Intelligent question and answer method and device, electronic equipment, storage medium and program product

Abstract

The invention provides an intelligent question answering method and device, electronic equipment, a storage medium and a program product, and relates to the technical field of data processing.The method comprises the steps that an initial question input by a target user is received, and historical data of the target user is obtained; performing intention recognition on the historical data and the initial question to obtain a target intention corresponding to the initial question; performing retrieval based on the target intention, and determining a target reply tool corresponding to the target intention; and generating target answer content corresponding to the initial question based on the target reply tool, and outputting the target answer content. According to the invention, the answering efficiency and the accuracy of the answering content can be improved.

Description

Technical Field

[0001] The present invention relates to the field of data processing technology, and in particular to an intelligent question-answering method, device, electronic device, storage medium, and program product. Background Art

[0002] In areas such as financial sales and customer service, automatically answering customer needs can effectively improve efficiency. In related technologies, different answer contents are preset for different functions in the answering system, and the questions input by the user are identified to obtain the identification results. Then, the corresponding answer content is selected from the preset answer contents based on the identification results. However, in related technologies, the questions input by the user are for different products or different functions, and the user needs to select the function corresponding to the question in order to accurately identify the user's intention. It is impossible to quickly identify the user's intention, resulting in low answer efficiency. At the same time, the answer content that is directly identified and determined for the user's question may not accurately identify the user's question intention, which in turn leads to a low accuracy rate of the answer content.

[0003] It can be seen that the relevant technology has problems such as low answering efficiency and low accuracy of answer content. Summary of the Invention

[0004] The embodiments of the present invention provide an intelligent question-answering method, device, electronic device, storage medium, and program product to address the problems of low answering efficiency and low accuracy of answer content in related technologies.

[0005] To solve the above initial problem, the present invention is achieved as follows:

[0006] In a first aspect, an embodiment of the present invention provides an intelligent question-answering method, comprising:

[0007] receiving an initial question input by a target user and obtaining historical data of the target user;

[0008] Performing intent recognition on the historical data and the initial question to obtain a target intent corresponding to the initial question;

[0009] Performing a search based on the target intent to determine a target response tool corresponding to the target intent;

[0010] Generate target answer content corresponding to the initial question based on the target reply tool, and output the target answer content.

[0011] In a second aspect, an embodiment of the present invention further provides an intelligent question-answering device, comprising:

[0012] A receiving module, configured to receive an initial question input by a target user and obtain historical data of the target user;

[0013] an identification module, configured to perform intent identification on the historical data and the initial question to obtain a target intent corresponding to the initial question;

[0014] A retrieval module, configured to perform retrieval based on the target intent and determine a target response tool corresponding to the target intent;

[0015] An output module is used to generate target answer content corresponding to the initial question based on the target reply tool and output the target answer content.

[0016] In a third aspect, the present invention also provides an electronic device comprising a processor, a memory, and a computer program stored in the memory and executable on the processor, wherein the computer program, when executed by the processor, implements the steps of the intelligent question-answering method as described in the first aspect above.

[0017] In a fourth aspect, the present invention further provides a computer-readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the steps in the intelligent question-answering method described in the first aspect above are implemented.

[0018] In a fifth aspect, the present invention further provides a computer program product comprising computer instructions, which, when executed by a processor, implement the steps of the intelligent question-answering method as described in the first aspect above.

[0019] In an embodiment of the present invention, an initial question input by a target user is received, and historical data of the target user is obtained; intent recognition is performed on the historical data and the initial question to obtain the target intent corresponding to the initial question; a search is performed based on the target intent to determine the target reply tool corresponding to the target intent; target answer content corresponding to the initial question is generated based on the target reply tool, and the target answer content is output. In this way, by performing intent recognition on the question input by the user to determine the corresponding target reply tool, it is avoided to directly determine the answer content through the intent recognition result, thereby improving the accuracy of intent recognition; then, the target answer content corresponding to the user's initial question is generated through the target reply tool, thereby achieving an accurate answer to the initial question input by the user, thereby improving the accuracy of answering the user's question. At the same time, there is no need to select the function corresponding to the question during the answering process, thereby improving the efficiency of answering. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments of the present invention. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative labor.

[0021] Figure 1 This is a flow chart of an intelligent question-answering method provided by an embodiment of the present invention;

[0022] Figure 2 Schematic diagram of intention recognition and determination of corresponding target tools provided by an embodiment of the present invention;

[0023] Figure 3 This is a schematic diagram of the structure of the professional knowledge type provided by an embodiment of the present invention;

[0024] Figure 4 This is a flowchart corresponding to the product knowledge intent provided by an embodiment of the present invention;

[0025] Figure 5 This is a flowchart corresponding to the company knowledge intent provided by an embodiment of the present invention;

[0026] Figure 6 This is a flowchart corresponding to the product catalog knowledge intent provided by an embodiment of the present invention;

[0027] Figure 7 This is a flowchart corresponding to the knowledge intent of the question-answering system provided by an embodiment of the present invention;

[0028] Figure 8 This is the process corresponding to the product channel knowledge intent provided by the embodiment of the present invention;

[0029] Figure 9 This is a flowchart corresponding to the tax type provided by an embodiment of the present invention;

[0030] Figure 10 This is a flow chart corresponding to the income type provided by an embodiment of the present invention;

[0031] Figure 11 is a flow chart for determining whether to answer an initial question provided by an embodiment of the present invention;

[0032] Figure 12 This is a flowchart corresponding to an unrecognizable intent provided by an embodiment of the present invention;

[0033] Figure 13 It is a flowchart corresponding to the product comparison intention, product recommendation intention, or product planning intention provided by an embodiment of the present invention;

[0034] Figure 14Schematic diagram of the intelligent question-answering process provided by an embodiment of the present invention;

[0035] Figure 15 is a structural diagram of an intelligent question-answering device provided by an embodiment of the present invention;

[0036] Figure 16 This is a structural diagram of an electronic device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0037] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of them. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0038] See Figure 1 , Figure 1 This is a flow chart of an intelligent question-answering method provided by an embodiment of the present invention. Figure 1 As shown, the following steps are included:

[0039] Step 101: Receive an initial question input by a target user and obtain historical data of the target user.

[0040] The initial question is the question the target user asks the intelligent question-answering system. The user may enter the initial question in text or in natural language. The historical data is the historical questions that the user has entered before entering the initial question into the intelligent question-answering system.

[0041] Among them, the intelligent question and answer system is a system that executes the intelligent question and answer method. Different response tools are configured in the intelligent question and answer system to answer different questions through different response tools.

[0042] Step 102: perform intent recognition on the historical data and the initial question to obtain the target intent corresponding to the initial question.

[0043] It should be noted that some initial questions include the user's intention, and the target intention can be obtained by directly identifying the user's initial questions; but there are also cases where some initial questions do not include the user's intention, or the included user intention is unclear. At this time, it is necessary to supplement and correct the content of the initial questions by combining the initial questions with historical data, and then perform intent identification on the supplemented or corrected questions to obtain the target intention corresponding to the initial questions.

[0044] In some embodiments, intent recognition is performed on the historical data and the initial question to obtain the target intent corresponding to the initial question. The historical data and the initial question can be retrieved through dynamic example retrieval and / or professional terminology retrieval to determine the target intent corresponding to the initial question.

[0045] Step 103: Search based on the target intent to determine the target response tool corresponding to the target intent.

[0046] The above-mentioned target reply tool is a preset reply tool in the intelligent question-answering system, which is used to reply to the questions input by the user. Figure 2 The intelligent question-answering system shown includes multiple response tools. Different response tools can respond to different questions. The target response tool that responds to the initial question needs to be determined based on the target intent.

[0047] In some embodiments, searching based on the target intent and determining the target response tool corresponding to the target intent can be achieved through a large model preset in the intelligent question-answering system, where the input of the preset large model is historical data and the initial question, and the output is the target response tool.

[0048] In one embodiment, searching based on the target intent and determining the target response tool corresponding to the target intent includes:

[0049] Searching a preset mapping table based on the target intent to determine the target response tool corresponding to the target intent;

[0050] Among them, the preset mapping table includes multiple intentions and the identification of the response tool corresponding to each intention, and the multiple intentions include at least one of unrecognizable intentions, product-related intentions, professional knowledge-related intentions, tax benefit-related intentions, product comparison intentions, product recommendation intentions, product planning intentions, product knowledge intentions, company knowledge intentions, common sense intentions, question-and-answer system knowledge intentions, product catalog knowledge intentions, and product channel knowledge intentions.

[0051] The above-mentioned preset mapping table includes at least one of unrecognizable intents, product-related intents, professional knowledge-related intents, tax benefit-related intents, product comparison intents, product recommendation intents, product planning intents, product knowledge intents, company knowledge intents, common sense intents, question-and-answer system knowledge intents, product catalog knowledge intents, and product channel knowledge intents, as well as the identifier of the response tool corresponding to each intent. The preset mapping table is used to determine the response tool corresponding to the target intent.

[0052] In an embodiment of the present invention, the different intents included in the preset mapping table are targeted at different application scenarios. By mapping different intents with different reply tools, a better reply effect can be achieved when replying to the initial question of the corresponding application scenario through the reply tool. For example, the target reply tool is a reply tool corresponding to the product recommendation intent, which is targeted at the application scenario of product recommendation. In the case where the initial question is for the user to find a suitable product based on their own needs, replying through the reply tool corresponding to the product recommendation intent can achieve a better reply effect.

[0053] Step 104: Generate target answer content corresponding to the initial question based on the target response tool, and output the target answer content.

[0054] In an embodiment of the present invention, an initial question input by a target user is received, and historical data of the target user is obtained; intent recognition is performed on the historical data and the initial question to obtain the target intent corresponding to the initial question; a search is performed based on the target intent to determine the target reply tool corresponding to the target intent; target answer content corresponding to the initial question is generated based on the target reply tool, and the target answer content is output. In this way, by performing intent recognition on the question input by the user to determine the corresponding target reply tool, it is avoided to directly determine the answer content through the intent recognition result, thereby improving the accuracy of intent recognition; then, the target answer content corresponding to the user's initial question is generated through the target reply tool, thereby achieving an accurate answer to the initial question input by the user, thereby improving the accuracy of answering the user's question. At the same time, there is no need to select the function corresponding to the question during the answering process, thereby improving the efficiency of answering.

[0055] In one embodiment, when the target intent is the professional knowledge-related intent, the product knowledge intent, the company knowledge intent, the common sense intent, the question-answering system knowledge intent, or the product catalog knowledge intent, the target response tool is a knowledge retrieval tool, and generating the target answer content corresponding to the initial question based on the target response tool includes:

[0056] Determine the type of expertise corresponding to the target intent;

[0057] Generate the target answer content based on the component corresponding to the professional knowledge type in the knowledge retrieval tool;

[0058] Among them, the professional knowledge types are general professional knowledge types for commercial maintenance, general professional knowledge types for traditional insurance, general professional knowledge types for individual maintenance, general professional knowledge types for dividend insurance, professional knowledge question and answer types for commercial maintenance, business development knowledge question and answer types for traditional insurance, professional knowledge question and answer types for individual maintenance or professional knowledge question and answer types for dividend insurance.

[0059] The above professional knowledge types are general professional knowledge types for raising, general professional knowledge types for traditional insurance, general professional knowledge types for individual raising, general professional knowledge types for dividend insurance, professional knowledge question and answer types for commercial raising, professional knowledge question and answer types for traditional insurance business development, professional knowledge question and answer types for individual raising or professional knowledge question and answer types for dividend insurance, such as Figure 3 As shown, each professional knowledge type corresponds to different components, and different questions can be answered through the corresponding components.

[0060] In one embodiment, when the target intent is product knowledge intent, generating the target answer content based on the component corresponding to the professional knowledge type in the knowledge retrieval tool includes:

[0061] Determining a first product corresponding to the target intent;

[0062] Completing the initial question based on the historical data using the large model in the knowledge retrieval tool to obtain a first completed question;

[0063] enhancing the initial question based on the database corresponding to the first product in the knowledge retrieval tool and the historical data to obtain a first enhanced question;

[0064] Combining the first completion question and the first enhancement question to obtain a first focusing question;

[0065] The target answer content corresponding to the first focused question is generated based on the answer knowledge base corresponding to the professional knowledge type in the knowledge retrieval tool.

[0066] In an embodiment of the present invention, the first product corresponding to the target intent is determined; the initial question is completed based on the historical data using the large model in the knowledge retrieval tool to obtain a first completed question; the initial question is enhanced based on the database corresponding to the first product in the knowledge retrieval tool and the historical data to obtain a first enhanced question; the first completed question and the first enhanced question are combined to obtain a first focused question; and the target answer content corresponding to the first focused question is generated based on the answer knowledge base corresponding to the professional knowledge type in the knowledge retrieval tool. In this way, by completing and enhancing the initial question to obtain the first focused question, the first focused question can focus more on the relevant knowledge corresponding to the product, thereby generating more accurate target answer content.

[0067] Specifically, such as Figure 4As shown, when the target intent is determined to be product knowledge intent, the large model is used to extract products and identify the first product. The large model is trained to identify products and can be obtained using a product database and specialized terminology corresponding to different product types. The large model is fed with the user's initial question and historical data, which it processes and outputs the identifier of the first product, thereby identifying the first product corresponding to the target intent.

[0068] Furthermore, after determining the first product, the initial question is completed and enhanced through different large models to obtain the first completed question and the first enhanced question. The first completed question and the first enhanced question are then combined through the large model to obtain the first focused question. The target answer content corresponding to the first focused question is then generated through the answer knowledge base.

[0069] In some embodiments, the first focus question can be used in the case of product recall. That is, when business personnel need to recall a product, they use the knowledge retrieval tool of the intelligent question-answering system to determine the products that need to be recalled and sort the products that need to be recalled, so that business personnel can determine the importance of product recall through recall sorting.

[0070] In some implementations, generating the target answer content corresponding to the first focused question based on the answer knowledge base corresponding to the professional knowledge type in the knowledge retrieval tool includes:

[0071] generating an intermediate answer content corresponding to the first focused question based on an answer knowledge base corresponding to the professional knowledge type in the knowledge retrieval tool;

[0072] Determine whether the first focused question is a preset type of question through the large model;

[0073] When the first focused question is not a question of the preset type, setting the intermediate answer content as the target answer content;

[0074] In the case where the first focused question is the preset type of question, the intermediate answer content is completed through the large model to obtain the target answer content.

[0075] It should be noted that for special types of questions (such as underwriting questions), simply generating answer content by answering the knowledge base cannot accurately meet the user's needs. The generated answer content needs to be further supplemented to obtain more accurate target answer content.

[0076] Among them, the above-mentioned large model is obtained by training professional knowledge fragments corresponding to preset type questions, and the professional knowledge fragments corresponding to preset type questions can be retrieved from the database corresponding to the preset type questions.

[0077] In one embodiment, when the target intent is the company knowledge intent, the common sense intent, the question-answering system knowledge intent, or the product catalog knowledge intent, generating the target answer content based on the component corresponding to the professional knowledge type in the knowledge retrieval tool includes:

[0078] Completing the initial question based on the historical data using the large model in the knowledge retrieval tool to obtain a second completed question;

[0079] Based on the answer knowledge base corresponding to the professional knowledge type, the target answer content corresponding to the second completion question is generated.

[0080] In this embodiment of the present invention, the large model in the knowledge retrieval tool completes the initial question based on the historical data to generate a second completed question. The target answer content corresponding to the second completed question is generated based on the answer knowledge base corresponding to the professional knowledge type. In this way, by completing the initial question and then generating the target answer content corresponding to the second completed question, the accuracy of the target answer content is improved.

[0081] Specifically, such as Figure 5 、 Figure 6 and Figure 7 As shown, when the target intent is the company knowledge intent, the common sense intent, the question-answering system knowledge intent, or the product catalog knowledge intent, the large model is used to complete the question, and the target answer content is generated using the corresponding answer knowledge base. The components corresponding to the company knowledge intent and the common sense intent have the same structure.

[0082] In one embodiment, Figure 5 As shown, when the target intent is the company knowledge intent or the common sense intent, the target answer content corresponding to the second completion question is generated based on the answer knowledge base corresponding to the professional knowledge type, including:

[0083] generating an intermediate answer content corresponding to the second completion question based on the answer knowledge base corresponding to the professional knowledge type;

[0084] Determining whether the intermediate answer content satisfies the initial question by using the large model;

[0085] If the intermediate answer content satisfies the requirement of answering the initial question, setting the intermediate answer content as the target answer content;

[0086] In the case that the intermediate answer content does not satisfy the requirements for answering the initial question, the intermediate answer content is supplemented by the large model to obtain the target answer content.

[0087] In an embodiment of the present invention, an intermediate answer corresponding to the second completed question is generated based on the answer knowledge base corresponding to the professional knowledge type; the macro model is used to determine whether the intermediate answer satisfies the initial question; if the intermediate answer satisfies the initial question, the intermediate answer is set as the target answer; if the intermediate answer does not satisfy the initial question, the macro model is used to complete the intermediate answer to obtain the target answer. In this way, the macro model is used to complete the intermediate answer to further improve the accuracy of the target answer.

[0088] Among them, the large model used to complete the intermediate answer content has the ability to connect to the Internet, and obtains relevant data from the network to complete the intermediate answer content and obtain the target answer content.

[0089] In some embodiments, as Figure 6 As shown, when the target intent is the product catalog knowledge intent, generating the target answer content corresponding to the second completion question based on the answer knowledge base corresponding to the professional knowledge type includes:

[0090] Determining, by a large model, a product attribute corresponding to the first product based on the initial question and the historical data;

[0091] Filtering from a preset product database based on the product attributes to obtain an initial product list;

[0092] A target product list is generated based on the initial product list and the completion question using a large model.

[0093] In an embodiment of the present invention, an initial product list is determined by product attributes, and then a target product list is generated, so that different products in the target product list all match the product attributes of the first product, thereby improving the accuracy of the target answer content.

[0094] The preset product database may include a database of products on sale and a database of products that have been discontinued. The specific product database may be determined based on the product data, and then an initial product list may be obtained.

[0095] In one embodiment, Figure 8As shown, when the target intent is the product channel knowledge intent, the component corresponding to the professional knowledge type in the knowledge retrieval tool generates the target answer content, including:

[0096] Identify the initial question using the large model in the knowledge retrieval tool to obtain the product channel type corresponding to the initial question;

[0097] In a case where the product channel type is a channel operation type, generating the target answer content corresponding to the initial question based on the answer knowledge base corresponding to the professional knowledge type in the knowledge retrieval tool;

[0098] In a case where the product channel type is a sales region type, determining a second product corresponding to the target intent;

[0099] Completing the initial question based on the historical data using the large model in the knowledge retrieval tool to obtain a third completed question;

[0100] enhancing the initial question based on the database corresponding to the second product in the knowledge retrieval tool and the historical data to obtain a second enhanced question;

[0101] Combining the third completion question and the second enhancement question to obtain a second focusing question;

[0102] Generate the target answer content corresponding to the second focused question based on the answer knowledge base corresponding to the professional knowledge type in the knowledge retrieval tool

[0103] In an embodiment of the present invention, the initial question is identified by the large model in the knowledge retrieval tool to obtain the product channel type corresponding to the initial question; when the product channel type is a channel operation type, the target answer content corresponding to the initial question is generated based on the answer knowledge base corresponding to the professional knowledge type in the knowledge retrieval tool; when the product channel type is a sales area type, the second product corresponding to the target intention is determined; the initial question is completed based on the historical data by the large model in the knowledge retrieval tool to obtain a third completed question; the initial question is enhanced based on the database corresponding to the second product in the knowledge retrieval tool and the historical data to obtain a second enhanced question; the third completed question and the second enhanced question are combined to obtain a second focused question; the target answer content corresponding to the second focused question is generated based on the answer knowledge base corresponding to the professional knowledge type in the knowledge retrieval tool. In this way, different answers are given for different product channel types, and question completion and question enhancement are combined to improve the accuracy of the obtained target answer content.

[0104] In some embodiments, when the product channel type is a question about a channel selling a product, a preset wording is set as the target answer content. For example, the preset wording may be "Thank you for your recognition of product XXX. Currently, this product does not have corresponding channel cooperation. If you need more information, please contact the channel staff."

[0105] In one embodiment, Figure 9 and Figure 10 As shown, when the target intention is the tax benefit-related intention, the target response tool is a knowledge retrieval tool and a product card tool, and the target answer content corresponding to the initial question generated based on the target response tool includes:

[0106] Determining a target type corresponding to the target intent;

[0107] Determining at least one product corresponding to the initial question based on the large model in the knowledge retrieval tool;

[0108] generating the target answer content based on the target type and the components corresponding to the at least one product in the product card tool;

[0109] The target type is a tax type or a revenue type.

[0110] In an embodiment of the present invention, a target type corresponding to the target intent is determined; at least one product corresponding to the initial question is determined based on the macro model in the knowledge retrieval tool; and target answer content is generated based on the target type and the components corresponding to the at least one product in the product card tool. The target type is a tax type or a revenue type. Thus, target answer content for the at least one product is generated using the tax type or revenue type card tool.

[0111] In some embodiments, as Figure 9 As shown, in the case where it is determined that the target type corresponding to the target intent is a tax type, before determining at least one product corresponding to the initial question based on the big model in the knowledge retrieval tool through the big model, the method further includes:

[0112] Determine through a large model whether the initial problem is a personal pension tax problem;

[0113] In the case where the initial question is about personal pension tax, the target answer content is generated through preset tax preferential cards and common sense tools;

[0114] Determining at least one product corresponding to the initial question based on the large model in the knowledge retrieval tool by using the large model includes:

[0115] In the case where the initial question is not a personal pension tax question and includes multiple products, determining at least one product corresponding to the initial question based on the large model in the knowledge retrieval tool through the large model;

[0116] When the initial question is not a personal pension tax issue and only includes one product, the preset script is output.

[0117] The above-mentioned at least one product may include personal pension tax products or other products. When personal pension tax is included, the target answer content is generated through preset tax incentive cards and common sense tools; when multiple other products are included, a product selection card is generated through other products, and the product selection card contains multiple other products for user selection.

[0118] The above preset scripts include the fact that there are no tax benefits for non-personal pension products.

[0119] It should be noted that the preset tax incentive cards and common sense tools are pre-configured components, which can be models or databases. The target answer content is generated through the preset cards and common sense tools.

[0120] In some embodiments, as Figure 10 As shown, when it is determined that the target type corresponding to the target intent is a revenue type, generating the target answer content based on the target type and the component corresponding to the at least one product in the product card tool includes:

[0121] If the quantity of the at least one product is greater than 1, generating a product selection card, the product selection card including the at least one product;

[0122] When the quantity of the at least one product is 1, determining whether the product is a commercial pension product;

[0123] In the case where the product is not a commercial pension product, the target answer content is generated through a trial calculation card corresponding to the product, and the trial calculation card is used to calculate the benefits of the product;

[0124] In the case where the product is a commercial elderly care product, determining whether the initial question includes product details corresponding to the product;

[0125] If the initial question does not include product details corresponding to the product, generating target answer content through a product details card, wherein the product details card is used to introduce the product;

[0126] In a case where the initial question includes product details corresponding to the product, target answer content is generated through a net value card, and the net value card includes the net value of the commercial pension product.

[0127] In some embodiments, when the determination based on the large model in the knowledge retrieval tool of at least one product corresponding to the initial question fails, that is, no product is extracted, target answer content is generated through preset words, and the preset words are used to request the user to provide more information.

[0128] In this way, the target answer content can be generated through the above-mentioned different cards and tools.

[0129] In some embodiments, as Figure 11 As shown, the initial question is secondary identified through the large model to determine whether it falls within the scope of allowed answers. If not allowed, the answer is refused. If allowed, the answer is given through the preset tool.

[0130] In one embodiment, Figure 12 In the case where the target intent is the unrecognizable intent, the target response tool is an intent clarification tool, and generating target answer content corresponding to the initial question based on the target response tool includes the following:

[0131] Analyze the initial question based on the large model in the intent clarification tool to obtain a first answer, where the first answer is used to request the user to provide additional descriptive information;

[0132] A second answer is obtained from a preset response in the intent clarification tool, where the second answer is used to display customer service contact information.

[0133] In this embodiment of the present invention, the initial question is analyzed based on the large model in the intent clarification tool to obtain a first answer, which is used to request the user to provide additional descriptive information. A second answer is obtained from the preset responses in the intent clarification tool, and the second answer is used to display customer service contact information. In this way, if the intent of the initial question cannot be identified, the first or second answer can be used to request the user to provide additional descriptive information or display customer service contact information, allowing the user to provide a clearer description or resolve the issue with customer service.

[0134] Among them, the customer service contact information can be displayed through a QR code image, or through a phone number.

[0135] In one embodiment, Figure 13As shown, when the target intention is the product comparison intention, the product recommendation intention, or the product planning intention, the target response tool is a product card tool, and the target answer content corresponding to the initial question is generated based on the target response tool, including:

[0136] determining a third product corresponding to the target intent;

[0137] Analyzing the product corresponding to the third product based on the preset components in the product card tool to obtain the target answer content;

[0138] Among them, the preset components include at least one of a product comparison component, a product recommendation component, a product planning component and a product preset speech component.

[0139] In an embodiment of the present invention, the product corresponding to the third product is analyzed by at least one of a product comparison component, a product recommendation component, a product planning component, and a product preset speech component to obtain the target answer content.

[0140] The product comparison component is used for product introduction and selection, while the product recommendation component is used for product selection. These two components are combined to generate the target answer content. The product planning component is used to plan the future use of different products. For example, for pension products, it can calculate basic pensions, occupational pensions, and pension gaps. The product preset script component package includes preset product recommendation scripts.

[0141] In some embodiments, as Figure 14 As shown, in the present invention, after receiving the initial question, the intelligent question-answering system obtains the target intent through professional terminology retrieval and dynamic example retrieval, combined with intent recognition, and determines the product corresponding to the target intent through product extraction. The initial question is optimized through question completion and question enhancement, and processed through preset product response tools such as product-related intent and professional knowledge-related intent to obtain the target answer content. Among them, the response tool can realize functions such as recall sorting and underwriting problem judgment through the first focused question, and then obtain the target answer content. In addition, the tax calculation tool corresponding to the tax benefit-related intent can be used, combined with the judgment of whether the question is related to personal tax pension products, to obtain the target answer content through the card tool.

[0142] In one embodiment, generating target answer content corresponding to the initial question based on the target response tool includes:

[0143] Focusing the initial problem to obtain a target focused problem;

[0144] Retrieving the target focus question based on the target response tool to obtain multiple products to be recalled and ranking information of the multiple products to be recalled;

[0145] The target answer content is generated based on the multiple products to be recalled and the ranking information.

[0146] In an embodiment of the present invention, the initial question is focused to obtain a target-focused question; the target-focused question is retrieved using the target-response tool to obtain multiple products to be recalled and their ranking information; and the target answer content is generated based on the multiple products to be recalled and the ranking information. In this way, the target-focused question is obtained through question focusing, and the multiple products to be recalled and their ranking information are obtained based on the target-focused question, and the target answer content is then generated. This allows the output target answer content to include multiple products, and the multiple products include content that meets user needs, thereby improving the accuracy of the answer content.

[0147] In some embodiments, the initial question is focused to obtain a target focused question. The initial question may be first enhanced to obtain an enhanced question, and the initial question may be completed to obtain a completed question, and then the enhanced question and the completed question may be combined to obtain the target focused question.

[0148] In some embodiments, the Figures 2 to 13 The flowchart shown builds the interface of the intelligent question-answering system, enabling full-link visual analysis of the question-answering process to improve the accuracy of answers.

[0149] See Figure 15 , Figure 15 This is a structural diagram of an intelligent question-answering device provided by an embodiment of the present invention. Figure 15 As shown, the intelligent question-answering device 1500 includes:

[0150] Receiving module 1501, configured to receive an initial question input by a target user and obtain historical data of the target user;

[0151] Identification module 1502, configured to perform intent identification on the historical data and the initial question to obtain a target intent corresponding to the initial question;

[0152] Retrieval module 1503, configured to perform retrieval based on the target intent and determine a target response tool corresponding to the target intent;

[0153] The output module 1504 is configured to generate target answer content corresponding to the initial question based on the target reply tool and output the target answer content.

[0154] In one embodiment, the retrieval module 1503 includes:

[0155] A retrieval submodule, configured to search a preset mapping table based on the target intent and determine the target response tool corresponding to the target intent;

[0156] Among them, the preset mapping table includes multiple intentions and the identification of the response tool corresponding to each intention, and the multiple intentions include at least one of unrecognizable intentions, product-related intentions, professional knowledge-related intentions, tax benefit-related intentions, product comparison intentions, product recommendation intentions, product planning intentions, product knowledge intentions, company knowledge intentions, common sense intentions, question-and-answer system knowledge intentions, product catalog knowledge intentions, and product channel knowledge intentions.

[0157] In one embodiment, when the target intent is the professional knowledge-related intent, the product knowledge intent, the company knowledge intent, the common sense intent, the question-answering system knowledge intent, or the product catalog knowledge intent, the target response tool is a knowledge retrieval tool, and the output module 1504 includes:

[0158] A first determination submodule is used to determine the professional knowledge type corresponding to the target intention;

[0159] A first generating submodule is configured to generate the target answer content based on the component corresponding to the professional knowledge type in the knowledge retrieval tool;

[0160] Among them, the professional knowledge types are general professional knowledge types for commercial maintenance, general professional knowledge types for traditional insurance, general professional knowledge types for individual maintenance, general professional knowledge types for dividend insurance, professional knowledge question and answer types for commercial maintenance, business development knowledge question and answer types for traditional insurance, professional knowledge question and answer types for individual maintenance or professional knowledge question and answer types for dividend insurance.

[0161] In one embodiment, when the target intent is product knowledge intent, the first generating submodule includes:

[0162] A first determining unit, configured to determine a first product corresponding to the target intent;

[0163] A first completion unit, configured to complete the initial question based on the historical data using a large model in the knowledge retrieval tool to obtain a first completed question;

[0164] A first enhancement unit, configured to enhance the initial question based on the database corresponding to the first product in the knowledge retrieval tool and the historical data to obtain a first enhanced question;

[0165] a first combining unit, configured to combine the first completion question and the first enhancement question to obtain a first focus question;

[0166] The first generating unit is configured to generate the target answer content corresponding to the first focused question based on the answer knowledge base corresponding to the professional knowledge type in the knowledge retrieval tool.

[0167] In one embodiment, when the target intent is the company knowledge intent, the common sense intent, the question-answering system knowledge intent, or the product catalog knowledge intent, the first generation submodule includes:

[0168] A second completion unit is configured to complete the initial question based on the historical data using the large model in the knowledge retrieval tool to obtain a second completed question;

[0169] The second generating unit is configured to generate the target answer content corresponding to the second completion question based on the answer knowledge base corresponding to the professional knowledge type.

[0170] In one embodiment, when the target intent is the company knowledge intent or the common sense intent, the second generating unit includes:

[0171] A first completion subunit is configured to generate an intermediate answer content corresponding to the second completion question based on an answer knowledge base corresponding to the professional knowledge type;

[0172] A judgment subunit, configured to judge whether the intermediate answer content satisfies the requirements for answering the initial question through the large model;

[0173] a first setting unit, configured to set the intermediate answer content as the target answer content if the intermediate answer content satisfies the requirement of answering the initial question;

[0174] The second completion sub-unit is used to complete the intermediate answer content through the large model to obtain the target answer content when the intermediate answer content does not meet the requirements of answering the initial question.

[0175] In one embodiment, when the target intent is the product channel knowledge intent, the first generating submodule includes:

[0176] an identification unit, configured to identify the initial question using the large model in the knowledge retrieval tool to obtain a product channel type corresponding to the initial question;

[0177] a third generating unit, configured to generate the target answer content corresponding to the initial question based on the answer knowledge base corresponding to the professional knowledge type in the knowledge retrieval tool when the product channel type is a channel operation type;

[0178] a second determining unit, configured to determine a second product corresponding to the target intent when the product channel type is a sales region type;

[0179] A third completion unit is configured to complete the initial question based on the historical data using the large model in the knowledge retrieval tool to obtain a third completed question;

[0180] A second enhancement unit, configured to enhance the initial question based on the database corresponding to the second product in the knowledge retrieval tool and the historical data to obtain a second enhanced question;

[0181] a second combining unit, configured to combine the third completion question and the second enhancement question to obtain a second focus question;

[0182] A fourth generating unit is configured to generate the target answer content corresponding to the second focused question based on the answer knowledge base corresponding to the professional knowledge type in the knowledge retrieval tool.

[0183] In one embodiment, when the target intent is the tax benefit-related intent, the target response tool is a knowledge search tool and a product card tool, and the output module 1504 includes:

[0184] A second determination submodule is used to determine the target type corresponding to the target intention;

[0185] A third determination submodule is configured to determine at least one product corresponding to the initial question based on the large model in the knowledge retrieval tool;

[0186] a second generating submodule, configured to generate the target answer content based on the target type in the product card tool and the component corresponding to the at least one product;

[0187] The target type is a tax type or a revenue type.

[0188] In one embodiment, when the target intent is the unrecognizable intent, the target response tool is an intent clarification tool, and the output module 1504 includes the following:

[0189] A first analysis submodule is configured to analyze the initial question based on the large model in the intent clarification tool to obtain a first answer, where the first answer is used to request the user to provide other descriptive information;

[0190] The acquisition submodule is used to obtain a second answer from the preset response in the intention clarification tool, and the second answer is used to display customer service contact information.

[0191] In one embodiment, when the target intention is the product comparison intention, the product recommendation intention, or the product planning intention, the target response tool is a product card tool, and the output module 1504 includes:

[0192] A fourth determination submodule, configured to determine a third product corresponding to the target intent;

[0193] A second analysis submodule is configured to analyze the product corresponding to the third product based on a preset component in the product card tool to obtain the target answer content;

[0194] Among them, the preset components include at least one of a product comparison component, a product recommendation component, a product planning component and a product preset speech component.

[0195] In one embodiment, the output module 1504 includes:

[0196] A focusing submodule, configured to focus the initial problem to obtain a target focused problem;

[0197] a processing submodule, configured to retrieve the target focus question based on the target response tool, and obtain a plurality of products to be recalled and ranking information of the plurality of products to be recalled;

[0198] The third generating submodule is configured to generate the target answer content based on the multiple products to be recalled and the ranking information.

[0199] The intelligent question-answering device provided in the embodiment of the present invention is capable of implementing each process of each embodiment of the above-mentioned intelligent question-answering method. The technical features correspond one to one and can achieve the same technical effect. To avoid repetition, they will not be described here.

[0200] It should be noted that the intelligent question-answering device in the embodiment of the present invention may be a device, or a component, integrated circuit, or chip in an electronic device.

[0201] The embodiment of the present invention further provides an electronic device, see Figure 16 , Figure 16 This is a schematic diagram of the structure of an electronic device provided by the present invention. The electronic device includes a memory 1601, a processor 1602, and a program or instruction stored in the memory 1601 and running on the memory 1601. When the program or instruction is executed by the processor 1602, Figure 1 Any steps in the corresponding intelligent question-answering method embodiment and the same beneficial effects are achieved will not be repeated here.

[0202] The processor 1602 may be a CPU, an ASIC, an FPGA, or a GPU.

[0203] Those skilled in the art will appreciate that all or part of the steps of implementing the intelligent question-answering method of the above embodiment can be completed based on hardware related to program instructions, and the program can be stored in a readable medium.

[0204] The embodiment of the present invention further provides a readable storage medium, which stores a computer program. When the computer program is executed by a processor, the above Figure 1 Any steps in the corresponding intelligent question-answering method embodiment can achieve the same technical effect and will not be described here to avoid repetition. The storage medium is, for example, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk.

[0205] The present invention also provides a computer program product, comprising computer instructions, which, when executed by a processor, implement the above Figure 1 The various processes of the corresponding intelligent question-answering method embodiments can achieve the same technical effects, so they will not be repeated here to avoid repetition.

[0206] The terms "first", "second" etc. in the embodiments of the present invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequential order. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions, for example, the process, method, system, product or equipment comprising a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or that are inherent to these processes, methods, products or equipment. In addition, "and / or" is used in this application to represent at least one of the connected objects, for example A and / or B and / or C, which means comprising 7 situations including single A, single B, single C, and both A and B exist, both B and C exist, both A and C exist, and both A, B and C exist.

[0207] It should be noted that, in this document, the terms "comprises," "includes," or any other variations thereof are intended to encompass non-exclusive inclusion, such that a process, method, article, or apparatus comprising a series of elements includes not only those elements but also other elements not explicitly listed, or elements inherent to such process, method, article, or apparatus. In the absence of further limitations, an element defined by the phrase "comprising a ..." does not exclude the presence of other identical elements in the process, method, article, or apparatus comprising the element.

[0208] Based on 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 the necessary general hardware platform, and of course can also be based on hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of the present application, 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, magnetic disk, optical disk), and includes a number of instructions for enabling a terminal (which can be a mobile phone, computer, server, air conditioner, or second terminal device, etc.) to execute the methods of each embodiment of the present application.

[0209] The embodiments of the present application are described above in conjunction with the accompanying drawings, but the present application is not limited to the above-mentioned specific implementation methods. The above-mentioned specific implementation methods are merely illustrative and not restrictive. Under the guidance of this application, ordinary technicians in this field can also make many forms without departing from the purpose of this application and the scope of protection of the claims, all of which are within the protection of this application.

Claims

1. An intelligent question-answering method, characterized in that: include: receiving an initial question input by a target user and obtaining historical data of the target user; Performing intent recognition on the historical data and the initial question to obtain a target intent corresponding to the initial question; Performing a search based on the target intent to determine a target response tool corresponding to the target intent; Generate target answer content corresponding to the initial question based on the target reply tool, and output the target answer content.

2. The method according to claim 1, wherein The searching based on the target intent and determining the target response tool corresponding to the target intent includes: Searching a preset mapping table based on the target intent to determine the target response tool corresponding to the target intent; Among them, the preset mapping table includes multiple intentions and the identification of the response tool corresponding to each intention, and the multiple intentions include at least one of unrecognizable intentions, product-related intentions, professional knowledge-related intentions, tax benefit-related intentions, product comparison intentions, product recommendation intentions, product planning intentions, product knowledge intentions, company knowledge intentions, common sense intentions, question-and-answer system knowledge intentions, product catalog knowledge intentions, and product channel knowledge intentions.

3. The method according to claim 2, wherein In a case where the target intent is the professional knowledge-related intent, the product knowledge intent, the company knowledge intent, the common sense intent, the question-answering system knowledge intent, or the product catalog knowledge intent, the target response tool is a knowledge retrieval tool, and generating target answer content corresponding to the initial question based on the target response tool includes: Determine the type of expertise corresponding to the target intent; Generate the target answer content based on the component corresponding to the professional knowledge type in the knowledge retrieval tool; Among them, the professional knowledge types are general professional knowledge types for commercial maintenance, general professional knowledge types for traditional insurance, general professional knowledge types for individual maintenance, general professional knowledge types for dividend insurance, professional knowledge question and answer types for commercial maintenance, business development knowledge question and answer types for traditional insurance, professional knowledge question and answer types for individual maintenance or professional knowledge question and answer types for dividend insurance.

4. The method according to claim 3, wherein In the case where the target intent is product knowledge intent, generating the target answer content based on the component corresponding to the professional knowledge type in the knowledge retrieval tool includes: Determining a first product corresponding to the target intent; Completing the initial question based on the historical data using the large model in the knowledge retrieval tool to obtain a first completed question; enhancing the initial question based on the database corresponding to the first product in the knowledge retrieval tool and the historical data to obtain a first enhanced question; Combining the first completion question and the first enhancement question to obtain a first focusing question; The target answer content corresponding to the first focused question is generated based on the answer knowledge base corresponding to the professional knowledge type in the knowledge retrieval tool.

5. The method according to claim 3, wherein In a case where the target intent is the company knowledge intent, the common sense intent, the question-answering system knowledge intent, or the product catalog knowledge intent, generating the target answer content based on the component corresponding to the professional knowledge type in the knowledge retrieval tool includes: Completing the initial question based on the historical data using the large model in the knowledge retrieval tool to obtain a second completed question; Based on the answer knowledge base corresponding to the professional knowledge type, the target answer content corresponding to the second completion question is generated.

6. The method according to claim 5, wherein In a case where the target intent is the company knowledge intent or the common sense intent, generating the target answer content corresponding to the second completion question based on the answer knowledge base corresponding to the professional knowledge type includes: generating an intermediate answer content corresponding to the second completion question based on the answer knowledge base corresponding to the professional knowledge type; Determining whether the intermediate answer content satisfies the initial question by using the large model; If the intermediate answer content satisfies the requirement of answering the initial question, setting the intermediate answer content as the target answer content; In the case that the intermediate answer content does not satisfy the requirements for answering the initial question, the intermediate answer content is supplemented by the large model to obtain the target answer content.

7. The method according to claim 3, wherein In the case where the target intent is the product channel knowledge intent, generating the target answer content based on the component corresponding to the professional knowledge type in the knowledge retrieval tool includes: Identify the initial question using the large model in the knowledge retrieval tool to obtain the product channel type corresponding to the initial question; In a case where the product channel type is a channel operation type, generating the target answer content corresponding to the initial question based on the answer knowledge base corresponding to the professional knowledge type in the knowledge retrieval tool; In a case where the product channel type is a sales region type, determining a second product corresponding to the target intent; Completing the initial question based on the historical data using the large model in the knowledge retrieval tool to obtain a third completed question; enhancing the initial question based on the database corresponding to the second product in the knowledge retrieval tool and the historical data to obtain a second enhanced question; Combining the third completion question and the second enhancement question to obtain a second focusing question; The target answer content corresponding to the second focused question is generated based on the answer knowledge base corresponding to the professional knowledge type in the knowledge retrieval tool.

8. The method according to claim 2, wherein In the case where the target intention is the tax benefit-related intention, the target response tool is a knowledge retrieval tool and a product card tool, and the target answer content corresponding to the initial question generated based on the target response tool includes: Determining a target type corresponding to the target intent; Determining at least one product corresponding to the initial question based on the large model in the knowledge retrieval tool; generating the target answer content based on the target type and the components corresponding to the at least one product in the product card tool; The target type is a tax type or a revenue type.

9. The method according to claim 2, wherein In the case where the target intent is the unrecognizable intent, the target response tool is an intent clarification tool, and generating target answer content corresponding to the initial question based on the target response tool includes the following: Analyze the initial question based on the large model in the intent clarification tool to obtain a first answer, where the first answer is used to request the user to provide additional descriptive information; A second answer is obtained from a preset response in the intent clarification tool, where the second answer is used to display customer service contact information.

10. The method according to claim 2, wherein In the case where the target intent is the product-related intent, the target response tool is a product card tool, and generating target answer content corresponding to the initial question based on the target response tool includes: determining a third product corresponding to the target intent; Analyzing the product corresponding to the third product based on the preset components in the product card tool to obtain the target answer content; Among them, the preset components include at least one of a product comparison component, a product recommendation component, a product planning component and a product preset speech component.

11. The method according to claim 2, wherein Generating target answer content corresponding to the initial question based on the target response tool includes: Focusing the initial problem to obtain a target focused problem; Retrieving the target focus question based on the target response tool to obtain multiple products to be recalled and ranking information of the multiple products to be recalled; The target answer content is generated based on the multiple products to be recalled and the ranking information.

12. An intelligent question-answering device, characterized in that: include: A receiving module, configured to receive an initial question input by a target user and obtain historical data of the target user; an identification module, configured to perform intent identification on the historical data and the initial question to obtain a target intent corresponding to the initial question; A retrieval module, configured to perform retrieval based on the target intent and determine a target response tool corresponding to the target intent; An output module is used to generate target answer content corresponding to the initial question based on the target reply tool and output the target answer content.

13. An electronic device, characterized in that: include: A processor, a memory, and a program stored in the memory and executable on the processor, wherein the program, when executed by the processor, implements the steps of the intelligent question-answering method according to any one of claims 1 to 11.

14. A computer-readable storage medium, characterized in that The computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps of the intelligent question-answering method according to any one of claims 1 to 11.

15. A computer program product, characterized in that The method comprises computer instructions, which, when executed by a processor, implement the steps of the intelligent question-answering method according to any one of claims 1 to 11.

Images