Interactive question-answering method, computer device, storage medium and program product

Abstract

The present invention relates to the field of artificial intelligence technology, and discloses an interactive question-answering method, a computer device, a storage medium and a program product. The interactive question-answering method comprises: obtaining a question text of a target user; performing intent recognition on the question text using an intent model to obtain intent information of the question text; judging whether the information of the question text is complete according to the intent information; when the information of the question text is incomplete, supplementing the question text with the intent information to generate an updated text; and generating a target answer text of the updated text using a generation model. The present invention can solve the problem of inaccurate intelligent answers when question information is missing.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence technology, and in particular to an interactive question-answering method, a computer device, a storage medium and a program product. Background Art

[0002] With the rapid popularization of the Internet and mobile terminals, online business is increasing, and users' demand for convenient online consultation and complaints is increasing day by day. Therefore, intelligent customer service is the most mature application scenario in the commercialization of artificial intelligence.

[0003] In the process of interacting with customer service systems, real users generally retain the habit of talking to people, including using pronouns to refer to content that appeared in the previous text, or even omitting it, resulting in the loss of key semantics. Existing customer service systems mostly use methods based on heuristic rules or deep learning to deal with semantic missing, but in actual applications, the above methods are difficult to adapt to scenarios with increasing demand for intelligence. Summary of the invention

[0004] In view of this, the present invention provides an interactive question-answering method, a computer device, a storage medium and a program product, which can solve the problem of inaccurate intelligent answers when question information is missing.

[0005] In a first aspect, the present invention provides an interactive question-and-answer method, which includes: obtaining a question text of a target user; using an intention model to perform intent recognition on the question text to obtain intent information of the question text; judging whether the information of the question text is complete based on the intent information; when the information of the question text is incomplete, using the intent information to supplement the question text to generate an updated text; and using a generation model to generate a target answer text of the updated text.

[0006] This embodiment uses the intention model to determine whether the user's question text is complete based on the intention information of the question text, thereby improving the accuracy of text recognition. When the text is incomplete, the text is supplemented with information to form a complete updated text, and the generation model recognizes the updated text to obtain a more accurate target answer text. This can solve the problem of inaccurate intelligent answers when question information is missing.

[0007] In an optional implementation, the intent model is used to perform intent recognition on the question text to obtain the intent information of the question text, including: using the intent model to perform intent recognition on the question text to obtain the intent type of the question text, the intent types include business consultation intent and chat intent; when the intent type is business consultation intent, judging the slot filling type of the business consultation intent, the slot filling type includes slot filling intent and non-slot filling intent; when the business consultation intent is business consultation slot filling intent, obtaining the text slot information of the question text.

[0008] This embodiment can take into account both the accuracy and flexibility of text recognition by identifying the intent type and slot information of the question text through the intent model, and can accurately identify the question intent of the target user and the slot information related to the question intent in the question text.

[0009] In an optional implementation, determining whether the information in the question text is complete based on the intent information includes: obtaining the target slot information corresponding to the business consultation slot filling intent; comparing the text slot information with the target slot information to determine whether there is missing information in the text slot information; when there is missing information in the text slot information, determining that the information in the question text is incomplete.

[0010] This implementation can determine different target slot information according to different business consultation slot filling intentions, and determine whether there is missing slot information in the question text based on the difference between the target slot information and the text slot information. This enables different processing of question texts with different intention information, thereby improving the accuracy of question text integrity recognition.

[0011] In an optional implementation, when the information in the question text is incomplete, the intention information is used to supplement the question text to generate an updated text, including: when the information in the question text is incomplete, the intention information is used to ask the target user to obtain the supplementary slot information corresponding to the missing slot information, and the supplementary slot information is used to fill in the question text to generate an updated text.

[0012] This implementation method determines the missing slot information based on the intent information, and generates supplementary slot information by asking questions to the user based on the slot information. This can solve the problem of low flexibility and efficiency in the traditional customer service system, where the traditional customer service system intelligently asks questions to users one by one through the set question-and-answer steps, and generates different questions for different problems, thereby improving the efficiency of questioning and enhancing the user experience.

[0013] In an optional embodiment, using intent information to ask questions to a target user to obtain supplementary slot information corresponding to the missing slot information includes: generating a question set based on text slot information and target slot information, the question set being a collection of inquiry texts used to inquire about the missing slot information; sending the question set to the target user to obtain user answer text corresponding to the question set; and extracting supplementary slot information corresponding to the missing slot information from the user answer text.

[0014] This implementation generates a corresponding query text based on each missing slot information, asks questions in sequence to generate supplementary slot information, asks questions accurately, improves the accuracy of the query text, and further improves the accuracy of the supplementary slot information.

[0015] In an optional implementation, the interactive question-and-answer method further includes: when the intent type is a small talk intent or a business consultation non-slot-filling intent, using the question text as an updated text.

[0016] This implementation performs different processing based on different intent types. When the intent type is a chat intent or a business consultation non-slot filling intent, the operation of obtaining slot information is not performed, thereby improving the efficiency of question and answering.

[0017] In an optional embodiment, using a generative model to generate a target answer text for an update text includes: obtaining a database text, where the database text is text in a database in the field where the question text is located; using a generative model to perform a similarity comparison between the update text and the database text; and generating a target answer text based on database text that has a high similarity to the update text.

[0018] The generation model of this embodiment generates answers based on the database of the field in which the question text is located, which can obtain the latest business knowledge and ensure the correctness of the answer. At the same time, the data in the database is updated in real time, avoiding the cost of frequent training due to business updates, and improving the usability and versatility of this interactive question-answering method.

[0019] In a second aspect, the present invention provides a computer device, comprising: a memory and a processor, the memory and the processor are communicatively connected to each other, the memory stores computer instructions, and the processor executes the interactive question-and-answer method of the first aspect or any corresponding embodiment thereof by executing the computer instructions.

[0020] In a third aspect, the present invention provides a computer-readable storage medium having computer instructions stored thereon, the computer instructions being used to enable a computer to execute the interactive question-and-answer method of the first aspect or any corresponding embodiment thereof.

[0021] In a fourth aspect, the present invention provides a computer program product, comprising computer instructions for causing a computer to execute the interactive question-and-answer method of the first aspect or any corresponding embodiment thereof. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] In order to more clearly illustrate the specific implementation methods of the present invention or the technical solutions in the prior art, the drawings required for use in the specific implementation methods or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some implementation methods of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0023] Figure 1 is a structural diagram of an interactive question-and-answer device according to an embodiment of the present invention;

[0024] Figure 2 is a flow chart of an interactive question-answering method according to an embodiment of the present invention;

[0025] Figure 3 is a flow chart of another interactive question-and-answer method according to an embodiment of the present invention;

[0026] Figure 4 It is a schematic diagram of the hardware structure of a computer device according to an embodiment of the present invention. DETAILED DESCRIPTION

[0027] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present invention.

[0028] With the rapid popularization of the Internet and mobile terminals, the scale of bank users and online business are growing, and users' demand for online consultation and complaints about bank-related business is increasing day by day. However, the existing bank intelligent customer service system has not yet met the expected requirements in terms of intelligence. One of the main reasons is that users generally retain the habit of talking to people during the interaction with the customer service system, including using pronouns to refer to the content that appeared in the previous text, or even omitting it, resulting in the loss of key semantics. The completion of key semantics is a difficult problem faced by the customer service system, especially in customer service systems with strong professionalism such as bank customer service systems. Due to the lack of familiarity of customers with banking business, incomplete or unclear expressions will significantly affect the accuracy of the customer service system's answers. It is necessary to complete semantics through interactive question-and-answer methods to help the system understand user intentions and get more accurate answers. In recent years, with the rise of big model technology, it has shown strong capabilities in semantic recognition and multi-round dialogue capabilities. Therefore, combining big model technology with intelligent customer service systems is expected to solve this problem.

[0029] The interactive question-and-answer method provided in this embodiment can be applied to the interaction between users and bank customer service systems. Figure 1 , Figure 1 It is a structural diagram of an interactive question-and-answer device according to an embodiment of the present invention.

[0030] The present application provides an interactive question-answering device, which includes a speech receiving module, a speech recognition module, a speech generating module, an analysis and processing module and a knowledge base.

[0031] The voice receiving module is used to receive the target user's question voice after the target user initiates an online conversation and asks a voice question.

[0032] The speech recognition module is used to convert the target user's question speech recognition into question text.

[0033] Among them, the speech generation module is used to convert specific text into speech and send the speech to the target user.

[0034] Among them, the knowledge base includes banking-related knowledge.

[0035] The analysis and processing module includes an intent model and a generation module. The intent model is used to identify the intent of the question text to obtain the corresponding intent type and text slot information. The generation model is used to obtain bank-related professional knowledge based on the knowledge base to generate the target answer text required by the target user. According to an embodiment of the present invention, an implementation example of an interactive question-answering method is provided. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and although a logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in an order different from that here.

[0036] In this embodiment, an interactive question-answering method is provided. Figure 2 , Figure 3 FIG. 1 is a flow chart of an interactive question-answering method according to an embodiment of the present invention. Figure 2 As shown, the process includes the following steps:

[0037] Step S201, obtaining the question text of the target user.

[0038] The question text is the text generated by the target user asking questions about topics of interest.

[0039] In some optional implementations, the question text is text data directly input by the target user; in other optional implementations, the question text is text data recognized based on voice data input by the target user.

[0040] Step S202: Use the intention model to perform intent recognition on the question text to obtain intent information of the question text.

[0041] Among them, the intent model is used to identify the text information of the question text, and obtain the target user's question intention and key information related to the question intention based on the text information.

[0042] Specifically, the intent information obtained by using the intent model to identify the question text includes the intent type of the question text, such as the target user's question intention is to chat or professional business consultation. The intent information also includes key information corresponding to the intent type. In one implementation, the key information is slot information.

[0043] Step S203: determine whether the information in the question text is complete based on the intention information.

[0044] Determine whether the information in the question text is complete based on the intent type and the key information corresponding to the intent type.

[0045] It can be understood that when the intent type of the question text is different, the corresponding key information is also different. When the question text lacks some or all of the key information, the corresponding answer cannot be generated based on the question text. When the key information corresponding to the intent type is complete, the information of the question text is judged to be complete; when the key information corresponding to the intent type is missing, the information of the question text is judged to be incomplete.

[0046] Step S204: When the information in the question text is incomplete, the intention information is used to supplement the question text to generate an updated text.

[0047] When the information in the question text is incomplete, the missing key information is determined according to the intent type and its corresponding key information, and the target user is requested to supplement the information based on the missing key information. The question text is supplemented based on the supplementary information to generate an updated text.

[0048] Step S205, using the generation model to generate a target answer text for the updated text.

[0049] Among them, the generation model recognizes the update text according to the knowledge base and obtains the target answer text corresponding to the update text.

[0050] This embodiment uses the intention model to determine whether the user's question text is complete based on the intention information of the question text, thereby improving the accuracy of text recognition. When the text is incomplete, the text is supplemented with information to form a complete updated text, and the generation model recognizes the updated text to obtain a more accurate target answer text. This can solve the problem of inaccurate intelligent answers when question information is missing.

[0051] See also Figure 3 , Figure 3 FIG. 1 is a flow chart of another interactive question-answering method according to an embodiment of the present invention. Figure 3 As shown, the process includes the following steps:

[0052] Step S301, obtaining the question text of the target user.

[0053] The voice receiving module obtains the target user's question voice, and the voice recognition module recognizes the question voice to obtain the target user's question text.

[0054] In a possible banking business consultation scenario, the target user's question text is obtained as: "I want to purchase foreign exchange, how do I go about doing it?"

[0055] Step S302: Use the intention model to perform intent recognition on the question text to obtain the intent type of the question text.

[0056] The intent model is the ChatGLM2 large model. Before using the intent model to identify the intent of the question text, the ChatGLM2 intent model is trained.

[0057] Specifically, in the instruction fine-tuning stage, according to the requirements of specific tasks, QA dialogues with specific prompts are used as training corpus to train the intent model to output answers that meet the task requirements. Among them, prompt refers to the text or instructions that provide input to the model to guide it to generate specific outputs. At the same time, DeepSpeed ​​technology is used to fine-tune all parameters of the ChatGLM2 large model, and DeepSpeed ​​is used for distributed training to obtain the trained intent model.

[0058] The format of the training instruction set is {user question, intent type, slot information}. "User question" is the specific question text raised by the user; "intent type" is the specific intent corresponding to the question text, where specific intents include chat intent and business consultation intent; "slot information" is information related to the intent type in the question text.

[0059] Among them, business consultation intentions include business consultation slot filling intentions and business consultation non-slot filling intentions. Business consultation slot filling intentions have corresponding slot information, business consultation non-slot filling intentions do not have corresponding slot information, and chatting intentions do not have corresponding slot information.

[0060] It is understandable that the intent type can be added, deleted, and modified according to the specific business. As the business develops and changes, the data in the intent type is updated and adjusted to adapt to new business needs.

[0061] For example, training text 1 is {"What is the amount of RMB that can be exchanged for US dollars?", "Foreign currency exchange", ""}, and its corresponding intent type is business consultation non-slot filling intent: foreign currency exchange.

[0062] Training text 2 is {"How is the weather today?", "Weather inquiry", ""}, and its corresponding intent type is chat intent-weather inquiry.

[0063] Training text 3 is {"I am a Chinese citizen and would like to apply for foreign currency purchase business. How can I do this?", "Foreign exchange purchase business consultation", "Customer type"}, and its corresponding intent type is business consultation slot filling intent - foreign exchange purchase business consultation, and the slot information is "Customer type".

[0064] Training text 4 is {"I am a Chinese citizen and would like to apply for foreign currency purchase for traveling abroad. How can I apply?", "Foreign exchange purchase business consultation", "Customer type, purpose"}, and its corresponding intent type is business consultation slot filling intent - foreign exchange purchase business consultation, and the slot information is "customer type" and "purpose".

[0065] After the present invention constructs a training corpus to train the model, the intent model can accurately identify the user's input, determine the user's specific intent, and the corresponding slot information. The model training is performed on multiple GPUs (Graphics Processing Units) to fully utilize computing resources and improve the scalability of the model.

[0066] The trained intent model is used to identify the intent of the question text to obtain the intent information of the question text, where the intent information includes the intent type and text slot information.

[0067] Specifically, the intent model is used to identify the intent of the question text to obtain the intent type of the question text, wherein the intent types include chat intent, business consultation slot filling intent, and business consultation non-slot filling intent.

[0068] First, determine whether the intent type is small talk intent or business consultation intent. If the intent type is business consultation intent, determine the slot filling type of the business consultation intent. The slot filling type includes slot filling intent and non-slot filling intent.

[0069] When the slot filling type of the business consulting intention is slot filling intention, the business consulting intention is judged to be business consulting slot filling intention; when the slot filling type of the business consulting intention is non-slot filling intention, the business consulting intention is judged to be business consulting non-slot filling intention.

[0070] For example, the intent model identifies the target user's question text "I want to apply for foreign exchange purchase business, how to apply for it", and obtains the intent type as business consultation intent: personal foreign exchange purchase business. Among them, the slot filling type of personal foreign exchange purchase business is slot filling intent, so it is determined that the target user's question text is a business consultation slot filling intent.

[0071] Step S303, when it is determined that the intent type is a business consultation slot filling intent, the text slot information of the question text is obtained.

[0072] Among them, when the intent type is a chat intent or a business consultation non-slot filling intent, the text slot information of the question text is not obtained.

[0073] For example, when the target user's question text is a business consultation slot filling intention, the text slot information in "I want to handle foreign exchange purchase business, how do I handle it" is obtained, and the text slot information is empty.

[0074] This embodiment recognizes the intent type and slot information of the question text through the intent model, which can take into account both the accuracy and flexibility of text recognition, and can accurately recognize the question intent of the target user and the slot information related to the question intent in the question text.

[0075] Step S304: determine whether the information of the question text is complete according to the text slot information.

[0076] Specifically, the target slot information corresponding to the business consultation slot filling intention is obtained; the text slot information and the target slot information are compared to determine whether there is missing information in the text slot information; when there is missing information in the text slot information, it is determined that the information of the question text is incomplete.

[0077] Exemplarily, the target slot information corresponding to the personal foreign exchange purchase business includes: customer type, quota, business type and purpose. The text slot information in "I want to purchase foreign exchange, how do I go about it" is empty. Compare the text slot information with the target slot information, and determine that the text slot information has missing information: customer type, quota, business type and purpose. Then it is determined that the text slot information has missing information, and the information in the question text "I want to purchase foreign exchange, how do I go about it" is incomplete.

[0078] This embodiment can determine different target slot information according to different business consultation slot filling intentions, and determine whether there is missing slot information in the question text based on the difference between the target slot information and the text slot information. This enables different processing of question texts with different intention information, thereby improving the accuracy of question text integrity recognition.

[0079] Step S305, when the information in the question text is incomplete, the intention information is used to ask the target user a question to obtain the supplementary slot information corresponding to the missing slot information, and the supplementary slot information is used to fill in the question text to generate an updated text.

[0080] Specifically, a question set is generated based on the text slot information and the target slot information. The question set is a collection of inquiry texts used to inquire about the missing slot information. That is, one inquiry text is generated according to each missing slot information, and multiple inquiry texts are used as the question set.

[0081] Send a question set to the target user to obtain the user answer text corresponding to each query text in the question set, and extract the supplementary slot information corresponding to the missing slot information from the user answer text. Use the supplementary slot information to supplement the missing slot information in the question text until the missing slot information is complete and generate an updated text.

[0082] Understandably, when the intent type is small talk intent or business consultation non-slot filling intent, the question text is used as the update text. Different processing is performed based on different intent types. When the intent type is small talk intent or business consultation non-slot filling intent, the operation of obtaining slot information is not performed to improve the efficiency of question and answer.

[0083] For example, the missing slot information of the question text "I want to purchase foreign exchange, how do I go about it?" includes: customer type, quota, business type and purpose. Based on each missing slot information, an inquiry text is generated, such as "Hello, are you a domestic customer or a foreign customer?" based on the missing customer type information; "Is the foreign exchange purchase quota you applied for within the facilitation quota?" based on the missing quota information; "Do you need to purchase foreign exchange or cash?" based on the missing business type information; and "What is the purpose of your foreign exchange purchase business?" based on the missing purpose information.

[0084] Send the query texts in the question set to the target users in turn to obtain the corresponding answer texts. The specific process is:

[0085] Customer service system: "Hello, are you a domestic customer or an overseas customer?"

[0086] Target users: “I am a domestic customer.”

[0087] Customer service system: "Please ask whether the foreign exchange purchase amount you applied for this time is within the facilitation quota."

[0088] Target user: “Yes”.

[0089] Customer service system: "Do you need foreign exchange or cash?"

[0090] Target users: “I want to handle foreign exchange business.”

[0091] Customer service system: "What is the purpose of your foreign exchange purchase business?"

[0092] Target users: "I want to exchange some US dollars for studying abroad."

[0093] Based on the target user's answer text, the supplementary slot information is determined as follows: the customer type is a domestic customer, the credit limit is within the facilitation credit limit, the business type is a spot business, and the purpose is to study abroad.

[0094] The missing slot information in the question text is supplemented with the supplementary slot information to generate an updated text: "This customer needs to apply for foreign exchange purchase business and inquire about the specific handling method of foreign exchange purchase business. The customer type of this customer is a domestic customer. The amount applied for this time is within the facilitation amount. The type of business handled is spot business, and the purpose is to study abroad."

[0095] This embodiment determines the missing slot information based on the intention information, and asks the user questions based on the slot information to generate supplementary slot information. This can solve the problem of low flexibility and efficiency in the traditional customer service system, which asks questions to users one by one through the set question-and-answer steps. Generate the corresponding inquiry text based on each missing slot information, ask questions in sequence to generate supplementary slot information, ask questions accurately, improve questioning efficiency, and enhance user experience.

[0096] Step S306: Generate a target answer text of the updated text using the generation model.

[0097] The generated model is connected to an external database, which is a database of the field corresponding to the question text. Specifically, the database includes external knowledge sources such as LangChain. LangChain supports various structured, unstructured, public and private data.

[0098] Get the database text, which is the text in the database of the field where the question text is located. Specifically, load the local database, which includes txt / json / doc / pdf / tables / web pages, etc., and use LangChain to read its relevant content.

[0099] Segment and vectorize the database text. Segment the loaded database text according to the proportion to generate documents. Then vectorize the documents and get the database text vector embedding of each document.

[0100] Vectorize the update text and generate an update text vector query.

[0101] The update text and database text are compared by using the generative model, and the target answer text is generated based on the database text with high similarity to the update text. Specifically, the database text vector embedding constructed above is used to perform similarity search with the update text vector query to obtain the top-k related documents. The documents are concatenated and combined with the update text vector query to form a prompt. Using the Chain component of the Retrieval QA model in Langchain, the selected large language model, the number of documents returned by the similarity search, the document concatenation method, the prompt template and the update text are combined to output the target answer text.

[0102] Referring to the example of step S304, the specific process is as follows: after generating the update text: "This customer needs to apply for foreign exchange purchase business and inquire about the specific handling method of foreign exchange purchase business. The customer type of this customer is a domestic customer, the amount applied for this time is within the facilitation quota, the type of business handled is spot foreign exchange business, and the purpose is to study abroad.", the model is generated to query the langChain database to obtain the handling method of spot foreign exchange business for domestic customers (the application amount is within the facilitation quota, and the purpose is to study abroad). Generate the target answer text.

[0103] Send the target answer text to the target user. In a possible implementation, the speech generation module converts the target answer text into a target answer speech, and sends the target answer speech to the target user.

[0104] The generation model of this embodiment is embedded with the langChain business knowledge database, which can obtain the latest business knowledge by querying the business knowledge database to ensure the correctness of the answer. At the same time, the banking business knowledge inside the business knowledge database can be updated in real time, avoiding the cost of frequent training due to business updates, and improving the usability and versatility of this interactive question-and-answer method.

[0105] The further functional description of each of the above modules and units is the same as that of the above corresponding embodiments and will not be repeated here.

[0106] The interactive question-and-answer device in this embodiment is presented in the form of a functional unit, where the unit refers to an ASIC (Application Specific Integrated Circuit) circuit, a processor and memory that executes one or more software or fixed programs, and / or other devices that can provide the above functions.

[0107] The embodiment of the present invention also provides a computer device having the above Figure 1 The interactive question-and-answer device shown.

[0108] See also Figure 4 , Figure 4 is a schematic diagram of the structure of a computer device provided by an optional embodiment of the present invention, such as Figure 4 As shown, the computer device includes: one or more processors 10, a memory 20, and interfaces for connecting various components, including high-speed interfaces and low-speed interfaces. Various components are connected to each other using different buses for communication, and can be installed on a common mainboard or installed in other ways as needed. The processor can process the instructions executed in the computer device, including instructions stored in or on the memory to display the graphical information of the GUI on an external input / output device (such as, a display device coupled to the interface). In some optional embodiments, if necessary, multiple processors and / or multiple buses can be used together with multiple memories and multiple memories. Similarly, multiple computer devices can be connected, and each device provides some necessary operations (for example, as a server array, a group of blade servers, or a multi-processor system). Figure 4 A processor 10 is taken as an example.

[0109] The processor 10 may be a central processing unit, a network processor or a combination thereof. The processor 10 may further include a hardware chip. The hardware chip may be a dedicated integrated circuit, a programmable logic device or a combination thereof. The programmable logic device may be a complex programmable logic device, a field programmable gate array, a general purpose array logic or any combination thereof.

[0110] The memory 20 stores instructions executable by at least one processor 10, so that at least one processor 10 executes the method shown in the above embodiment.

[0111] The memory 20 may include a program storage area and a data storage area, wherein the program storage area may store an operating system, an application required for at least one function; the data storage area may store data created according to the use of the computer device, etc. In addition, the memory 20 may include a high-speed random access memory, and may also include a non-transient memory, such as at least one disk storage device, a flash memory device, or other non-transient solid-state storage device. In some optional embodiments, the memory 20 may optionally include a memory remotely arranged relative to the processor 10, and these remote memories may be connected to the computer device via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0112] The memory 20 may include a volatile memory, such as a random access memory; the memory may also include a non-volatile memory, such as a flash memory, a hard disk or a solid state drive; the memory 20 may also include a combination of the above types of memory.

[0113] The computer device also includes an input device 30 and an output device 40. The processor 10, the memory 20, the input device 30 and the output device 40 may be connected via a bus or other means. Figure 4 The example of connecting through bus is taken in the following.

[0114] The input device 30 can receive input digital or character information, and generate key signal input related to the user settings and function control of the computer device, such as a touch screen, a keypad, a mouse, a track pad, a touch pad, an indicator bar, one or more mouse buttons, a trackball, a joystick, etc. The output device 40 may include a display device, an auxiliary lighting device (e.g., an LED) and a tactile feedback device (e.g., a vibration motor), etc. The above-mentioned display device includes but is not limited to a liquid crystal display, a light emitting diode, a display and a plasma display. In some optional embodiments, the display device can be a touch screen.

[0115] The computer device further comprises a communication interface 30 for the computer device to communicate with other devices or a communication network.

[0116] The embodiment of the present invention also provides a computer-readable storage medium. The method according to the embodiment of the present invention can be implemented in hardware, firmware, or can be implemented as a computer code that can be recorded in a storage medium, or can be implemented as a computer code that is originally stored in a remote storage medium or a non-temporary machine-readable storage medium and will be stored in a local storage medium through a network download, so that the method described herein can be stored in such software processing on a storage medium using a general-purpose computer, a dedicated processor, or programmable or dedicated hardware. Among them, the storage medium can be a magnetic disk, an optical disk, a read-only storage memory, a random access memory, a flash memory, a hard disk or a solid-state hard disk, etc.; further, the storage medium can also include a combination of the above types of memories. It can be understood that a computer, a processor, a microprocessor controller, or programmable hardware includes a storage component that can store or receive software or computer code. When the software or computer code is accessed and executed by a computer, a processor, or hardware, the method shown in the above embodiment is implemented.

[0117] A part of the present invention may be applied as a computer program product, such as a computer program instruction, which, when executed by a computer, can call or provide the method and / or technical solution according to the present invention through the operation of the computer. Those skilled in the art should understand that the existence of the computer program instruction in a computer-readable medium includes, but is not limited to, a source file, an executable file, an installation package file, etc., and accordingly, the way in which the computer program instruction is executed by the computer includes, but is not limited to: the computer directly executes the instruction, or the computer compiles the instruction and then executes the corresponding compiled program, or the computer reads and executes the instruction, or the computer reads and installs the instruction and then executes the corresponding installed program. Here, the computer-readable medium may be any available computer-readable storage medium or communication medium accessible to the computer.

[0118] Although the embodiments of the present invention have been described in conjunction with the accompanying drawings, those skilled in the art may make various modifications and variations without departing from the spirit and scope of the present invention, and such modifications and variations are all within the scope defined by the appended claims.

Claims

1. An interactive question-answering method, characterized in that: The method comprises: Get the target user's question text; Using the intent model to perform intent recognition on the question text, and obtain the intent type of the question text, the intent type including business consultation intent and chat intent; When the intention type is the business consultation intention, determining a slot filling type of the business consultation intention, the slot filling type including slot filling intention and non-slot filling intention; When the business consultation intention is a business consultation slot filling intention, obtaining text slot information of the question text; Obtain target slot information corresponding to the business consultation slot filling intention; Comparing the text slot information with the target slot information to determine whether there is missing information in the text slot information; When there is missing information in the text slot information, it is determined that the information of the question text is incomplete; When the information of the question text is incomplete, using the intent type to ask the target user a question to obtain supplementary slot information corresponding to the missing slot information, and using the supplementary slot information to fill in the question text to generate an updated text; A target answer text for the updated text is generated using a generative model.

2. The interactive question-answering method according to claim 1, wherein: The step of using the intent type to ask the target user a question to obtain the supplementary slot information corresponding to the missing slot information includes: generating a question set based on the text slot information and the target slot information, wherein the question set is a collection of inquiry texts for inquiring about the missing slot information; Sending the question set to the target user to obtain user answer text corresponding to the question set; The supplementary slot information corresponding to the missing slot information is extracted from the user answer text.

3. The interactive question-answering method according to claim 1, wherein: The method further comprises: When the intent type is a small talk intent or a business consultation non-slot filling intent, the question text is used as the update text.

4. The interactive question-answering method according to any one of claims 1 to 3, characterized in that: The target answer text of the updated text generated by the generation model includes: Obtaining a database text, where the database text is a text in a database in the field where the question text is located; Using the generative model, the updated text and the database text are compared for similarity; The target answer text is generated based on the database text having a high degree of similarity to the update text.

5. A computer device, characterized in that: include: A memory and a processor, wherein the memory and the processor are communicatively connected to each other, the memory stores computer instructions, and the processor executes the interactive question-answering method according to any one of claims 1 to 4 by executing the computer instructions.

6. A computer-readable storage medium, characterized in that The computer-readable storage medium stores computer instructions, and the computer instructions are used to enable a computer to execute the interactive question-answering method according to any one of claims 1 to 4.

7. A computer program product, characterized in that The method comprises computer instructions for causing a computer to execute the interactive question-answering method according to any one of claims 1 to 4.

Images