Data processing method and device, electronic equipment and computer program product

Abstract

The invention discloses a data processing method and device, electronic equipment and a computer program product. The method comprises the steps of obtaining target problem data described by using a natural language; semantic analysis is conducted on the target problem data and the description data corresponding to the multiple preset intelligent agents, a candidate intelligent agent set matched with the target problem data is obtained, and candidate intelligent agents in the candidate intelligent agent set are the preset intelligent agents with the semantic similarity between the description data and the target problem data meeting a preset condition; in the candidate agent set, selecting a target agent with the maximum semantic similarity with the target problem data; and replying the target question data by using the target agent to generate target reply data. According to the method and the device, the technical problem of poor reply accuracy caused by incapability of intelligently selecting a proper agent for a question needing to be replied to reply in the prior art is solved.

Description

Technical Field

[0001] The present invention relates to the field of data processing, and in particular to a data processing method, device, electronic device and computer program product. Background Art

[0002] In recent years, with the development of natural language processing and artificial intelligence technologies, question-answering systems based on machine learning have gradually become a research hotspot and are widely used in customer service, search engines, intelligent assistants and other fields.

[0003] However, because existing question-answering systems typically only support fixed rules or predefined semantic models, when the semantics of a user's question are inconsistent with the system's predefined semantic models, the system cannot correctly identify the user's intent and provide an appropriate response. Furthermore, because existing question-answering systems often rely on a single agent to process user questions, if that agent cannot understand the user's question, the system will be unable to provide an answer.

[0004] Currently, no effective solution has been proposed to the problem that the above-mentioned existing technologies are unable to intelligently select appropriate agents to answer questions that need to be answered, resulting in poor answer accuracy. Summary of the Invention

[0005] Embodiments of the present invention provide a data processing method, device, electronic device, and computer program product to at least solve the technical problem that the prior art is unable to intelligently select a suitable intelligent agent to answer questions that need to be answered, resulting in poor answer accuracy.

[0006] According to one aspect of an embodiment of the present invention, a data processing method is provided, comprising: obtaining target question data described in natural language; performing semantic analysis on the target question data and the description data corresponding to a plurality of preset agents, to obtain a set of candidate agents matching the target question data, wherein the preset agent is used to reply to the preset question data, the description data is used to describe the function of the preset agent in natural language, and the candidate agents in the candidate agent set are the preset agents whose semantic similarity between the description data and the target question data meets preset conditions; selecting the target agent with the greatest semantic similarity to the target question data from the candidate agent set; and using the target agent to reply to the target question data to generate target reply data.

[0007] Optionally, semantic analysis is performed on the target question data and the descriptive data corresponding to multiple preset agents to obtain a set of candidate agents that match the target question data, including: performing semantic analysis on the preset question data to obtain question intent; performing semantic analysis on the descriptive data pre-configured for each preset agent to obtain agent intent; detecting the semantic similarity between the question intent and the intent of each agent; and determining a set of candidate agents based on the agent intent that meets the preset conditions with the semantic similarity.

[0008] Optionally, determining the candidate agent set based on the agent intention whose semantic similarity meets the preset conditions includes: determining at least one of the agent intentions whose semantic similarity meets the preset conditions as a candidate agent intention; and determining the preset agent corresponding to the candidate agent intention as the candidate agent in the candidate agent set.

[0009] Optionally, determining at least one of the agent intentions whose semantic similarity meets the preset conditions as a candidate agent intention includes: determining at least one of the agent intentions whose semantic similarity exceeds a preset similarity threshold as a candidate agent intention; or selecting a preset number of the agent intentions as candidate agent intentions in order of the semantic similarity from high to low.

[0010] Optionally, the descriptive data includes: descriptive text and keywords, and semantic analysis is performed on the descriptive data corresponding to the target problem data and multiple preset agents respectively to obtain a set of candidate agents matching the target problem data, including: semantic analysis is performed on the descriptive text corresponding to the target problem data and multiple preset agents respectively to obtain a first analysis result corresponding to each of the preset agents, wherein the first analysis result is at least used to indicate whether the preset agent matches the target problem data; the preset agent that does not match the target problem data is determined as an alternative agent; semantic analysis is performed on the keywords corresponding to the target problem data and the alternative agents to obtain a second analysis result corresponding to each of the alternative agents, wherein the second analysis result is at least used to indicate whether the alternative agent matches the target problem data; and the candidate agent in the candidate agent set is determined among the preset agents and the alternative agents that match the target problem data.

[0011] Optionally, after performing semantic analysis on the target question data and the keywords corresponding to the alternative agents to obtain the second analysis result corresponding to each of the alternative agents, the method further includes: analyzing the first analysis results and the second analysis results of all the preset agents to determine whether there is a situation where all the preset agents cannot match the target question data; if it is determined that all the preset agents cannot match the target question data, generating an alarm message, wherein the alarm message is used to indicate that multiple preset agents cannot respond to the target question data.

[0012] Optionally, after using the target agent to reply to the target question data and generate target answer data, the method further includes: receiving an agent switching instruction, wherein the agent switching instruction is triggered by a switching plug-in, the switching plug-in corresponds one-to-one to the candidate agent, and the switching plug-in and the target answer data are displayed in the same interface; in response to the agent switching instruction, updating the target agent to the candidate agent corresponding to the switching plug-in; using the updated target agent to reply to the target question data and updating the target question data.

[0013] According to another aspect of an embodiment of the present invention, a data processing device is also provided, including: an acquisition module for acquiring target question data described in natural language; a semantic analysis module for performing semantic analysis on the target question data and the description data corresponding to multiple preset agents, to obtain a set of candidate agents matching the target question data, wherein the preset agent is used to reply to the preset question data, and the description data is used to describe the function of the preset agent using natural language, and the candidate agents in the candidate agent set are the preset agents whose semantic similarity between the description data and the target question data meets preset conditions; a selection module for selecting, from the candidate agent set, a target agent with the greatest semantic similarity to the target question data; and a generation module for using the target agent to reply to the target question data and generate target reply data.

[0014] According to another aspect of an embodiment of the present invention, an electronic device is provided, including a memory and a processor, wherein the memory stores a computer program, and the processor is configured to execute the above data processing method through the computer program.

[0015] According to another aspect of an embodiment of the present invention, a computer program product is provided, comprising computer instructions, which implement the steps of the above data processing method when executed by a processor.

[0016] In an embodiment of the present invention, semantic analysis is performed on the question data and the description data of the agent, which can identify the key information of the question and the functional scope of the agent. This analysis not only includes the surface text of the question, but also deeply understands the background, context and potential needs of the question. It can quickly screen out the most relevant set of candidate agents from a large number of preset agents, and then use the candidate agent with the greatest semantic similarity as the target agent to reply to the target question data, thereby achieving the purpose of accurately matching the questions that need to be answered and the agents required to answer the questions. The agent obtained by the precise match is used to reply, achieving the technical effect of improving the accuracy of the answer to the question, and solving the technical problem that the existing technology cannot intelligently select a suitable agent to reply to the questions that need to be answered, resulting in poor answer accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] The drawings described herein are used to provide a further understanding of the present invention and constitute a part of this application. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention. In the drawings:

[0018] Figure 1 is a flow chart of a data processing method according to an embodiment of the present invention;

[0019] Figure 2 is a schematic diagram of a process of a multi-scenario intention recognition mechanism based on semantic understanding and keyword matching according to an embodiment of the present invention;

[0020] Figure 3 is a schematic diagram of a description text pre-configured by a preset intelligent agent according to an embodiment of the present invention;

[0021] Figure 4 is a schematic diagram of keywords pre-configured by a preset agent according to an embodiment of the present invention;

[0022] Figure 5 is a schematic diagram showing target reply data and switching plug-ins according to an embodiment of the present invention;

[0023] Figure 6 is a schematic diagram of generating alarm information according to an embodiment of the present invention;

[0024] Figure 7 is a schematic diagram of a data processing device according to an embodiment of the present invention;

[0025] Figure 8 It is a structural block diagram of a computer terminal according to an embodiment of the present invention. DETAILED DESCRIPTION

[0026] In order to enable those skilled in the art to better understand the solutions of the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the embodiments described are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts should fall within the scope of protection of the present invention.

[0027] It should be noted that the terms "first", "second", etc. in the description and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that the numbers used in this way can be interchanged where appropriate, so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions. For example, a process, method, system, product or device that includes 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 inherent to these processes, methods, products or devices.

[0028] According to an embodiment of the present invention, an embodiment of a data processing 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 shown here.

[0029] Figure 1 is a flow chart of a data processing method according to an embodiment of the present invention. Figure 1 As shown, the method includes the following steps:

[0030] Step S102, obtaining target question data described in natural language;

[0031] Step S104: semantically analyze the target question data and the description data corresponding to each of the plurality of preset agents to obtain a set of candidate agents that match the target question data. The preset agents are used to answer the preset question data, and the description data are used to describe the functions of the preset agents in natural language. Candidate agents in the candidate agent set are preset agents whose semantic similarity between the description data and the target question data meets a preset condition.

[0032] Step S106: Select the target agent with the greatest semantic similarity to the target question data from the candidate agent set;

[0033] Step S108: Use the target agent to respond to the target question data and generate target response data.

[0034] In an embodiment of the present invention, semantic analysis is performed on the question data and the description data of the agent, which can identify the key information of the question and the functional scope of the agent. This analysis not only includes the surface text of the question, but also deeply understands the background, context and potential needs of the question. It can quickly screen out the most relevant set of candidate agents from a large number of preset agents, and then use the candidate agent with the greatest semantic similarity as the target agent to reply to the target question data, thereby achieving the purpose of accurately matching the questions that need to be answered and the agents required to answer the questions. The agent obtained by the precise match is used to reply, achieving the technical effect of improving the accuracy of the answer to the question, and solving the technical problem that the existing technology cannot intelligently select a suitable agent to reply to the questions that need to be answered, resulting in poor answer accuracy.

[0035] In the above step S104, semantic analysis includes but is not limited to word embedding models and semantic similarity calculation algorithms, which can understand the deep meaning of the question and the agent description.

[0036] In the above step S104, the preset conditions include at least: the semantic similarity exceeds the preset similarity threshold, or the ranking of the semantic similarity in descending order exceeds the preset ranking, wherein the preset ranking may require the numerical values ​​of the selected candidate agents to be the same.

[0037] In the above step S108, the target agent is used to reply to the target question data. After the target reply data is generated, the target reply data can be displayed in an interface capable of user interaction. In addition to the target reply data, the interface also includes a switching plug-in for each candidate agent. By triggering the switching plug-in, the target agent can be switched to the candidate agent corresponding to the switching plug-in, thereby realizing the update of the target agent. The updated target agent's reply to the target question data will serve as the new target reply number, and replace the target reply data originally displayed in the interface, thereby realizing the update of the target reply data.

[0038] As an optional embodiment, semantic analysis is performed on the target question data and the descriptive data corresponding to multiple preset agents to obtain a set of candidate agents that match the target question data, including: performing semantic analysis on the preset question data to obtain the question intent; performing semantic analysis on the pre-configured descriptive data of each preset agent to obtain the agent intent; detecting the semantic similarity between the question intent and the intent of each agent; and determining the candidate agent set based on the agent intent whose semantic similarity meets the preset conditions.

[0039] In the above-mentioned embodiment of the present application, through semantic analysis technology, the system can understand the essential intention of the question and the questions that the preset intelligent agent can answer. By evaluating the semantic similarity between the intelligent agent intention and the question intention, the most suitable set of candidate intelligent agents can be screened out from the numerous preset intelligent agents, which improves the pertinence and efficiency of problem solving, ensures the accuracy of the matching process, avoids invalid answers caused by misunderstanding the problem, and improves the accuracy of intelligent agent selection.

[0040] Optionally, determining the set of candidate agents based on the agent intentions whose semantic similarity meets preset conditions can be achieved through an intention matching algorithm, wherein the intention matching algorithm includes but is not limited to intention-based semantic similarity calculation and intention recognition model, which can accurately evaluate the degree of matching between the problem intention and the agent intention, thereby ensuring the accuracy and efficiency of the agent selection.

[0041] As an optional embodiment, determining the candidate agent set based on the agent intentions whose semantic similarity meets the preset conditions includes: determining at least one agent intention whose semantic similarity meets the preset conditions as a candidate agent intention; and determining the preset agent corresponding to the candidate agent intention as a candidate agent in the candidate agent set.

[0042] In the above-mentioned embodiment of the present application, the agent intent is the analysis result of the descriptive data of each preset agent. Therefore, the agent intent corresponds to the preset agent one-to-one. Then, when selecting candidate agents from the preset agents, the semantic similarity between the problem intent corresponding to the target problem data and the intention of each agent can be determined, and multiple agent intentions can be preliminarily screened based on the semantic similarity to obtain at least one candidate agent intention that meets the preset conditions. The preset agent corresponding to each candidate agent intention is the candidate agent in the candidate agent set, which ensures the rationality and effectiveness of the candidate agent set, realizes the effective management and optimization of the candidate agent set, and solves the problems of low processing efficiency and waste of resources caused by too many agents.

[0043] As an optional embodiment, determining at least one intelligent agent intention whose semantic similarity meets preset conditions as a candidate intelligent agent intention includes: determining at least one intelligent agent intention whose semantic similarity exceeds a preset similarity threshold as a candidate intelligent agent intention; or selecting a preset number of intelligent agent intentions as candidate intelligent agent intentions in order of semantic similarity from high to low.

[0044] The above-mentioned embodiments of the present application extract candidate agent intentions with higher semantic similarity by setting a preset similarity threshold or according to a preset number, thereby completing the screening of candidate agents by screening the candidate agent intentions, realizing refined management and optimization of the candidate agent set, and solving the problems of low response efficiency and low answer quality caused by unclear standards in the agent selection process.

[0045] As an optional embodiment, the description data includes: description text and keywords, and semantic analysis is performed on the description data corresponding to the target problem data and multiple preset agents respectively to obtain a set of candidate agents that match the target problem data, including: semantic analysis of the description text corresponding to the target problem data and multiple preset agents respectively to obtain a first analysis result corresponding to each preset agent, wherein the first analysis result is at least used to indicate whether the preset agent matches the target problem data; the preset agent that does not match the target problem data is determined as an alternative agent; semantic analysis is performed on the keywords corresponding to the target problem data and the alternative agent to obtain a second analysis result corresponding to each alternative agent, wherein the second analysis result is at least used to indicate whether the alternative agent matches the target problem data; and among the preset agents and alternative agents that match the target problem data, the candidate agents in the candidate agent set are determined.

[0046] The above-mentioned embodiments of the present application, by adopting a multi-level semantic analysis method, can more comprehensively and accurately evaluate the matching degree between the preset intelligent agent and the target question data, thereby solving the problems of inaccurate intelligent agent matching and low answer quality caused by a single analysis method.

[0047] Specifically, we will describe the combination of text analysis technology (including but not limited to deep learning-based text understanding models and text similarity calculation based on semantic vectors) and keyword analysis technology (including but not limited to keyword-based semantic similarity calculation and keyword recognition algorithms) to ensure the comprehensiveness and accuracy of agent selection, improve problem-solving efficiency and user experience. At the same time, this mechanism can also promote the continuous optimization and upgrading of agents. Through user feedback and system learning, the accuracy of semantic analysis and the ability to recognize agent intent are continuously improved, thereby enhancing the intelligence and adaptability of the entire system.

[0048] As an optional embodiment, after performing semantic analysis on the keywords corresponding to the target question data and the alternative intelligent agents to obtain the second analysis result corresponding to each alternative intelligent agent, the method also includes: analyzing the first analysis results and the second analysis results of all preset intelligent agents to determine whether there is a situation where all preset intelligent agents cannot match the target question data; when it is determined that all preset intelligent agents cannot match the target question data, generating an alarm message, wherein the alarm message is used to indicate that multiple preset intelligent agents cannot respond to the target question data.

[0049] In the above-mentioned embodiment of the present application, when all preset intelligent agents cannot be matched, an alarm message will be generated to promptly remind the administrator or developer, thereby promoting the continuous optimization and upgrading of the system. By introducing an alarm mechanism, the problem of insufficient intelligent agent functions can be discovered and solved in a timely manner, thereby improving the stability and reliability of the system.

[0050] As an optional embodiment, after using the target agent to reply to the target question data and generate the target reply data, the method also includes: receiving an agent switching instruction, wherein the agent switching instruction is triggered by a switching plug-in, the switching plug-in corresponds one-to-one to the candidate agent, and the switching plug-in and the target reply data are displayed in the same interface; in response to the agent switching instruction, the target agent is updated to the candidate agent corresponding to the switching plug-in; and the updated target agent is used to reply to the target question data and update the target question data.

[0051] In the above-mentioned embodiment of the present application, the introduction of the switching plug-in enables the user to quickly try the solutions of other intelligent agents when the user is not satisfied with the answer of the current intelligent agent, thereby enhancing the flexibility of the system and the user experience.

[0052] Optionally, after using the target agent to answer the target question data, the target answer data and the switching plug-in can be displayed in the same interface, wherein there can be multiple switching plug-ins displayed in the interface, and each switching plug-in corresponds to a candidate agent; if the user is not satisfied with the target answer data obtained using the current target agent, the agent switching instruction can be triggered through the switching plug-in, and the candidate agent corresponding to the switching plug-in is called to realize the update of the target agent.

[0053] Optionally, after the target agent is updated by the candidate agent, the target agent before the update will replace the candidate agent updated to the target agent, such as matching the target agent before the update with the switching plug-in corresponding to the candidate agent updated to the target agent; or using the target agent before the update as the new candidate agent and reallocating the switching plug-in.

[0054] The present invention also provides a preferred embodiment, which provides a multi-scenario intent recognition mechanism based on semantic understanding and keyword matching. This mechanism uses an intent recognition agent to serve multiple scenarios. It can switch between different agents for multiple scenarios and different scenario questions, and provide multiple candidate agents to serve different customer questions. If the intent recognition agent identifies an error in the agent serving different businesses, it can manually switch to the candidate agent to continue answering the same customer question, thereby providing better service to the customer.

[0055] Figure 2 is a schematic diagram of a process of a multi-scenario intention recognition mechanism based on semantic understanding and keyword matching according to an embodiment of the present invention, such as Figure 2 As shown, by creating a global agent and linking multiple sub-agents (i.e., preset agents), multiple application scenarios are included and covered. The ability of the large language model is used to detect users' questions and usage scenarios through the intent recognition of different sub-agents, and different questions are automatically distributed to different sub-agents to serve different scenarios. Specifically, the global agent of intent recognition can be regarded as the front desk of the company, and users only need to interact and communicate with this agent. This agent will then distribute tasks to other sub-agents, and the sub-agents will respond to users' questions. The entire process is as follows: The global agent and each sub-agent are equipped with settings for the agent's prompt words, descriptions, and keywords for the scenarios and problems it handles.

[0056] Figure 3 is a schematic diagram of a description text pre-configured by a preset agent according to an embodiment of the present invention, such as Figure 3 As shown, when a user interacts with the global agent of intent recognition, the agent will first perform semantic retrieval and judgment on all set and bound sub-agent descriptions (i.e., the description text pre-configured by the preset agent), thereby determining the candidate agent based on the first analysis result.

[0057] Figure 4 is a schematic diagram of keywords pre-configured by a preset agent according to an embodiment of the present invention, such as Figure 4 As shown, when a user interacts with the global agent of intent recognition, the agent can first perform semantic retrieval and judgment on the pre-configured keywords of all sub-agents (i.e., preset agents), thereby determining the candidate agents based on the second analysis results.

[0058] Figure 5 is a schematic diagram showing target reply data and switching plug-ins according to an embodiment of the present invention, such as Figure 5As shown, when the large language model recognizes the intent of user questions and can detect relevant agents through the description of sub-agents (that is, determine the candidate agents based on the first analysis result), or when the large language model cannot detect relevant agents, but can perform semantic retrieval on the set keywords through semantic search (that is, determine the candidate agents based on the second analysis result), the intent recognition mechanism will sort the sub-agents (that is, preset agents) according to the relevance of semantics, select the top 3 most relevant sub-agents (that is, candidate agents), and use the top 1 agent (that is, the target agent) to give answers to customer questions and application scenarios. Top 2 and Top 3 are used as candidate agents for customers to manually switch.

[0059] Figure 6 is a schematic diagram of generating alarm information according to an embodiment of the present invention, such as Figure 6 As shown, if no candidate agent can be found through the description text and keywords, the user will not be given an answer, and the user will be replied that there is no agent that can handle the relevant problem (i.e., an alarm message) to reduce the illusion of the agent.

[0060] The above-mentioned embodiments of the present application achieve rapid understanding of complex problems and accurate matching of intelligent agents through deep learning and natural language processing technologies, and solve the problems of low matching efficiency and inaccurate answers in traditional question-answering systems due to the inability to accurately understand the intention of the question.

[0061] According to an embodiment of the present invention, a data processing device embodiment is also provided. It should be noted that the data processing device can be used to execute the data processing method in the embodiment of the present invention, and the data processing method in the embodiment of the present invention can be executed in the data processing device.

[0062] Figure 7 is a schematic diagram of a data processing device according to an embodiment of the present invention. Figure 7 As shown, the device may include: an acquisition module 72 for acquiring target question data described in natural language; a semantic analysis module 74 for performing semantic analysis on the target question data and the description data corresponding to multiple preset agents, to obtain a set of candidate agents that match the target question data, wherein the preset agent is used to reply to the preset question data, and the description data is used to describe the function of the preset agent using natural language, and the candidate agents in the candidate agent set are preset agents whose semantic similarity between the description data and the target question data meets preset conditions; a selection module 76 for selecting a target agent with the greatest semantic similarity to the target question data in the candidate agent set; a generation module 78 for using the target agent to reply to the target question data and generate target reply data.

[0063] It should be noted that the acquisition module 72 in this embodiment can be used to execute step S102 in the embodiment of the present application, the semantic analysis module 74 in this embodiment can be used to execute step S104 in the embodiment of the present application, the selection module 76 in this embodiment can be used to execute step S106 in the embodiment of the present application, and the generation module 78 in this embodiment can be used to execute step S108 in the embodiment of the present application. The examples and application scenarios implemented by the above modules and corresponding steps are the same, but are not limited to the contents disclosed in the above embodiments.

[0064] In an embodiment of the present invention, semantic analysis is performed on the question data and the description data of the agent, which can identify the key information of the question and the functional scope of the agent. This analysis not only includes the surface text of the question, but also deeply understands the background, context and potential needs of the question. It can quickly screen out the most relevant set of candidate agents from a large number of preset agents, and then use the candidate agent with the greatest semantic similarity as the target agent to reply to the target question data, thereby achieving the purpose of accurately matching the questions that need to be answered and the agents required to answer the questions. The agent obtained by the precise match is used to reply, achieving the technical effect of improving the accuracy of the answer to the question, and solving the technical problem that the existing technology cannot intelligently select a suitable agent to reply to the questions that need to be answered, resulting in poor answer accuracy.

[0065] As an optional embodiment, the semantic analysis module includes: a first semantic analysis unit, used to perform semantic analysis on preset question data to obtain question intent; a second semantic analysis unit, used to perform semantic analysis on the pre-configured description data of each preset agent to obtain agent intent; a detection unit, used to detect the semantic similarity between the question intent and the intent of each agent; a first determination unit, used to determine a set of candidate agents based on the agent intent whose semantic similarity meets preset conditions.

[0066] As an optional embodiment, the first determination unit includes: a first determination subunit, used to determine at least one agent intention whose semantic similarity meets preset conditions as a candidate agent intention; a second determination subunit, used to determine the preset agent corresponding to the candidate agent intention as a candidate agent in the candidate agent set.

[0067] As an optional embodiment, the first determination subunit includes: a third determination subunit, used to determine at least one intelligent agent intention whose semantic similarity exceeds a preset similarity threshold as a candidate intelligent agent intention; or a selection subunit, used to select a preset number of intelligent agent intentions as candidate intelligent agent intentions in order of semantic similarity from high to low.

[0068] As an optional embodiment, the description data includes: description text and keywords, and the semantic analysis module includes: a third semantic analysis unit, which performs semantic analysis on the description texts corresponding to the target question data and multiple preset agents, and obtains a first analysis result corresponding to each preset agent, wherein the first analysis result is at least used to indicate whether the preset agent matches the target question data; a second determination unit, which is used to determine the preset agent that does not match the target question data as an alternative agent; a fourth semantic analysis unit, which is used to perform semantic analysis on the keywords corresponding to the target question data and the alternative agent, and obtains a second analysis result corresponding to each alternative agent, wherein the second analysis result is at least used to indicate whether the alternative agent matches the target question data; a fifth determination subunit, which is used to determine the candidate agent in the candidate agent set among the preset agents and alternative agents that match the target question data.

[0069] As an optional embodiment, the device also includes: a sixth determination subunit, which is used to perform semantic analysis on the target question data and the keywords corresponding to the alternative intelligent agents, and after obtaining the second analysis result corresponding to each alternative intelligent agent, analyze the first analysis results and the second analysis results of all preset intelligent agents to determine whether there is a situation where all preset intelligent agents cannot match the target question data; a seventh determination subunit, which is used to generate an alarm message when it is determined that all preset intelligent agents cannot match the target question data, wherein the alarm message is used to indicate that multiple preset intelligent agents cannot respond to the target question data.

[0070] As an optional embodiment, the device also includes: a receiving sub-module, which is used to receive an agent switching instruction after using the target agent to reply to the target question data and generate target reply data, wherein the agent switching instruction is triggered by a switching plug-in, the switching plug-in corresponds one-to-one to the candidate agent, and the switching plug-in and the target reply data are displayed in the same interface; a first updating sub-module, which is used to update the target agent to the candidate agent corresponding to the switching plug-in in response to the agent switching instruction; and a second updating sub-module, which is used to use the updated target agent to reply to the target question data and update the target question data.

[0071] An embodiment of the present invention may provide an electronic device, which may be a computer terminal, and the computer terminal may be any computer terminal device in a computer terminal group. Optionally, in this embodiment, the computer terminal may also be replaced by a terminal device such as a mobile terminal.

[0072] Optionally, in this embodiment, the computer terminal may be located in at least one network device among a plurality of network devices of a computer network.

[0073] In this embodiment, the above-mentioned computer terminal can execute the program code of the following steps in the data processing method: obtaining target question data described using natural language; performing semantic analysis on the target question data and the description data corresponding to multiple preset agents, and obtaining a set of candidate agents that match the target question data, wherein the preset agent is used to reply to the preset question data, and the description data is used to describe the function of the preset agent using natural language, and the candidate agents in the candidate agent set are preset agents whose semantic similarity between the description data and the target question data meets the preset conditions; in the candidate agent set, selecting the target agent with the greatest semantic similarity with the target question data; using the target agent to reply to the target question data to generate target reply data.

[0074] Figure 8 is a structural block diagram of a computer terminal according to an embodiment of the present invention, such as Figure 8 As shown, the computer terminal 80 may include: one or more (only one is shown in the figure) processors 82 and a memory 84.

[0075] The memory can be used to store software programs and modules, such as program instructions / modules corresponding to the data processing method and device in the embodiments of the present invention. The processor executes various functional applications and data processing by running the software programs and modules stored in the memory, that is, implementing the above-mentioned data processing method. The memory may include a high-speed random access memory and may also include a non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some instances, the memory may further include a memory remotely located relative to the processor, and these remote memories may be connected to the terminal 80 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.

[0076] The processor can call the information and application programs stored in the memory through the transmission device to perform the following steps: obtaining target question data described in natural language; performing semantic analysis on the target question data and the description data corresponding to multiple preset agents to obtain a set of candidate agents that match the target question data, wherein the preset agent is used to reply to the preset question data, and the description data is used to describe the function of the preset agent using natural language, and the candidate agents in the candidate agent set are preset agents whose semantic similarity between the description data and the target question data meets the preset conditions; in the candidate agent set, selecting the target agent with the greatest semantic similarity to the target question data; using the target agent to reply to the target question data to generate target reply data.

[0077] Optionally, the processor may also execute the program code of the following steps: performing semantic analysis on preset question data to obtain question intent; performing semantic analysis on the pre-configured description data of each preset agent to obtain agent intent; detecting the semantic similarity between the question intent and the intent of each agent; and determining a set of candidate agents based on the agent intents that meet preset conditions with semantic similarity.

[0078] Optionally, the processor may also execute the program code of the following steps: determining at least one agent intention whose semantic similarity meets preset conditions as a candidate agent intention; and determining the preset agent corresponding to the candidate agent intention as a candidate agent in the candidate agent set.

[0079] Optionally, the above-mentioned processor can also execute the program code of the following steps: determining at least one intelligent agent intention whose semantic similarity exceeds a preset similarity threshold as a candidate intelligent agent intention; or selecting a preset number of intelligent agent intentions as candidate intelligent agent intentions in order of semantic similarity from high to low.

[0080] Optionally, the description data includes: description text and keywords, and the above-mentioned processor can also execute the program code of the following steps: perform semantic analysis on the description texts corresponding to the target problem data and multiple preset intelligent agents, and obtain a first analysis result corresponding to each preset intelligent agent, wherein the first analysis result is at least used to indicate whether the preset intelligent agent matches the target problem data; determine the preset intelligent agent that does not match the target problem data as an alternative intelligent agent; perform semantic analysis on the keywords corresponding to the target problem data and the alternative intelligent agent, and obtain a second analysis result corresponding to each alternative intelligent agent, wherein the second analysis result is at least used to indicate whether the alternative intelligent agent matches the target problem data; determine the candidate intelligent agent in the candidate intelligent agent set among the preset intelligent agents and alternative intelligent agents that match the target problem data.

[0081] Optionally, the processor may also execute the program code of the following steps: analyzing the first analysis results and the second analysis results of all preset intelligent agents to determine whether there is a situation where all preset intelligent agents cannot match the target question data; when it is determined that all preset intelligent agents cannot match the target question data, generating an alarm message, wherein the alarm message is used to indicate that multiple preset intelligent agents cannot respond to the target question data.

[0082] Optionally, the processor can also execute the program code of the following steps: receiving an agent switching instruction, wherein the agent switching instruction is triggered by a switching plug-in, the switching plug-in corresponds one-to-one to the candidate agent, and the switching plug-in and the target answer data are displayed in the same interface; in response to the agent switching instruction, the target agent is updated to the candidate agent corresponding to the switching plug-in; the updated target agent is used to answer the target question data, and the target question data is updated.

[0083] It can be understood by those skilled in the art that Figure 8 The structure shown is for illustration only, and the computer terminal may also be a smart phone (such as an Android phone, an iOS phone, etc.), a tablet computer, a handheld computer, a mobile Internet device (MID), a PAD, or other terminal devices. Figure 8 It does not limit the structure of the above electronic device. For example, the computer terminal 80 may also include Figure 8 More or fewer components (such as network interfaces, display devices, etc.) shown in, or with Figure 8 Different configurations shown.

[0084] A person skilled in the art can understand that all or part of the steps in the various methods of the above embodiments can be completed by instructing the hardware related to the terminal device through a computer program. The computer program can be stored in a non-volatile medium. The non-volatile storage medium may include: a flash drive, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, etc.

[0085] The embodiment of the present invention further provides a non-volatile storage medium. Optionally, in this embodiment, the non-volatile storage medium can be used to store the program code executed by the data processing method provided in the above embodiment.

[0086] Optionally, in this embodiment, the non-volatile storage medium may be located in any computer terminal in a computer terminal group in a computer network, or in any mobile terminal in a mobile terminal group.

[0087] Optionally, in this embodiment, the non-volatile storage medium is configured to store program code for executing the following steps: obtaining target question data described using natural language; performing semantic analysis on the target question data and the description data corresponding to multiple preset agents, to obtain a set of candidate agents that match the target question data, wherein the preset agent is used to reply to the preset question data, and the description data is used to describe the functions of the preset agent using natural language, and the candidate agents in the candidate agent set are preset agents whose semantic similarity between the description data and the target question data meets preset conditions; in the candidate agent set, selecting the target agent with the greatest semantic similarity to the target question data; using the target agent to reply to the target question data to generate target reply data.

[0088] Optionally, in this embodiment, the non-volatile storage medium is configured to store program code for executing the following steps: performing semantic analysis on preset question data to obtain question intent; performing semantic analysis on the pre-configured description data of each preset agent to obtain agent intent; detecting the semantic similarity between the question intent and the intent of each agent; and determining a set of candidate agents based on the agent intents whose semantic similarity meets preset conditions.

[0089] Optionally, in this embodiment, the non-volatile storage medium is configured to store program code for performing the following steps: determining at least one agent intention whose semantic similarity meets preset conditions as a candidate agent intention; determining the preset agent corresponding to the candidate agent intention as a candidate agent in the candidate agent set.

[0090] Optionally, in this embodiment, the non-volatile storage medium is configured to store program code for performing the following steps: determining at least one agent intention whose semantic similarity exceeds a preset similarity threshold as a candidate agent intention; or selecting a preset number of agent intentions as candidate agent intentions in order of semantic similarity from high to low.

[0091] Optionally, in this embodiment, the description data includes: description text and keywords, and the non-volatile storage medium is configured to store program code for executing the following steps: performing semantic analysis on the description texts corresponding to the target problem data and multiple preset agents, respectively, to obtain a first analysis result corresponding to each preset agent, wherein the first analysis result is at least used to indicate whether the preset agent matches the target problem data; determining the preset agent that does not match the target problem data as an alternative agent; performing semantic analysis on the keywords corresponding to the target problem data and the alternative agent, to obtain a second analysis result corresponding to each alternative agent, wherein the second analysis result is at least used to indicate whether the alternative agent matches the target problem data; and determining a candidate agent in the candidate agent set among the preset agents and alternative agents that match the target problem data.

[0092] Optionally, in this embodiment, the non-volatile storage medium is configured to store program code for executing the following steps: analyzing the first analysis results and the second analysis results of all preset intelligent agents to determine whether there is a situation where all preset intelligent agents cannot match the target question data; when it is determined that all preset intelligent agents cannot match the target question data, generating an alarm message, wherein the alarm message is used to indicate that multiple preset intelligent agents cannot respond to the target question data.

[0093] Optionally, in this embodiment, the non-volatile storage medium is configured to store program code for performing the following steps: receiving an agent switching instruction, wherein the agent switching instruction is triggered by a switching plug-in, the switching plug-in corresponds one-to-one to the candidate agent, and the switching plug-in and the target answer data are displayed in the same interface; in response to the agent switching instruction, updating the target agent to the candidate agent corresponding to the switching plug-in; using the updated target agent to answer the target question data, thereby updating the target question data.

[0094] An embodiment of the present invention further provides a computer program product, including a computer program. Optionally, in this embodiment, when the computer program is executed by a processor, the steps of the data processing method provided in the above embodiment are implemented.

[0095] The serial numbers of the above embodiments of the present invention are for description only and do not represent the advantages or disadvantages of the embodiments.

[0096] In the above embodiments of the present invention, the description of each embodiment has its own focus. For parts that are not described in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0097] In the several embodiments provided in this application, it should be understood that the disclosed technical content can be implemented in other ways. Among them, the device embodiments described above are only exemplary. For example, the division of the units can be a logical function division. In actual implementation, there may be other division methods, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of units or modules, which can be electrical or other forms.

[0098] The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple units. Some or all of the units may be selected according to actual needs to achieve the purpose of the present embodiment.

[0099] In addition, the functional units in the various embodiments of the present invention may be integrated into a single processing unit, each unit may exist physically separately, or two or more units may be integrated into a single unit. The aforementioned integrated units may be implemented in the form of hardware or software functional units.

[0100] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a non-volatile storage medium. Based on this understanding, the technical solution of the present invention is essentially or the part that contributes to the prior art or all or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a non-volatile storage medium and includes several instructions for enabling a computer device (which can be a personal computer, server or network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present invention. The aforementioned non-volatile storage medium includes: U disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), mobile hard disk, magnetic disk or optical disk and other media that can store program code.

[0101] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present invention. These improvements and modifications should also be regarded as within the scope of protection of the present invention.

Claims

1. A data processing method, characterized in that: include: Obtain target problem data described in natural language; Performing semantic analysis on the target question data and the description data corresponding to each of a plurality of preset agents to obtain a set of candidate agents matching the target question data, wherein the preset agents are used to answer the preset question data, the description data are used to describe the functions of the preset agents using natural language, and the candidate agents in the candidate agent set are the preset agents whose semantic similarity between the description data and the target question data meets a preset condition; Selecting a target agent with the greatest semantic similarity to the target problem data from the candidate agent set; The target agent is used to answer the target question data to generate target answer data.

2. The method according to claim 1, characterized in that Performing semantic analysis on the target question data and the description data corresponding to the plurality of preset agents, and obtaining a set of candidate agents matching the target question data, including: Performing semantic analysis on the preset question data to obtain the question intent; Performing semantic analysis on the description data pre-configured for each of the preset intelligent agents to obtain the agent intention; Detecting the semantic similarity between the question intent and each of the agent intents; A set of candidate agents is determined based on the agent intentions whose semantic similarity meets the preset conditions.

3. The method according to claim 2, characterized in that Determining a set of candidate agents based on the agent intentions whose semantic similarity meets the preset conditions includes: Determine at least one of the agent intentions whose semantic similarity meets the preset condition as a candidate agent intention; The preset agent corresponding to the intention of the candidate agent is determined as the candidate agent in the candidate agent set.

4. The method according to claim 3, characterized in that Determining at least one of the agent intentions whose semantic similarity meets the preset condition as a candidate agent intention includes: Determine at least one of the agent intentions whose semantic similarity exceeds a preset similarity threshold as a candidate agent intention; or A preset number of the agent intentions are selected in descending order of the semantic similarity as candidate agent intentions.

5. The method according to claim 1, wherein The description data includes: description text and keywords. Semantic analysis is performed on the description data corresponding to the target question data and multiple preset agents, and a set of candidate agents matching the target question data is obtained, including: Performing semantic analysis on the target question data and the description texts corresponding to the plurality of preset agents, respectively, to obtain a first analysis result corresponding to each of the preset agents, wherein the first analysis result is used to at least indicate whether the preset agent matches the target question data; Determine the preset intelligent agent that does not match the target question data as an alternative intelligent agent; Performing semantic analysis on the target question data and the keywords corresponding to the candidate agents to obtain a second analysis result corresponding to each candidate agent, wherein the second analysis result is at least used to indicate whether the candidate agent matches the target question data; The candidate agent in the candidate agent set is determined among the preset agents and the candidate agents that match the target question data.

6. The method according to claim 5, characterized in that After performing semantic analysis on the target question data and the keywords corresponding to the candidate agents to obtain a second analysis result corresponding to each candidate agent, the method further includes: Analyzing the first analysis results and the second analysis results of all the preset agents to determine whether there is a situation where all the preset agents cannot match the target question data; When it is determined that all of the preset agents cannot match the target question data, an alarm message is generated, wherein the alarm message is used to indicate that multiple preset agents cannot answer the target question data.

7. The method according to claim 1, characterized in that After using the target agent to answer the target question data and generate target answer data, the method further includes: receiving an agent switching instruction, wherein the agent switching instruction is triggered by a switching plug-in, the switching plug-in corresponds to the candidate agent one-to-one, and the switching plug-in and the target response data are displayed in the same interface; In response to the agent switching instruction, updating the target agent to the candidate agent corresponding to the switching plug-in; The updated target agent is used to respond to the target question data, thereby updating the target question data.

8. A data processing device, characterized in that: include: An acquisition module is used to obtain target problem data described in natural language; a semantic analysis module for performing semantic analysis on the target question data and the description data corresponding to each of the plurality of preset agents, to obtain a set of candidate agents matching the target question data, wherein the preset agents are used to answer the preset question data, the description data are used to describe the functions of the preset agents using natural language, and the candidate agents in the candidate agent set are the preset agents whose semantic similarity between the description data and the target question data meets a preset condition; A selection module is used to select a target agent with the greatest semantic similarity to the target question data from the candidate agent set; A generation module is used to use the target agent to respond to the target question data and generate target response data.

9. An electronic device comprising a memory and a processor, characterized in that: A computer program is stored in the memory, and the processor is configured to execute the data processing method according to any one of claims 1 to 7 through the computer program.

10. A computer program product comprising computer instructions, characterized in that When the computer instructions are executed by a processor, the steps of the data processing method according to any one of claims 1 to 7 are implemented.

Images