Vehicle insurance case reporting method based on telephone robot

Abstract

The invention discloses a car insurance case reporting method based on a telephone robot, which comprises the following steps: a telephone robot system receives a call from a caller, and the system routes the call to a robot or an artificial agent; when the call is connected to the robot, the robot converts the voice of the caller into text using a voice recognition technique. According to the invention, the realization of a new-generation telephone robot platform is explained by taking car insurance case reporting as a scene and aiming at a scene that a customer makes a customer service phone to report a case. In order to solve the problems that a traditional telephone robot NLP model cannot extract a multi-slot result through a single model and does not have the capabilities of context association and multi-round information modification, the telephone robot is enabled by introducing a large model through analysis and comparison, so that the telephone robot has the capabilities of multi-task solution and context association, and the efficiency of the telephone robot is improved. Semantics spoken by the user can be really understood, and the recognition success rate of various case reporting in a car insurance case reporting scene and the satisfaction degree of the customer are improved.

Description

Technical Field

[0001] The present invention relates to the field of communication technology, and in particular to a car insurance reporting method based on a telephone robot. Background Art

[0002] A phone robot is an automated system designed to interact with users over the phone. They can be used for a variety of tasks, such as customer service, marketing, research, appointment scheduling, etc. The core technologies used by phone robots include automatic speech recognition (ASR), natural language processing (NLP), and text-to-speech conversion (TTS).

[0003] Automatic Speech Recognition (ASR): enables the system to understand and process voice commands entered by the user over the phone. The development of this technology has greatly improved the interactive capabilities of telephone robots, allowing them to process natural language input rather than relying solely on predetermined options.

[0004] Natural Language Processing (NLP): By analyzing and understanding the user’s language, NLP enables phone bots to have more complex conversations. Advances in NLP allow phone bots to understand context, answer questions, and perform tasks.

[0005] Text-to-speech (TTS): This technology enables telephony bots to generate natural-sounding voice responses, thereby improving the user experience.

[0006] Large models, especially language models based on deep learning, such as GPT-3 and GPT-4, are an important development in the field of artificial intelligence in recent years. These models consist of a large number of neural network layers, which can generate and understand natural language text, perform complex reasoning and prediction by training on large-scale data sets. Large models are usually based on deep neural networks, especially the Transformer architecture. The performance of large models is highly dependent on the scale and quality of training data. Typically, large models are pre-trained on large-scale text data, which may include: Wikipedia articles, books and news articles, social media content, scientific research papers, etc. The training process of large models is usually divided into two stages:

[0007] Pre-training: Training on large-scale unsupervised data to learn the basic structure and semantics of the language. The goal is to enable the model to generate meaningful text and understand context. Fine-tuning: Supervised training on data from a specific task or domain to adapt to specific application requirements. For example, for sentiment analysis, text classification, or machine translation.

[0008] In the existing technology, traditional telephone robots have problems such as a single speech node can only collect a single information, a feature node can only collect specific information, keywords are misabsorbed, a single factor extraction requires separate training, and the model requires a large amount of labeled data. Summary of the invention

[0009] The purpose of the present invention is to solve the problems that a single speech node of a traditional telephone robot can only collect single information, a feature node can only collect specific information, keywords are misabsorbed, single element extraction requires separate training, and the model requires a large amount of labeled data, and proposes a car insurance reporting method based on a telephone robot.

[0010] The purpose of the present invention can be achieved through the following technical solutions:

[0011] A method for reporting automobile insurance cases based on a telephone robot comprises the following steps:

[0012] The telephony robot system receives calls from callers and the system routes the call to a robot or a human agent;

[0013] When a call is connected to the bot, the bot uses speech recognition technology to convert the caller’s speech into text;

[0014] The robot uses a large model speech flow to understand the caller's intention, analyzes the text after speech conversion to extract its meaning, and generates a response based on the understanding of the caller's intention, hitting the pre-set large model speech flow, and broadcasting a pre-recorded voice message and / or a text generated in real time;

[0015] If the robot cannot understand the caller’s intent, the call is routed to a human agent.

[0016] As a further aspect of the present invention, the routing decision is based on the caller's input or the type of call.

[0017] As a further aspect of the present invention: speech recognition technology decomposes speech signals into individual utterances and then converts these utterances into text.

[0018] As a further solution of the present invention: the large model speech process includes the following steps:

[0019] S1. Collect basic information about auto insurance. When the user enters the start node process, the robot will ask for basic information about auto insurance;

[0020] S2, policy query, the robot checks whether the basic information of the auto insurance has been collected;

[0021] If it has been collected, proceed to the next step after verifying the address with the user;

[0022] If it is not collected, then ask the user for the information again;

[0023] S3: The robot asks the user whether he is at the scene, and asks the user about casualties and vehicle damage, and collects information about the losses;

[0024] S4. The robot asks the user whether the vehicle can be driven and whether rescue is needed. It then further confirms the identity of the reporter, verifies the user’s contact information, and broadcasts the reporting instructions.

[0025] As a further solution of the present invention: in step S1, the basic information of the car insurance includes whether to report the case, personal injury situation, collection time, collection address, and cause of the accident.

[0026] As a further solution of the present invention: in step S2, the information inquired from the user includes collecting time, collecting and verifying the address, and collecting the cause of the accident.

[0027] Beneficial effects of the present invention: The present invention uses car insurance reporting as a scenario, and describes the implementation of a new generation of telephone robot platform for the scenario where a customer calls the customer service phone to report an insurance claim. In order to solve the problem that the traditional telephone robot NLP model cannot extract multi-slot results through a single model, does not have the ability of context association, and does not have the ability of multi-round information modification, the present invention, through analysis and comparison, introduces a large model to empower the telephone robot, so that the telephone robot has multi-task solving and context association capabilities, can truly understand the semantics of what the user says, and improves the recognition success rate of various reports in the car insurance reporting scenario and customer satisfaction. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] The present invention will be further described below in conjunction with the accompanying drawings.

[0029] Figure 1 It is a schematic diagram of the large model speech flow of the present invention. DETAILED DESCRIPTION

[0030] 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.

[0031] See also Figure 1 As shown, the present invention is a method for reporting car insurance based on a telephone robot, comprising the following steps:

[0032] The telephony robot system receives calls from callers and the system routes the call to a robot or a human agent;

[0033] When a call is connected to the bot, the bot uses speech recognition technology to convert the caller’s speech into text;

[0034] The robot uses a large model speech flow to understand the caller's intention, analyzes the text after speech conversion to extract its meaning, and generates a response based on the understanding of the caller's intention, hitting the pre-set large model speech flow, and broadcasting a pre-recorded voice message and / or a text generated in real time;

[0035] If the robot cannot understand the caller’s intent, the call is routed to a human agent.

[0036] Routing decisions are based on the caller's input or the type of call.

[0037] Speech recognition technology breaks down speech signals into individual utterances and then converts those utterances into text.

[0038] The big model dialogue process includes the following steps:

[0039] S1. Collect basic information about auto insurance. When the user enters the start node process, the robot will ask for basic information about auto insurance;

[0040] S2, policy query, the robot checks whether the basic information of the auto insurance has been collected;

[0041] If it has been collected, proceed to the next step after verifying the address with the user;

[0042] If it is not collected, then ask the user for the information again;

[0043] S3: The robot asks the user whether he is at the scene, and asks the user about casualties and vehicle damage, and collects information about the losses;

[0044] S4. The robot asks the user whether the vehicle can be driven and whether rescue is needed. It then further confirms the identity of the reporter, verifies the user’s contact information, and broadcasts the reporting instructions.

[0045] The basic information of auto insurance includes whether a report has been made, personal injury situation, collection time, collection address, and cause of the accident.

[0046] The information asked from the user includes the collection time, address collection and verification, and collection of the cause of the accident.

[0047] This invention uses car insurance reporting as a scenario, and describes the implementation of a new generation of telephone robot platform for the scenario where a customer calls the customer service phone to report an insurance claim. In order to solve the problem that the traditional telephone robot NLP model cannot extract multi-slot results through a single model, does not have the ability of context association, and does not have the ability of multi-round information modification, this invention empowers the telephone robot through analysis and comparison, and introduces a large model to enable the telephone robot, so that the telephone robot has multi-task solving and context association capabilities, can truly understand the semantics of what the user says, and improve the recognition success rate of various reports in the car insurance reporting scenario and customer satisfaction.

[0048] Large model training

[0049] The large model training flowchart is a diagram that outlines the large model training process. It generally includes the following steps:

[0050] Data preparation: Collecting and preprocessing training data. This may include collecting data from various sources, cleaning it, and converting it into a format that the model can understand. The dataset should be large and diverse enough to allow the model to learn general patterns.

[0051] Model architecture design: Choose an appropriate model architecture. This depends on the type of training data and the task to be performed. Some common model architectures include Transformer, CNN, and RNN.

[0052] Model training: This involves training the model using training data. This involves adjusting the model parameters using an optimization algorithm to minimize the loss function. The loss function measures the difference between the model's predictions and the actual results.

[0053] Model evaluation: Evaluate the performance of the model using a validation dataset. This helps ensure that the model generalizes well on data outside of the training data. Common evaluation metrics include accuracy, precision, and recall.

[0054] Model fine-tuning: Fine-tune the model as needed to perform a specific task. This may include adjusting the model architecture or training data. Model fine-tuning can improve the performance of the model on a specific task.

[0055] Model deployment: The model is deployed to a production environment. This may include packaging the model as a software application or deploying it to a cloud platform. Model deployment enables the model to be used in real-world applications.

[0056] Talk Design

[0057] Through multiple rounds of individual collection of questions and finally transferring to manual review, it is transformed into a first round with multi-slot extraction and subsequent process combined with understanding.

[0058] Telephone robot design

[0059] A telephony bot, also known as a call center robot or virtual assistant, is a software program that uses speech recognition and natural language processing technology to simulate human conversation. They are often used in customer service applications, such as answering questions, providing support, and processing orders.

[0060] The phone robot process usually includes the following steps:

[0061] Call Routing: The telephony bot system first receives a call from a caller. The system then routes the call to the appropriate bot or human agent. Routing decisions are usually based on the caller's input or the type of call.

[0062] Speech Recognition: When a call is connected to the bot, the bot converts the caller's speech into text using speech recognition technology. Speech recognition technology breaks down the voice signal into individual utterances and then converts those utterances into text.

[0063] In the auto insurance reporting scenario, when the user enters the start node process, the robot will ask for basic information about the auto insurance, and the user will say the corresponding reporting information, such as whether to report the case, personal injury situation, collection time, collection address, cause of occurrence, etc.

[0064] Big model semantic recognition: Once the speech is converted to text, the bot uses big model technology to understand the caller’s intent. This involves analyzing the text to extract its meaning.

[0065] Response generation: Based on its understanding of the caller’s intent, the robot generates a response. This can be a pre-set robot script flow, a pre-recorded voice message, or a text generated in real time. If the robot cannot understand the caller’s intent, it may route the call to a human agent.

[0066] The above is a detailed description of an embodiment of the present invention, but the content is only a preferred embodiment of the present invention and cannot be considered to limit the scope of implementation of the present invention. All equivalent changes and improvements made according to the scope of application of the present invention should still fall within the scope of the claims of the present invention.

Claims

1. A method for reporting car insurance based on a telephone robot, characterized in that: The following steps are involved: The telephony robot system receives calls from callers and the system routes the call to a robot or a human agent; When a call is connected to the bot, the bot uses speech recognition technology to convert the caller’s speech into text; The robot uses a large model speech flow to understand the caller's intention, analyzes the text after speech conversion to extract its meaning, and generates a response based on the understanding of the caller's intention, hitting the pre-set large model speech flow, and broadcasting a pre-recorded voice message and / or a text generated in real time; If the robot cannot understand the caller’s intent, the call is routed to a human agent.

2. The method for reporting automobile insurance case based on telephone robot according to claim 1 is characterized in that: Routing decisions are based on the caller's input or the type of call.

3. The method for reporting automobile insurance case based on telephone robot according to claim 1 is characterized in that: Speech recognition technology breaks down speech signals into individual utterances and then converts those utterances into text.

4. A method for reporting automobile insurance based on a telephone robot according to any one of claims 1 to 3, characterized in that: The big model dialogue process includes the following steps: S1. Collect basic information about auto insurance. When the user enters the start node process, the robot will ask for basic information about auto insurance; S2, policy query, the robot checks whether the basic information of the auto insurance has been collected; If it has been collected, proceed to the next step after verifying the address with the user; If it is not collected, then ask the user for the information again; S3: The robot asks the user whether he is at the scene, and asks the user about casualties and vehicle damage, and collects information about the losses; S4. The robot asks the user whether the vehicle can be driven and whether rescue is needed. It then further confirms the identity of the reporter, verifies the user’s contact information, and broadcasts the reporting instructions.

5. The method for reporting automobile insurance case based on telephone robot according to claim 4 is characterized in that: In step S1, the basic information of the car insurance includes whether a report has been made, personal injury situation, collection time, collection address, and cause of the accident.

6. The method for reporting automobile insurance case based on telephone robot according to claim 4 is characterized in that: In step S2, the user is asked for information including the collection time, address collection and verification, and collection of the cause of the accident.

Images