Information processing and classification method and device based on large model, equipment and medium

Abstract

The invention discloses an information processing and classifying method, device and system based on a large model and a medium, and the method comprises the steps: obtaining the dialogue data of a user, and inputting the dialogue data into a large language model; receiving and analyzing the dialogue data by the large language model, and outputting key information; the key information is converted into a key text vector through the NLP; calculating the similarity between the key text vector and a predefined category to obtain a similarity value; and classifying the key information according to the similarity value to obtain a classification result. According to the method, the dialogue content of the user is analyzed, the key information is automatically identified and extracted, and then the appeal of the user is deeply analyzed and classified, so that the working efficiency of processing the appeal of the user is improved, and meanwhile, the appeal processing accuracy is also improved.

Description

Technical Field

[0001] The present invention relates to the field of data processing technology, which is applied to dialogue scenarios such as finance or medical treatment, and in particular to a large model-based information processing and classification method, device, equipment and storage medium. Background Art

[0002] With the booming development of the new energy industry, the number of related claims cases is also growing. For insurance companies, how to efficiently and accurately handle these increasing claims cases, as well as how to deeply explore and meet customer demands, has become an urgent problem to be solved. The traditional claims processing process mainly relies on manual review, which is inefficient and error-prone, and can no longer meet the needs of modern insurance business for rapid response and precise processing. Therefore, insurance companies need to explore new technical means and methods to improve the efficiency and accuracy of claims processing, ensure better service to customers and maintain continuous growth of business. Summary of the invention

[0003] The purpose of the embodiments of the present invention is to propose a method, device, equipment and storage medium for information processing and classification based on a large model to solve the problems of inaccurate mining of user demands and low work efficiency in the prior art.

[0004] In order to solve the above technical problems, the embodiment of the present invention provides an information processing and classification method based on a large model, which adopts the following technical solutions:

[0005] A method for information processing and classification based on a large model comprises the following steps:

[0006] Acquire user conversation data, and input the conversation data into a large language model;

[0007] The large language model receives the conversation data for analysis and outputs key information;

[0008] Converting the key information into a key text vector by the NLP;

[0009] Calculating the similarity between the key text vector and the predefined category to obtain a similarity value;

[0010] The key information is classified according to the similarity value to obtain a classification result.

[0011] In one embodiment, the large language model receives the conversation data for analysis and outputs key information, including:

[0012] Setting a preset instruction for the prompt in the large language model;

[0013] Enter the dialog data into the text of prompt;

[0014] The conversation data is analyzed according to the preset instructions, and key information is output.

[0015] In one embodiment, converting the key information into a key text vector by using the NLP includes:

[0016] Preprocessing the key information and adding a mark to each sentence of the key information; wherein the mark includes a start mark, an end mark and a middle mark;

[0017] The pre-processed key information is input into the NLP for processing, and the NLP converts the key information into a key text vector.

[0018] In one embodiment, calculating the similarity between the key text vector and the predefined category to obtain a similarity value includes:

[0019] Calculate the dot product of the key text vector and the predefined category vector to obtain the dot product value;

[0020] Calculate the modulus of the key text vector and the predefined category vector respectively to obtain the value of the modulus of the key text vector and the value of the modulus of the predefined category vector;

[0021] The cosine similarity value is calculated according to the dot product value, the value of the key text vector modulus and the value of the predefined category vector modulus to obtain a similarity value.

[0022] In one embodiment, classifying the key information according to the similarity value to obtain a classification result includes:

[0023] comparing the similarity values ​​between the key information and each of the preset defined categories;

[0024] When the similarity value is greater than the category threshold, the key information is classified into a preset defined category and the classification result is output.

[0025] In one embodiment, after classifying the key information according to the similarity value and obtaining the classification result, the method further includes:

[0026] Matching a plurality of classification labels to the key information;

[0027] The key information of the same label is assigned to the same category to form a label system.

[0028] In one embodiment, the classification tags include loss degree classification, compensation reason analysis, user type classification and time factor classification.

[0029] In order to solve the above technical problems, the embodiment of the present invention also provides an information processing and classification device based on a large model, which adopts the following technical solution:

[0030] An information processing and classification device based on a large model, comprising:

[0031] A data acquisition module, used to acquire user's conversation data and input the conversation data into a large language model;

[0032] A key information module, configured to receive the conversation data through the large language model for analysis and output key information;

[0033] A text vector module, used for converting the key information into a key text vector through the NLP;

[0034] A similarity module, used to calculate the similarity between the key text vector and the predefined category to obtain a similarity value;

[0035] The information classification module is used to classify the key information according to the similarity value to obtain a classification result.

[0036] In order to solve the above technical problems, an embodiment of the present invention further provides a computer device, which adopts the following technical solution:

[0037] A computer device includes a memory and a processor, wherein the memory stores computer-readable instructions, and the processor implements the steps of the above-mentioned large model-based information processing and classification method when executing the computer-readable instructions.

[0038] In order to solve the above technical problems, an embodiment of the present invention further provides a computer-readable storage medium, which adopts the following technical solution:

[0039] A computer-readable storage medium stores computer-readable instructions, which, when executed by a processor, implement the steps of the large model-based information processing and classification method as described above.

[0040] Compared with the prior art, the embodiments of the present invention mainly have the following beneficial effects:

[0041] The information processing and classification method based on the large model described in the embodiment of the present invention obtains the user's conversation data and inputs the conversation data into the large language model; the large language model receives the conversation data for analysis and outputs key information; the key information is converted into a key text vector through the NLP; the similarity between the key text vector and the predefined category is calculated to obtain a similarity value; the key information is classified according to the similarity value to obtain a classification result. The present invention analyzes the user's conversation content, automatically identifies and extracts key information, and then conducts in-depth analysis and classification of the user's demands, thereby improving the work efficiency of processing user demands and the accuracy of demand processing. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] In order to more clearly illustrate the scheme in the present invention, a brief introduction is given below to the drawings required for use in the description of the embodiments of the present invention. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0043] Figure 1 is an exemplary system architecture diagram in which the present invention can be applied;

[0044] Figure 2 is a flow chart of an embodiment of a method for information processing and classification based on a large model according to the present invention;

[0045] Figure 3 is a schematic structural diagram of an embodiment of a large model-based information processing and classification device according to the present invention;

[0046] Figure 4 FIG. 1 is a schematic structural diagram of an embodiment of a computer device according to the present invention. DETAILED DESCRIPTION

[0047] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by technicians in the technical field of the present invention; the terms used in the specification of the application herein are only for the purpose of describing specific embodiments and are not intended to limit the present invention; the terms "including" and "having" and any variations thereof in the specification and claims of the present invention and the above-mentioned drawings are intended to cover non-exclusive inclusions. The terms "first", "second", etc. in the specification and claims of the present invention or the above-mentioned drawings are used to distinguish different objects, not to describe a specific order.

[0048] Reference to "embodiments" herein means that a particular feature, structure, or characteristic described in conjunction with the embodiments may be included in at least one embodiment of the present invention. The appearance of the phrase in various places in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment that is mutually exclusive with other embodiments. It is explicitly and implicitly understood by those skilled in the art that the embodiments described herein may be combined with other embodiments.

[0049] In order to enable those skilled in the art to better understand the solutions of the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings.

[0050] like Figure 1 As shown, the system architecture 100 may include terminal devices 101, 102, 103, a network 104 and a server 105. The network 104 is used to provide a medium for communication links between the terminal devices 101, 102, 103 and the server 105. The network 104 may include various connection types, such as wired, wireless communication links or optical fiber cables, etc.

[0051] Users can use terminal devices 101, 102, 103 to interact with server 105 through network 104 to receive or send messages, etc. Various communication client applications can be installed on terminal devices 101, 102, 103, such as web browser applications, shopping applications, search applications, instant messaging tools, email clients, social platform software, etc.

[0052] Terminal devices 101, 102, 103 can be various electronic devices with display screens and supporting web browsing, including but not limited to smart phones, tablet computers, e-book readers, MP3 players (Moving Picture Experts Group Audio Layer III, Moving Picture Experts Group Audio Layer 3), MP4 (Moving Picture Experts Group Audio Layer IV, Moving Picture Experts Group Audio Layer 4) players, laptop computers and desktop computers, etc.

[0053] The server 105 may be a server that provides various services, such as a background server that provides support for web pages displayed on the terminal devices 101 , 102 , and 103 .

[0054] It should be noted that the information processing and classification method based on the big model provided in the embodiment of the present invention is generally executed by a server / terminal device, and accordingly, the information processing and classification device based on the big model is generally arranged in the server / terminal device.

[0055] It should be understood that Figure 1The number of terminal devices, networks and servers in the embodiment is only for illustration. Any number of terminal devices, networks and servers may be provided according to implementation requirements.

[0056] Continue to refer Figure 2 , shows a flow chart of an embodiment of a method for information processing and classification based on a large model according to the present invention. The method for information processing and classification based on a large model comprises the following steps:

[0057] S1. Obtaining user conversation data, and inputting the conversation data into a large language model;

[0058] S2, the large language model receives the conversation data for analysis and outputs key information;

[0059] S3, converting the key information into a key text vector through the NLP;

[0060] S4, calculating the similarity between the key text vector and the predefined category to obtain a similarity value;

[0061] S.5 classifies the key information according to the similarity value to obtain a classification result.

[0062] In this embodiment, in the scenario of processing user conversation data, the application of large language model (LLM) and natural language processing (NLP) technology makes the extraction, analysis and classification of information more efficient and accurate. First, collect user conversation data from various channels (such as customer service conversations, social media interactions, etc.). These data are the basis for subsequent analysis. Input the collected conversation data into the pre-trained large language model. This model can understand and process natural language and is a key tool for analyzing the content of the conversation. After receiving the conversation data, the large language model analyzes the text content through its complex algorithms and learning mechanisms to extract key information in the conversation. This key information may include user needs, questions, opinions, etc.

[0063] Using NLP technology, the extracted key information is converted into text vectors. Text vectorization is the process of converting text data into numerical form, which enables computers to process and analyze this information. Calculate the similarity between the key text vector and the predefined category vector. This is usually achieved by calculating the cosine similarity between the two vectors, obtaining a value between -1 and 1 to measure their proximity in the vector space. Based on the calculated similarity value, the key information is classified into the most similar predefined category. This step usually involves setting a threshold, and only when the similarity value exceeds this threshold is the key information considered similar enough to a category for classification. Finally, based on the above, the classification results of the key information are obtained. These results can be used for further data analysis, decision support or automated response.

[0064] Through this process, key information can be systematically extracted from a large amount of conversation data and classified to support various application scenarios, such as customer service optimization, market trend analysis, product feedback collection, etc. This approach improves the automation level of data processing, reduces manual intervention, and ensures consistency and accuracy of processing.

[0065] In one embodiment, the large language model receives the conversation data for analysis and outputs key information, including:

[0066] Setting a preset instruction for the prompt in the large language model;

[0067] Enter the dialog data into the text of prompt;

[0068] The conversation data is analyzed according to the preset instructions, and key information is output.

[0069] In this embodiment, Baichuan2-13B is a new open source large-scale language model launched by Baichuan Intelligence. It has achieved excellent performance in general and field benchmarks in Chinese, English and multiple languages ​​through training with 2.6 trillion high-quality tokens. This model not only achieves the best results among models of the same size, but also has powerful dialogue processing capabilities and can be quickly put into use through simple code deployment.

[0070] In practical applications, Baichuan2-13B can guide the model to process and analyze user conversation data through specific prompt design. For example, in the insurance claim scenario, the following prompt template can be used to mine and classify customer demands:

[0071] “The current conversation between customer service and customers is as follows:

[0072] {text}

[0073] Please ensure the accuracy of your intent and do not be affected by ASR speech transcription errors.

[0074] Customer intent is considered valid only when it is confirmed by customer service.

[0075] Based on the above text, list the intent label results at multiple points (such as 1, 2, etc.). At the same time, the final output intent label results do not need to explain unnecessary reasons, nor do they need to include () brackets and colons.

[0076] Please note that this is an insurance claim scenario. Please do not take the customer service's inquiry as the customer's intention. Please note that the output form of the intention is a verb-object phrase, and the claim process in the result is as detailed as possible. From the above dialogue, it can be seen that the customer's detailed intention is..."

[0077] In this template, {text} is a placeholder for user conversation data, and specific conversation content needs to be filled in. The model will analyze the conversation content based on these instructions, identify the customer's intention, and output it in the form of verb-object phrases, such as "apply for claims", "ask about claims progress", etc. In this way, the Baichuan2-13B model can automatically extract key information from conversation data, accurately mine customer demands, and classify them, thereby significantly improving work efficiency and processing accuracy. The application of this technology is of great significance to improving customer service quality and optimizing the claims process.

[0078] Furthermore, in the application of large language models, prompt plays a vital role, providing the model with the instructions and context required to process the input data. These instructions detail how the model should process the input text, what types of information to focus on, and the expected output format. In actual application scenarios, preset instructions may require the model to identify the customer's intent, key demands, or specific questions in the conversation. These instructions are embedded in the prompt template to guide the model to perform targeted text analysis.

[0079] To ensure that the model can understand correctly, the conversation data needs to be formatted into text strings and meet the model's input specifications, including text encoding and segmentation. When the model receives the formatted conversation data, it will perform in-depth analysis based on preset instructions, which may involve complex natural language processing tasks such as language understanding, intent recognition, and sentiment analysis.

[0080] The model extracts key information from the conversation based on the instructions, such as the reason for the claim, the products involved, or the customer's specific requirements. The output key information will follow the format specified in the prompt, such as the verb-object phrase form. Such an output format is conducive to direct application in a variety of scenarios such as automatic response generation, customer service optimization, data analysis and reporting.

[0081] In summary, large language models combined with preset prompts provide a powerful tool for automatically extracting valuable information from large amounts of conversation data. This approach not only improves processing efficiency, but also improves service quality, especially in accurately mining user demands and classifying demand types, significantly improving work efficiency and accuracy.

[0082] In one embodiment, converting the key information into a key text vector by using the NLP includes:

[0083] Preprocessing the key information and adding a mark to each sentence of the key information; wherein the mark includes a start mark, an end mark and a middle mark;

[0084] The pre-processed key information is input into the NLP for processing, and the NLP converts the key information into a key text vector.

[0085] In this embodiment, SentenceBERT (Sentence-BERT, abbreviated as SBERT) is an advanced natural language processing technology that aims to convert sentences into high-quality vector representations. This method is based on the BERT model and optimizes the vector representation of sentences through a fine-tuning process. SBERT provides three main strategies to obtain the vector representation of sentences: CLS vector, average pooling, and maximum pooling. In the CLS vector strategy, the [CLS] tag vector in the BERT model is used as the vector representation of the entire sentence. This special tag is designed to capture the contextual information of the entire input sequence, so it can serve as a holistic representation of the sentence.

[0086] The process of sentence vectorization using SBERT consists of the following steps:

[0087] 1. Input processing: One or more sentences are input into the BERT model. A [CLS] tag is added at the beginning of each sentence and a [SEP] tag is added at the end. If two sentences are processed, a [SEP] tag is also required between them.

[0088] 2. BERT model processing: The input sequence is encoded by the BERT model, and the model outputs the context-dependent representation of each token. BERT's Transformer architecture can understand the relationship between each token in the input sequence.

[0089] 3.CLS vector extraction: The vector corresponding to the [CLS] token is extracted from the output of the BERT model and is used as the representation of the sentence.

[0090] In addition to the CLS vector strategy, SBERT also uses average pooling and maximum pooling strategies to process the output of the BERT model to obtain the vector representation of the sentence. These strategies generate a comprehensive sentence vector by aggregating the tag vectors output by BERT. Compared with the traditional BERT method, using SBERT has significant advantages, especially when a large number of text similarity calculations or text matching tasks are required. SBERT is able to convert each sentence into a fixed-length embedding vector, which significantly improves computational efficiency and processing speed.

[0091] In order to make the text data meet the input requirements of the NLP model, a preprocessing step is essential. This includes adding the [CLS] tag at the beginning of the sentence, the [SEP] tag at the end, and specific tags in the middle of the sentence when necessary. The key information after preprocessing is input into the NLP model, such as SBERT, which converts it into a vector representation. Text vectorization is the process of converting text data into numerical vectors that are able to represent the semantic content of the text in a vector space, facilitating mathematical operations and comparisons. Through vectorization, the model is able to retain the core semantic information of the text while simplifying the structure of the text data for easy machine learning and automated processing.

[0092] In summary, SBERT provides an efficient way to extract key information from text and convert it into vector representation, which is particularly useful for text analysis and processing, especially in application scenarios that need to process large amounts of text data, such as customer service and support systems.

[0093] In one embodiment, calculating the similarity between the key text vector and the predefined category to obtain a similarity value includes:

[0094] Calculate the dot product of the key text vector and the predefined category vector to obtain the dot product value;

[0095] Calculate the modulus of the key text vector and the predefined category vector respectively to obtain the value of the modulus of the key text vector and the value of the modulus of the predefined category vector;

[0096] The cosine similarity value is calculated according to the dot product value, the value of the key text vector modulus and the value of the predefined category vector modulus to obtain a similarity value.

[0097] In this embodiment, in the context of text vector analysis, the dot product can be used to measure the similarity of two texts in the vector space. The dot product, also known as the inner product or scalar product, is an operation that measures the sum of the products of the corresponding components of two vectors. For key text vectors and predefined category vectors, similarity can be evaluated by calculating their dot products. The dot product is obtained by multiplying and accumulating the values ​​in each dimension of the two vectors.

[0098] The modulus of the key text vector and the predefined category vector are calculated respectively to obtain the value of the modulus of the key text vector and the value of the modulus of the predefined category vector. The modulus of a vector, that is, the length of the vector, can be obtained by calculating the square root of the sum of the squares of the components of the vector. Specifically, to calculate the modulus of the key text vector, it is necessary to find the square root of the sum of the squares of the values ​​of each dimension of the vector.

[0099] Cosine similarity is an indicator that measures the similarity of the directions of two vectors, and its value range is between -1 and 1. When the cosine similarity of two vectors is 1, it means that their directions are exactly the same; when it is -1, it means that their directions are completely opposite; when it is 0, it means that the two vectors are orthogonal, that is, there is no correlation between them. The calculation formula of cosine similarity is:

[0100]

[0101] Among them, A and B are two vectors, A·B represents their dot product, and ∥A∥·∥B∥ are their moduli. Through the cosine similarity formula, combined with the calculated dot product value and the modulus of the two vectors, the cosine similarity between the key text vector and the predefined category vector can be calculated. This similarity value reflects the degree of semantic similarity between the key text and the predefined category. In text classification tasks, the category with the highest cosine similarity value is usually selected as the classification result of the text. Through this process, a systematic and vector-based method can be used to quantify the similarity between texts, which is a crucial step in natural language processing and text analysis.

[0102] In one embodiment, classifying the key information according to the similarity value to obtain a classification result includes:

[0103] comparing the similarity values ​​between the key information and each of the preset defined categories;

[0104] When the similarity value is greater than the category threshold, the key information is classified into a preset defined category and the classification result is output.

[0105] In this embodiment, in the text processing and natural language understanding tasks, key information and predefined categories are converted into text. Each key information and each predefined category corresponds to a unique text vector, which can capture the semantic features of the text. In order to measure the similarity between the key information vector and the predefined category vector, we use the cosine similarity formula for calculation.

[0106] Specifically, the calculated cosine similarity value is compared with this preset threshold. If the similarity value is greater than or equal to the threshold, then the key information is considered to match the corresponding category. When the similarity between the key information and a predefined category exceeds this threshold, the system automatically classifies the key information into that category. This decision based on quantitative similarity provides an objective text classification method. When the classification is completed, the system outputs the content of the key information, the category label assigned to it, and the similarity value used for classification. This threshold-based classification method can automatically process large amounts of text data, such as automatically classifying inquiries, feedback, or complaints in customer service, and quickly assigning them to the corresponding processing flow or response strategy.

[0107] Automated text classification not only improves processing efficiency and reduces the need for manual intervention, but also ensures the consistency and accuracy of classification. In addition, the category threshold can be adjusted to optimize the precision and recall of classification according to the performance and needs in the actual application.

[0108] In summary, this text classification system based on similarity measurement and threshold decision provides an effective solution for efficient text processing and data management.

[0109] In one embodiment, after classifying the key information according to the similarity value and obtaining the classification result, the method further includes:

[0110] Matching a plurality of classification labels to the key information;

[0111] The key information of the same label is assigned to the same category to form a label system.

[0112] In this embodiment, when processing the conversation data, the extracted key information is first analyzed in depth to identify its topic, intent or other important features. This step is the basis for understanding customer inquiries and needs, and is crucial for subsequent data processing and response strategies.

[0113] Next, the key information is matched with predefined classification tags according to its content and characteristics. These tags represent different topics, types of questions, service requirements, etc., and are the basis for information classification. Through the matching process, the key information is assigned corresponding tags, thereby achieving preliminary classification of the information. Once the key information is matched to one or more tags, all key information with the same tags will be classified together. This means that all key information about the same topic or issue will be organized in the same category for further processing and analysis. In this way, different information categories can be formed, each containing key information with the same or similar characteristics. This not only helps in the systematic management of information, but also facilitates rapid retrieval and response.

[0114] Building a comprehensive tagging system is the key to effective information management. This system includes all predefined tags and their corresponding key information categories, providing a framework for information classification and retrieval. In the field of customer service, such a tagging system can quickly identify and respond to common customer questions or needs, thereby improving service efficiency and customer satisfaction. In addition, the tagging system also facilitates more in-depth data analysis, such as trend analysis, problem identification, etc., to support decision-making. By analyzing the use of the tagging system and user feedback, the definition and classification logic of tags can be continuously adjusted and optimized to improve the accuracy and efficiency of information matching.

[0115] As new conversation data is continuously generated, the tagging system can be dynamically updated to include emerging topics or issues. This ensures that the information management system can adapt to the ever-changing data environment and maintain its relevance and effectiveness. In summary, the tag matching and classification-based system provides an effective solution for the effective organization and management of key information, especially for processing large amounts of diverse data, such as customer feedback, market research, or social media monitoring. In this way, the efficiency and quality of data processing can be improved, providing strong support for decision-making.

[0116] In one embodiment, the classification tags include loss degree classification, compensation reason analysis, user type classification and time factor classification.

[0117] In this embodiment, tags play an important role in classifying key information in detail.

[0118] The following classification labels and their applications:

[0119] 1. Classification of loss degree:

[0120] This classification label categorizes information based on the severity of the loss suffered by the customer. In scenarios such as insurance claims, understanding the extent of the loss is critical to determining the amount of compensation and the processing process. The extent of the loss can be broken down into subcategories such as minor loss, moderate loss, and major loss to more accurately process claims.

[0121] 2. Classification of compensation reasons:

[0122] This label is used to distinguish different causes of loss or claim. The cause of compensation may include product quality problems, human operational errors, external environmental factors (such as natural disasters, accidents, etc.). By correctly classifying the causes of compensation, insurance companies can better assess responsibilities and optimize risk management strategies.

[0123] 3. User type classification:

[0124] This classification label organizes information based on the identity or type of claimant. User types may include individual customers, corporate customers, partners, etc. This classification helps customize services and communication strategies because different types of users may have different needs and expectations.

[0125] 4. Classification of time factors:

[0126] This labeling takes into account the time factor of the claim or issue, as well as the customer's expectations for processing speed. Time factor classification may include short-term claims (such as emergency assistance, quick damage assessment) and long-term claims (such as long-term treatment, ongoing financial support). This classification helps prioritize time-sensitive cases.

[0127] By automatically classifying key information into these predefined tags, processing efficiency can be improved to ensure timely responses to customer requests. These classification tags also help with data analysis, such as identifying the most common or severe types of losses, analyzing claim trends over a specific time period, etc. Understanding the types of customers and their needs can help optimize customer service processes and provide more personalized services. Integrating these classification tags into a unified tag system makes it possible to systematically manage and retrieve all key information. As the business develops and the market changes, this tag system should be updated regularly to include new classification standards and customer needs.

[0128] In this way, a multi-dimensional classification system can be constructed, which helps to fully understand and effectively manage key information, especially in scenarios where large amounts of complex data need to be quickly processed and responded to. This systematic approach improves the efficiency and quality of data processing and provides strong support for decision-making.

[0129] The embodiments of the present invention can acquire and process relevant data based on artificial intelligence technology. Artificial Intelligence (AI) is the theory, method, technology and application system that uses digital computers or machines controlled by digital computers to simulate, extend and expand human intelligence, perceive the environment, acquire knowledge and use knowledge to obtain the best results.

[0130] The basic technologies of artificial intelligence generally include sensors, dedicated artificial intelligence chips, cloud computing, distributed storage, large model technology, operation / interaction systems, mechatronics, etc. Artificial intelligence software technologies mainly include computer vision technology, robotics technology, biometrics technology, speech processing technology, natural language processing technology, and machine learning / deep learning.

[0131] In an embodiment of the present invention, the user's conversation data is obtained and input into a large language model; the large language model receives the conversation data for analysis and outputs key information; the key information is converted into a key text vector through the NLP; the similarity between the key text vector and a predefined category is calculated to obtain a similarity value; the key information is classified according to the similarity value to obtain a classification result. The present invention analyzes the user's conversation content, automatically identifies and extracts key information, and then conducts in-depth analysis and classification of the user's demands, thereby improving the work efficiency of processing user demands and the accuracy of demand processing.

[0132] Further references Figure 3 , as a response to the above Figure 3 The present invention provides an embodiment of an information processing and classification device based on a large model, and the device embodiment is similar to Figure 3 Corresponding to the method embodiment shown, the device can be specifically applied to various electronic devices.

[0133] like Figure 3 As shown, the information processing and classification device based on the large model described in this embodiment includes: a data acquisition module 11, a key information module 12, a text vector module 13, a similarity module 14 and an information classification module 15. Among them:

[0134] A data acquisition module 11 is used to acquire the user's conversation data and input the conversation data into the large language model;

[0135] A key information module 12, configured to receive the conversation data through the large language model for analysis and output key information;

[0136] A text vector module 13, used for converting the key information into a key text vector through the NLP;

[0137] A similarity module 14 is used to calculate the similarity between the key text vector and the predefined category to obtain a similarity value;

[0138] The information classification module 15 is used to classify the key information according to the similarity value to obtain a classification result.

[0139] like Figure 3 As shown, the data acquisition module 11, the key information module 12, the text vector module 13, the similarity module 14 and the information classification module 15 are connected in sequence. The module referred to in the present invention refers to a series of computer program instruction segments that can perform specific functions, which are more suitable for describing the information processing and classification execution process based on a large model than a program. For the specific implementation method of each module, please refer to the corresponding method embodiment above, which will not be repeated here.

[0140] In one embodiment, the key information module 12 includes:

[0141] A preset instruction unit, used to set a preset instruction for the prompt in the large language model;

[0142] A preset instruction unit is used to input the dialogue data into the text of prompt;

[0143] The data analysis unit is used to analyze the conversation data according to the preset instructions and output key information.

[0144] In one embodiment, the text vector module 13 includes:

[0145] A tag adding unit, used for preprocessing the key information and adding a tag to each sentence of the key information; wherein the tag includes a start tag, an end tag and an intermediate tag;

[0146] The vector conversion unit is used to input the pre-processed key information into the NLP for processing, and the NLP converts the key information into a key text vector.

[0147] In one embodiment, the similarity module 14 includes:

[0148] A dot product unit is used to calculate the dot product of the key text vector and the predefined category vector to obtain a dot product value;

[0149] A modulus value unit, used to calculate the modulus of the key text vector and the predefined category vector respectively, and obtain the value of the modulus of the key text vector and the value of the modulus of the predefined category vector;

[0150] The similarity value unit is used to calculate the cosine similarity value according to the dot product value, the value of the key text vector modulus and the value of the predefined category vector modulus to obtain the similarity value.

[0151] In one embodiment, the information classification module 15 includes:

[0152] A similarity comparison unit, used for comparing the similarity value between the key information and each preset defined category;

[0153] The information classification unit is used to classify the key information into a preset defined category and output a classification result when the similarity value is greater than a category threshold.

[0154] In one embodiment, after classifying the key information according to the similarity value and obtaining the classification result, the method further includes:

[0155] A label matching unit, used to match a plurality of classification labels to the key information;

[0156] The label system unit is used to allocate the key information of the same label to the same category to form a label system.

[0157] In one embodiment, the classification tags include loss degree classification, compensation reason analysis, user type classification and time factor classification.

[0158] In the above embodiment, the present invention discloses an information processing and classification device based on a large model, which belongs to the field of data processing technology. The present invention obtains the user's conversation data, and inputs the conversation data into a large language model; the large language model receives the conversation data for analysis and outputs key information; the key information is converted into a key text vector through the NLP; the similarity between the key text vector and the predefined category is calculated to obtain a similarity value; the key information is classified according to the similarity value to obtain a classification result. The present invention analyzes the user's conversation content, automatically identifies and extracts key information, and then conducts in-depth analysis and classification of the user's demands, thereby improving the work efficiency of processing user demands and the accuracy of demand processing.

[0159] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiments can be implemented by instructing the relevant hardware through computer-readable instructions, and the computer-readable instructions can be stored in a computer-readable storage medium. When the program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, the aforementioned storage medium can be a non-volatile storage medium such as a disk, an optical disk, a read-only memory (ROM), or a random access memory (RAM).

[0160] It should be understood that, although the steps in the flowchart of the accompanying drawings are displayed in sequence as indicated by the arrows, these steps are not necessarily executed in sequence in the order indicated by the arrows. Unless otherwise specified herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least a part of the steps in the flowchart of the accompanying drawings may include multiple sub-steps or multiple stages, and these sub-steps or stages are not necessarily executed at the same time, but can be executed at different times, and their execution order is not necessarily sequential, but can be executed in turn or alternately with other steps or at least a part of the sub-steps or stages of other steps.

[0161] To solve the above technical problems, the embodiment of the present invention also provides a computer device. Figure 4 , Figure 4 This is a basic structural block diagram of the computer device in this embodiment.

[0162] The computer device 8 includes a memory 8a, a processor 8b, and a network interface 8c that are interconnected through a system bus. It should be noted that the figure only shows a computer device 8 having components 8a-8c, but it should be understood that it is not required to implement all the components shown, and more or fewer components can be implemented instead. Among them, those skilled in the art can understand that the computer device here is a device that can automatically perform numerical calculations and / or information processing according to pre-set or stored instructions, and its hardware includes but is not limited to microprocessors, application specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), digital processors (DSPs), embedded devices, etc.

[0163] The computer device may be a computing device such as a desktop computer, a notebook, a PDA, a cloud server, etc. The computer device may interact with a user through a keyboard, a mouse, a remote controller, a touch pad, or a voice control device.

[0164] The memory 8a includes at least one type of readable storage medium, and the readable storage medium includes flash memory, hard disk, multimedia card, card-type memory (for example, SD or DX memory, etc.), random access memory (RAM), static random access memory (SRAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), programmable read-only memory (PROM), magnetic memory, magnetic disk, optical disk, etc. In some embodiments, the memory 8a can be an internal storage unit of the computer device 8, such as a hard disk or memory of the computer device 8. In other embodiments, the memory 8a can also be an external storage device of the computer device 8, such as a plug-in hard disk equipped on the computer device 8, a smart memory card (Smart Med ia Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (FlashCard), etc. Of course, the memory 8a can also include both the internal storage unit of the computer device 8 and its external storage device. In this embodiment, the memory 8a is generally used to store the operating system and various application software installed on the computer device 8, such as a computer-readable instruction of an information processing and classification method based on a large model, etc. In addition, the memory 8a can also be used to temporarily store various data that have been output or are to be output.

[0165] The processor 8b may be a central processing unit (CPU), a controller, a microcontroller, a microprocessor, or other information processing and classification chips based on a large model in some embodiments. The processor 8b is generally used to control the overall operation of the computer device 8. In this embodiment, the processor 8b is used to run the computer-readable instructions stored in the memory 8a or process data, such as computer-readable instructions for running the information processing and classification method based on the large model.

[0166] The network interface 8c may include a wireless network interface or a wired network interface. The network interface 8c is generally used to establish a communication connection between the computer device 8 and other electronic devices.

[0167] The computer device proposed in this embodiment belongs to the field of data processing technology and is applied to dialogue scenarios such as finance or medical treatment. The present invention obtains the user's dialogue data and inputs the dialogue data into a large language model; the large language model receives the dialogue data for analysis and outputs key information; the key information is converted into a key text vector through the NLP; the similarity between the key text vector and the predefined category is calculated to obtain a similarity value; the key information is classified according to the similarity value to obtain a classification result. The present invention analyzes the user's dialogue content, automatically identifies and extracts key information, and then conducts in-depth analysis and classification of the user's demands, thereby improving the work efficiency of processing user demands and the accuracy of demand processing.

[0168] The present invention also provides another embodiment, namely, providing a computer-readable storage medium, which stores computer-readable instructions, and the computer-readable instructions can be executed by a processor to enable the processor to perform the steps of the large model-based information processing and classification method as described above.

[0169] The computer-readable storage medium proposed in this embodiment belongs to the field of data processing technology and is applied to conversation scenarios such as finance or medical treatment. The present invention obtains the user's conversation data and inputs the conversation data into a large language model; the large language model receives the conversation data for analysis and outputs key information; the key information is converted into a key text vector through the NLP; the similarity between the key text vector and the predefined category is calculated to obtain a similarity value; the key information is classified according to the similarity value to obtain a classification result. The present invention analyzes the user's conversation content, automatically identifies and extracts key information, and then conducts in-depth analysis and classification of the user's demands, thereby improving the work efficiency of processing user demands and the accuracy of demand processing.

[0170] Through the description of the above implementation methods, those skilled in the art can clearly understand that the above-mentioned embodiment methods can be implemented by means of software plus a necessary general hardware platform, and of course by hardware, but in many cases the former is a better implementation method. Based on such an understanding, the technical solution of the present invention, or the part that contributes to the prior art, can be embodied in the form of a software product, which is stored in a storage medium (such as ROM / RAM, a magnetic disk, or an optical disk), and includes a number of instructions for enabling a terminal device (which can be a mobile phone, a computer, a server, an air conditioner, or a network device, etc.) to execute the methods described in each embodiment of the present invention.

[0171] Obviously, the embodiments described above are only some embodiments of the present invention, rather than all embodiments. The preferred embodiments of the present invention are given in the accompanying drawings, but they do not limit the patent scope of the present invention. The present invention can be implemented in many different forms. On the contrary, the purpose of providing these embodiments is to make the understanding of the disclosure of the present invention more thorough and comprehensive. Although the present invention has been described in detail with reference to the aforementioned embodiments, for those skilled in the art, it is still possible to modify the technical solutions recorded in the aforementioned specific embodiments, or to perform equivalent replacements for some of the technical features therein. Any equivalent structure made using the contents of the specification and drawings of the present invention, directly or indirectly used in other related technical fields, is also within the scope of patent protection of the present invention.

Claims

1. A method for information processing and classification based on a large model, characterized in that: The steps include: Acquire user conversation data, and input the conversation data into a large language model; The large language model receives the conversation data for analysis and outputs key information; Converting the key information into a key text vector by the NLP; Calculating the similarity between the key text vector and the predefined category to obtain a similarity value; The key information is classified according to the similarity value to obtain a classification result.

2. The information processing and classification method based on a large model according to claim 1 is characterized in that: The large language model receives the conversation data for analysis and outputs key information, including: Setting a preset instruction for the prompt in the large language model; Enter the dialog data into the text of prompt; The conversation data is analyzed according to the preset instructions, and key information is output.

3. The information processing and classification method based on a large model according to claim 2 is characterized in that: The converting the key information into a key text vector by the NLP comprises: Preprocessing the key information and adding a mark to each sentence of the key information; wherein the mark includes a start mark, an end mark and a middle mark; The pre-processed key information is input into the NLP for processing, and the NLP converts the key information into a key text vector.

4. The information processing and classification method based on a large model according to claim 3 is characterized in that: The calculating the similarity between the key text vector and the predefined category to obtain a similarity value includes: Calculate the dot product of the key text vector and the predefined category vector to obtain the dot product value; Calculate the modulus of the key text vector and the predefined category vector respectively to obtain the value of the modulus of the key text vector and the value of the modulus of the predefined category vector; The cosine similarity value is calculated according to the dot product value, the value of the key text vector modulus and the value of the predefined category vector modulus to obtain a similarity value.

5. The method for information processing and classification based on a large model according to claim 4, characterized in that: The classifying the key information according to the similarity value to obtain a classification result includes: comparing the similarity values ​​between the key information and each of the preset defined categories; When the similarity value is greater than the category threshold, the key information is classified into a preset defined category and the classification result is output.

6. The information processing and classification method based on a large model according to claim 5 is characterized in that: After the key information is classified according to the similarity value and the classification result is obtained, the method further includes: Matching a plurality of classification labels to the key information; The key information of the same label is assigned to the same category to form a label system.

7. The information processing and classification method based on a large model according to claim 6 is characterized in that: The classification labels include loss degree classification, compensation cause analysis, user type classification and time factor classification.

8. An information processing and classification device based on a large model, characterized in that: include: A data acquisition module, used to acquire user's conversation data and input the conversation data into a large language model; A key information module, configured to receive the conversation data through the large language model for analysis and output key information; A text vector module, used for converting the key information into a key text vector through the NLP; A similarity module, used to calculate the similarity between the key text vector and the predefined category to obtain a similarity value; The information classification module is used to classify the key information according to the similarity value to obtain a classification result.

9. A computer device, characterized in that: It comprises a memory and a processor, wherein the memory stores computer-readable instructions, and when the processor executes the computer-readable instructions, the steps of the large model-based information processing and classification method as described in any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer-readable instructions, which, when executed by a processor, implement the steps of the large model-based information processing and classification method according to any one of claims 1 to 7.

Images