trains the corpus learning model, and completes the intent learning of the batch of corpus data; After the corpus learning model training is completed, the parameters of the corpus learning model are displayed on the data learning page. The corpus learning model parameters include the version number and version name, training status, creation time, the amount of data included in the model, and the list of intents involved in the training; The learning workbench provides the trained corpus learning model to the intelligent agent for self-learning.
6. An intelligent agent self-learning system based on corpus learning, characterized in that: Including database, corpus learning device; The database is used to store the conversation data; the system generates the conversation, and after the conversation is generated, the conversation is stored in the database; The corpus learning device extracts samples based on customer response sentences for intention judgment verification; Checking the intention obtained by the agent through the corpus learning device to recognize the customer's response sentence, marking and correcting the intention data with recognition errors, and sending the marked intention data to the agent as the corpus learning data of the agent; The corpus learning device trains the corpus learning model using the corpus learning data, and the intelligent agent performs self-learning using the corpus learning model; After receiving the labeled corpus learning data, the intelligent agent uses it to adjust the corpus learning model and uses the corpus learning model for self-learning.
7. The intelligent agent self-learning system based on corpus learning according to claim 6 is characterized in that: in, The corpus learning device includes a learning workbench and a data learning platform; corpus learning is performed through the corpus learning device; Among them, the learning workbench is used to create corpus learning batches and train the learning model through the learning workbench; the data learning platform is a working environment for corpus learning labeling, and the data learning platform includes a data learning page, which displays the corpus data learning results.
8. According to the corpus learning-based intelligent agent self-learning system of claim 7, It is characterized in that in, Corpus learning is performed through a corpus learning device, including: The learning workbench creates a corpus learning batch. After the corpus learning batch is created, the corpus is marked in the data learning platform. Among them, corpus tagging includes determining the number and range of corpus data that need to be labeled and extracting corpus data samples, and marking the extracted samples in the data learning platform; marking the batch of corpus data as completed tagging after tagging is completed; using the marked corpus data for training the corpus learning model, and providing the corpus learning model to the intelligent agent for self-learning.
9. According to claim 8, the intelligent agent self-learning system based on corpus learning, It is characterized in that in, Corpus tagging includes importing intent list, setting data filtering conditions and push methods, and intent annotation. Importing the intent list includes first importing the intent list before performing intent annotation on the corpus data, and initializing the intent list after importing, wherein a preset intent list is selected to be imported from the flowchart, or a new intent list is created by the user; Among them, setting data screening conditions and push methods includes: displaying qualified corpus data by filling in screening conditions, including time range, number of data words, keywords, and intent; marking the screened corpus data; setting advanced screening mode and refining screening conditions when the data volume is large; Among them, intent labeling includes displaying relevant parameters for each customer response sentence, including sentence text, time, whether it has been learned, whether it has been labeled, and recommended intent judgment; when the intent judgment is correct, it is directly marked as completed; when the intent judgment is incorrect, it is corrected first and then marked as completed.
10. According to claim 6, the intelligent agent self-learning system based on corpus learning, It is characterized in that The learning workbench aggregates the corpus data that have been learned to create a corpus learning model, trains the corpus learning model, and completes the intent learning of the batch of corpus data; After the corpus learning model training is completed, the parameters of the corpus learning model are displayed on the data learning page. The corpus learning model parameters include the version number and version name, training status, creation time, the amount of data included in the model, and the list of intents involved in the training; The learning workbench provides the trained corpus learning model to the intelligent agent for self-learning.