Question and answer type customer service information intelligent processing system, method and equipment and medium

Abstract

The invention discloses a question and answer type customer service information intelligent processing system, method and device and a medium, and the system comprises an input and output module which is in communication connection with a pre-training large-scale language model through a langchain framework module, and can receive question information input by a user and output answer information given by the pre-training large-scale language model; the policy document vector library is used for storing clause vectors converted from clauses processed by the langchain framework module for clause processing of the document related to the customer service policy; the langchain framework module is used for converting input question information into an input vector, comparing the input vector with a clause vector in a policy document vector library, searching k clauses closest to the input vector through minimum vector distance, and outputting the k clauses to a pre-trained large language model; the method comprises the following steps: pre-training a large language model, taking k clauses as known information, and reasoning answer information corresponding to question information together with the input question information; according to the system, the knowledge of the manual customer service question and answer record document can be efficiently utilized, and corresponding answers conforming to current customer service policy information are provided for users.

Description

Technical Field

[0001] The present invention relates to the field of intelligent customer service systems, and in particular to a question-and-answer type intelligent processing system, method, device and medium for customer service information. Background Art

[0002] Existing service technologies combined with artificial intelligence often stay at the level of customer service products, and require prior design of operation processes such as dialogues, with great limitations. For example, the current traditional customer service system is essentially a question-and-answer system developed based on a process-based system, and cannot respond in a timely manner to changes in customer service policy information. It can only be adjusted manually, with low efficiency. That is, there are problems such as the inability to make good use of unstructured and non-processed knowledge such as artificial customer service question-and-answer records, narrow application scenarios, low data utilization rate, and high manual maintenance costs.

[0003] In view of this, the present invention is specifically proposed. Summary of the Invention

[0004] The purpose of the present invention is to provide a question-and-answer type intelligent processing system, method, device and medium for customer service information, which can make full use of the advantages of large language models, efficiently utilize the knowledge of artificial customer service question-and-answer record documents, realize directly replacing the old customer service policy information with new customer service policy information, give correct answers that conform to the current customer service policy information to users, improve the timeliness and correctness of customer service information processing, and thus solve the problems of narrow application scenarios, low data utilization rate, and high manual maintenance costs of existing intelligent customer service technologies.

[0005] The purpose of the present invention is achieved through the following technical solutions:

[0006] A question-and-answer type intelligent processing system for customer service information, comprising:

[0007] An input / output module, a langchain framework module, a policy document vector library, and a pre-trained large language model; wherein,

[0008] The input / output module is communicatively connected to the pre-trained large language model through the langchain framework module, and can receive question information input by a user and output answer information given by the pre-trained large language model;

[0009] The policy document vector library is communicatively connected to the langchain framework module, and can store the sentence vectors converted from the sentences after the langchain framework module performs sentence segmentation processing on customer service policy-related documents;

[0010] The langchain framework module can convert the question information input by the input-output module into an input vector, compare it with the sentence vectors in the policy document vector library, and search for the k sentences closest to the input vector according to the smallest vector distance after comparison, and output the k sentences and the question information input by the user to the pre-trained large language model;

[0011] The pre-trained large language model can use the k sentences output by the langchain framework module as known information, and jointly reason with the question information input by the user to obtain the answer information corresponding to the question information.

[0012] A processing device includes:

[0013] At least one memory for storing one or more programs;

[0014] At least one processor capable of executing the one or more programs stored in the memory. When the one or more programs are executed by the processor, the processor can implement the method of the present invention.

[0015] A readable storage medium stores a computer program, and when the computer program is executed by a processor, the method of the present invention can be implemented.

[0016] Compared with the prior art, the beneficial effects of the question-and-answer type customer service information intelligent processing system and method provided by the present invention include:

[0017] By using the langchain framework in cooperation with the pre-trained large language model, the knowledge of the artificial customer service question-and-answer records document can be efficiently utilized, directly replacing the old customer service policy information with the new customer service policy information, or separately accessing the customer service policy information in multiple time periods in the form of a time stamp, and giving accurate answers that conform to the current customer service policy rules to the questions raised by users, thereby solving the problems that the existing customer service information processing system cannot automatically perform information extraction, information induction and summary of unformatted document data, and requires separate question-and-answer logic design and question-and-answer system design, etc. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the following drawings are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0019] Figure 1 It is a schematic diagram of the composition of the question-and-answer type customer service information intelligent processing system provided by the embodiment of the present invention.

[0020] Figure 2 This is a flowchart of the intelligent processing method for Q&A customer service information provided by the embodiments of the present invention. Detailed implementation manners

[0021] Next, in combination with the specific content of the present invention, the technical solutions in the embodiments of the present invention will be described clearly and completely; obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments, which does not constitute a limitation to the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts fall within the protection scope of the present invention.

[0022] First, the following explanations will be given to the terms that may be used in this article:

[0023] The term "and / or" means that either one of the two or both can be realized. For example, X and / or Y means that it includes both the case of "X" or "Y" and the three cases of "X and Y".

[0024] Descriptions with semantic meanings such as "comprising", "including", "containing", "having" or other similar terms should be interpreted as non-exclusive inclusion. For example: including a certain technical feature element (such as raw materials, components, ingredients, carriers, dosage forms, materials, dimensions, parts, components, mechanisms, devices, steps, processes, methods, reaction conditions, processing conditions, parameters, algorithms, signals, data, products or articles, etc.) should be interpreted as not only including the clearly listed certain technical feature element, but also including other technical feature elements well-known in the art that are not clearly listed.

[0025] The term "consisting of" means excluding any technical feature element that is not clearly listed. If this term is used in a claim, this term will make the claim a closed type, making it not contain technical feature elements other than the clearly listed ones, except for related conventional impurities. If this term only appears in a sub-clause of a claim, then it only limits the elements clearly listed in that sub-clause, and the elements recorded in other sub-clauses are not excluded from the overall claim.

[0026] Unless otherwise clearly specified or limited, terms such as "installed", "connected", "joined", "fixed", etc. should be understood in a broad sense. For example: it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be directly connected or indirectly connected through an intermediate medium, and it can be the communication inside two elements. For those of ordinary skill in the art, the specific meanings of the above terms in this article can be understood according to specific situations.

[0027] The terms "center", "longitudinal", "lateral", "length", "width", "thickness", "upper", "lower", "front", "rear", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", "outer", "clockwise", "counterclockwise", etc. indicate the orientation or positional relationship based on the orientation or positional relationship shown in the drawings. They are only for the convenience of description and simplification of the description, rather than explicitly or implicitly indicating that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation. Therefore, it should not be construed as a limitation to this article.

[0028] The following will provide a detailed description of the intelligent processing system and method for Q&A customer service information provided by the present invention. The content not described in detail in the embodiments of the present invention belongs to the prior art well-known to those skilled in the art. For the conditions not specified in the embodiments of the present invention, they are carried out according to the conventional conditions in the art or the conditions recommended by the manufacturer. For the reagents or instruments not specified in the embodiments of the present invention for the manufacturer, they are all conventional products that can be obtained through commercial purchase.

[0029] As Figure 1 shown, the embodiments of the present invention provide a Q&A customer service information intelligent processing system, including:

[0030] An input / output module, a langchain framework module, a policy document vector library, and a pre-trained large language model; wherein,

[0031] The input / output module is communicatively connected to the pre-trained large language model through the langchain framework module, and can receive the question information input by the user and output the answer information given by the pre-trained large language model;

[0032] The policy document vector library is communicatively connected to the langchain framework module, and can store the clause vectors converted from the clauses after the langchain framework module processes the customer service policy-related documents into clauses;

[0033] The langchain framework module can convert the question information input by the input / output module into an input vector, compare it with the clause vectors in the policy document vector library, and search for the k clauses closest to the input vector according to the smallest vector distance after comparison, and output the k clauses and the question information input by the user to the pre-trained large language model;

[0034] The pre-trained large language model can use the k clauses output by the langchain framework module as known information, and jointly reason with the question information input by the user to obtain the answer information corresponding to the question information.

[0035] Preferably, in the above system, the sentences obtained by the langchain framework module after clause segmentation of the customer service policy-related documents are converted into sentence vectors by the text2vec method.

[0036] The problem information input by the user through the input-output module of the langchain framework module is also converted into an input vector by the text2vec method.

[0037] Preferably, in the above system, during the inference process of the pre-trained large language model, the Agent module in the langchain framework module is used to maintain the inference process, and the Memory framework in the langchain framework module is used to maintain the historical records before and after.

[0038] Preferably, in the above system, during the inference process of the pre-trained large language model, if the exact answer information cannot be obtained according to the prompt words, the Agent module of the langchain framework module will automatically increase the value of the number of clauses k for the minimum vector distance search until the value of k reaches the maximum preset value. If the exact answer information still cannot be inferred when the value of k reaches the maximum preset value, a question will be asked according to the inference process.

[0039] Preferably, in the above system, the pre-trained large language model uses the ChatGLM open-source large language model.

[0040] Preferably, in the above system, the content of the customer service policy-related documents is the historical customer service question-and-answer information or customer service policy information.

[0041] As Figure 2 shown, the embodiment of the present invention also provides a method for intelligent processing of question-and-answer customer service information, using the above-mentioned question-and-answer customer service information intelligent processing system, including the following steps:

[0042] Receive the problem information input by the user through the input-output module of the system and send it to the langchain framework module of the system;

[0043] Convert the problem information input by the input-output module through the langchain framework module into an input vector, compare it with the sentence vectors converted from the sentences obtained by the langchain framework module after clause segmentation of the customer service policy-related documents stored in the policy document vector library of the system, and search for the k sentences closest to the input vector according to the minimum vector distance after comparison, and output them together with the problem information input by the user to the pre-trained large language model of the system;

[0044] Using the pre-trained large language model, the k clauses output by the langchain framework module are used as known information, and together with the question information input by the user, reasoning is performed to give the answer information corresponding to the question information, and the answer information is sent to the input / output module for output to the user.

[0045] Preferably, in the above method, during the reasoning process of the pre-trained large language model, if no exact answer information can be obtained according to the prompt words, the Agent module of the langchain framework module will automatically increase the value of the number of clauses k for searching until the k value reaches the maximum preset value. If the exact answer information still cannot be inferred when the k value reaches the maximum preset value, questions will be asked according to the reasoning process.

[0046] Preferably, in the above method, the content of the customer service policy-related document is the historical customer service Q&A information or customer service policy information.

[0047] An embodiment of the present invention further provides a processing device, which is characterized by including:

[0048] At least one memory for storing one or more programs;

[0049] At least one processor capable of executing the one or more programs stored in the memory. When the one or more programs are executed by the processor, the processor can implement the above method.

[0050] An embodiment of the present invention further provides a readable storage medium storing a computer program, which can implement the above method when executed by a processor.

[0051] In summary, the system and method of the embodiments of the present invention, by using the langchain framework in cooperation with the pre-trained large language model, can solve problems such as information extraction, information induction and summary, Q&A logic design and Q&A system design of non-formatted document data such as historical stored customer service policy-related documents, facilitating the convenient and efficient utilization of documents such as artificial customer service Q&A records in the intelligent processing system of customer service information, directly realizing the replacement of old policy information with new policy information, or enabling the policy information in multiple time periods to be separately accessed and used in the form of timestamp stamping, and efficiently and automatically providing accurate customer service policy information for users.

[0052] In order to more clearly show the technical solutions provided by the present invention and the technical effects produced, the following takes specific embodiments to describe in detail the Q&A-based intelligent processing system and method of customer service information provided by the embodiments of the present invention.

[0053] Embodiment 1

[0054] Such as Figure 1As shown in the figure, an intelligent processing system for Q&A customer service information provided by an embodiment of the present invention includes:

[0055] An input / output module 1, a langchain framework module 2, a policy document vector library 3, and a pre-trained large language model 4; among them,

[0056] The pre-trained large language model 4 and the langchain framework module 2 constitute the backend core components, which can analyze the knowledge extraction and relationships of knowledge carriers such as customer service policy-related documents. The input / output module can adopt a user interaction interface;

[0057] Among them, the processing of each part is as follows:

[0058] (1) Processing of customer service policy-related documents: Perform clause splitting on the content of customer service policy-related documents. According to the specific content of the input customer service policy-related documents, this clause splitting can use non-machine learning methods or machine learning methods. After clause splitting, use the text2vec method to convert each clause into a vector and save it in the policy document vector library.

[0059] (2) Input / output module: Use the text chat box of the user interaction interface as the input / output module to interact with the user in the form of a text chat box. To achieve unmanned policy-related Q&A, but the most appropriate answer may not be obtained through just one question. For the rigor of question answering, it is also necessary to further ask the user for relevant information when the complete information is not known. The present invention uses the chat box method to interact with the user, so that more accurate answers can be obtained through multiple Q&A sessions when the information is not clear enough.

[0060] (3) Q&A derivation chain based on langchain: This part combines the langchain framework module and the ChatGLM open-source large language model. The combined operation logic of the two is as follows:

[0061] Convert the question text input by the user into an input vector using the text2vec method, and compare it with the clause vectors obtained from the clause splitting of customer service policy-related documents in the policy document vector library. Through the minimum vector distance search, obtain the k clauses closest to the input vector. These k clauses are usually existing Q&A case or customer service policy clause information.

[0062] Take k clauses as known information and input them together with the text of the user's question into the ChatGLM large language model to obtain the text output in the chat box. During the processing, use the Agent module in the langchain framework to maintain the inference process, and use the Memory framework module to maintain the historical records before and after. In prompt engineering, require the large language model to draw reliable conclusions. If reliable conclusions cannot be drawn, the Agent module will automatically increase the value of k for searching until k reaches 10 (this value can be determined by the actual computing power of the configurator). If an exact answer still cannot be analyzed, ask questions based on the inference process.

[0063] Connection relationships of each part: The ChatGLM large language model is communicatively connected to the langchain framework module and can be called by the langchain framework module, and is encapsulated as a module with only input and output interfaces (i.e., the text chat box of the user interaction interface). The front end only uses this input and output interface to maintain the interaction interface, and this interface can be encapsulated as an application, a web page, or a small program.

[0064] The operation process of the system of the present invention adopts a simple question-and-answer method for users, and is coordinated by the ChatGLM large language model and the langchain framework module to infer the correct answer.

[0065] The system and method of the present invention utilize the natural language processing ability of the large language model to realize the extraction of knowledge from unformatted and non-procedural texts and the intelligent processing of customer service information in the form of question and answer based on this knowledge in the specific field of customer service policy information question and answer processing.

[0066] Those of ordinary skill in the art can understand that all or part of the processes in the methods of the above embodiments can be completed by instructing relevant hardware through a program. The program can be stored in a computer-readable storage medium. When the program is executed, it can include the processes of the above method embodiments. Among them, the storage medium can be a magnetic disk, an optical disk, a read-only memory (ROM), or a random access memory (RAM), etc.

[0067] The above is only a preferred specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any changes or substitutions that can be easily thought of by those skilled in the art within the technical scope disclosed by the present invention should be covered by the protection scope of the present invention. Therefore, the protection scope of the present invention should be subject to the protection scope of the claims. The information disclosed in the background art part of this article is only intended to deepen the understanding of the overall background art of the present invention, and should not be regarded as an admission or any form of implication that this information constitutes the prior art known to those skilled in the art.

Claims

1. A question-and-answer customer service information intelligent processing system, characterized in that: include: Input and output modules, langchain framework modules, policy document vector library and pre-trained large language model; among them, The input-output module is connected to the pre-trained large language model through the langchain framework module, and can receive question information input by the user and output answer information given by the pre-trained large language model; The policy document vector library is in communication with the langchain framework module and can store the sentence vectors converted from the sentences processed by the langchain framework module on the customer service policy related documents; The langchain framework module can convert the question information input by the input-output module into an input vector, compare it with the sentence vector in the policy document vector library, and after comparison, search for the k sentences closest to the input vector by the minimum vector distance, and output them to the pre-trained large language model together with the question information input by the user; The pre-trained large language model can use the k sentences output by the langchain framework module as known information and infer the answer information corresponding to the question information together with the question information input by the user.

2. The question-and-answer customer service information intelligent processing system according to claim 1, characterized in that: During the inference process of the pre-trained large language model, the Agent module in the langchain framework module is used to maintain the inference process, and the Memory framework in the langchain framework module is used to maintain the previous and next historical records.

3. The question-and-answer customer service information intelligent processing system according to claim 2, characterized in that: During the reasoning process of the pre-trained large language model, if the exact answer information cannot be obtained based on the prompt word, the Agent module of the langchain framework module will automatically expand the number of sentences k to perform a vector distance minimum search until the k value reaches the maximum preset value. If the exact answer information cannot be inferred when the k value reaches the maximum preset value, questions will be asked according to the reasoning process.

4. The question-and-answer customer service information intelligent processing system according to any one of claims 1 to 3, characterized in that: The pre-trained large language model adopts ChatGLM open source large language model.

5. The question-and-answer customer service information intelligent processing system according to any one of claims 1 to 3, characterized in that: The content of the customer service policy-related document is historically recorded customer service question and answer information or customer service policy information.

6. A question-and-answer customer service information intelligent processing method, characterized in that: The question-and-answer customer service information intelligent processing system according to any one of claims 1 to 5 comprises the following steps: Receive the question information input by the user through the input and output module of the system, and send it to the langchain framework module of the system; The question information input by the input-output module is converted into an input vector through the langchain framework module, and compared with the sentence vectors converted from the sentences processed by the langchain framework module on the customer service policy-related documents stored in the policy document vector library of the system. After the comparison, the k sentences closest to the input vector are searched according to the minimum vector distance, and are output to the pre-trained large language model of the system together with the question information input by the user; The k sentences output by the langchain framework module are used as known information through the pre-trained large language model, and are inferred together with the question information input by the user to give answer information corresponding to the question information, and the answer information is sent to the input-output module to output to the user.

7. The intelligent processing method for question-and-answer customer service information according to claim 6, characterized in that: During the reasoning process of the pre-trained large language model, if the exact answer information cannot be obtained based on the prompt word, the Agent module of the langchain framework module will automatically expand the number of sentences k to search until the k value reaches the maximum preset value. If the exact answer information cannot be inferred after the k value reaches the maximum preset value, questions will be asked according to the reasoning process.

8. The intelligent processing method for question-and-answer customer service information according to claim 6 or 7, characterized in that: The content of the customer service policy-related document is historically recorded customer service question and answer information or customer service policy information.

9. A processing device, characterized in that: include: at least one memory for storing one or more programs; At least one processor can execute one or more programs stored in the memory, and when the one or more programs are executed by the processor, the processor can implement the method described in any one of claims 6-8.

10. A readable storage medium, characterized in that: A computer program is stored, and when the computer program is executed by a processor, the method described in any one of claims 6 to 8 can be implemented.

Images