WO2022126965A1 - 智能问答方法、装置、设备及存储介质 - Google Patents

智能问答方法、装置、设备及存储介质 Download PDF

Info

Publication number
WO2022126965A1
WO2022126965A1 PCT/CN2021/090197 CN2021090197W WO2022126965A1 WO 2022126965 A1 WO2022126965 A1 WO 2022126965A1 CN 2021090197 W CN2021090197 W CN 2021090197W WO 2022126965 A1 WO2022126965 A1 WO 2022126965A1
Authority
WO
WIPO (PCT)
Prior art keywords
question
business
initial
questions
type
Prior art date
Application number
PCT/CN2021/090197
Other languages
English (en)
French (fr)
Inventor
邓柏伟
谢万富
李婧宇
Original Assignee
平安科技（深圳）有限公司
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 平安科技（深圳）有限公司 filed Critical 平安科技（深圳）有限公司
Publication of WO2022126965A1 publication Critical patent/WO2022126965A1/zh

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
    • G06F16/33—Querying
    • G06F16/332—Query formulation
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
    • G06F16/35—Clustering; Classification
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
    • G06F16/36—Creation of semantic tools, e.g. ontology or thesauri

Definitions

  • the present application relates to the field of artificial intelligence, and in particular, to an intelligent question answering method, apparatus, device and storage medium.
  • the embodiments of the present application provide an intelligent question answering method, apparatus, device and storage medium, which can improve the question answering efficiency of the intelligent question answering system and optimize the user experience.
  • the embodiments of the present application provide an intelligent question answering method, the method comprising:
  • the terminal device obtains, through the user interface, user operation data for triggering the activation of the target service, and determines the initial service problem associated with the above-mentioned user operation data;
  • the initial standard questions corresponding to the above-mentioned initial business questions, the initial standard answers corresponding to the above-mentioned initial standard questions, and the business feature labels contained in the initial standard questions are determined from the standard question base;
  • the first type of related question sets semantically related to the above-mentioned initial standard questions are determined from the question-and-answer knowledge aggregation graph, and the first type of related question sets that are semantically related to the above-mentioned initial standard questions, as well as those contained in the above-mentioned initial standard questions
  • an intelligent question answering device comprising:
  • a problem acquisition module configured to acquire user operation data for triggering the start of the target service through the user interface, and determine the initial service problem associated with the above-mentioned user operation data
  • the semantic analysis module is used to determine the initial standard question corresponding to the above-mentioned initial business question, the initial standard answer corresponding to the above-mentioned initial standard question, and the business features contained in the initial standard question from the standard question base based on the semantic analysis of the above-mentioned initial business question Label;
  • the association aggregation module is used to determine the first type of association question set that is semantically related to the above initial standard question from the question-and-answer knowledge aggregation graph according to the above-mentioned initial standard question and the business feature labels contained in the above-mentioned initial standard question, and the set of related questions related to the above-mentioned initial standard question.
  • the second type of associated problem set associated with the business feature labels included in the initial standard problem;
  • the result output module is used to determine the initial standard answer of the above-mentioned initial standard question, the above-mentioned first type of related question set and the above-mentioned second type of related question set as the answer text of the above-mentioned initial business question, and the answer to the above-mentioned initial business question.
  • the text is output to the user interface that initiates the above target service.
  • an embodiment of the present application provides a terminal device, where the terminal device includes a processor and a memory, and the processor and the memory are connected to each other.
  • the memory is used to store a computer program, the computer program includes program instructions, and the processor is configured to invoke the above program instructions to execute the following method:
  • the initial standard questions corresponding to the above-mentioned initial business questions, the initial standard answers corresponding to the above-mentioned initial standard questions, and the business feature labels contained in the initial standard questions are determined from the standard question base;
  • the first type of related question sets semantically related to the above-mentioned initial standard questions are determined from the question-and-answer knowledge aggregation graph, and the first type of related question sets that are semantically related to the above-mentioned initial standard questions, as well as those contained in the above-mentioned initial standard questions
  • an embodiment of the present application provides a computer-readable storage medium, where the computer-readable storage medium stores a computer program, the computer program includes program instructions, and the program instructions, when executed by a processor, cause the processor to execute The following methods:
  • the initial standard questions corresponding to the above-mentioned initial business questions, the initial standard answers corresponding to the above-mentioned initial standard questions, and the business feature labels contained in the initial standard questions are determined from the standard question base;
  • the first type of related question sets semantically related to the above-mentioned initial standard questions are determined from the question-and-answer knowledge aggregation graph, and the first type of related question sets that are semantically related to the above-mentioned initial standard questions, as well as those contained in the above-mentioned initial standard questions
  • the embodiments of the present application can optimize the user experience and improve the question-answering efficiency of the intelligent question-answering system.
  • FIG. 1 is a schematic flowchart of an intelligent question answering method provided by an embodiment of the present application.
  • FIG. 2 is a schematic flowchart of constructing a question-and-answer knowledge aggregation graph provided by an embodiment of the present application
  • FIG. 3 is a schematic structural diagram of a question and answer aggregation map provided in an embodiment of the present application.
  • Fig. 4 is another schematic flowchart of the intelligent question answering method provided by the embodiment of the present application.
  • FIG. 5 is a schematic structural diagram of an intelligent question answering device provided by an embodiment of the present application.
  • FIG. 6 is another schematic structural diagram of an intelligent question answering device provided by an embodiment of the present application.
  • FIG. 7 is a schematic structural diagram of a terminal device provided by an embodiment of the present application.
  • the technical solutions of the present application may relate to the fields of artificial intelligence and/or big data technologies, such as natural language processing technologies, to realize intelligent question answering, thereby promoting the construction of smart cities.
  • the data involved in this application such as business questions, standard questions, standard answers, labels and/or answer texts, may be stored in a database, or may be stored in a blockchain, such as distributed storage through a blockchain, This application is not limited.
  • the use of artificial intelligence technology to build an intelligent question answering system for information in a certain field can better promote the development of technology in this field.
  • the construction of an intelligent question answering system for insurance business information can help people quickly understand the applicable groups of a certain insurance, the scope of protection, the scope of claims, and the methods of claims.
  • the application scope of the intelligent question answering system is very wide. This application only uses the construction of the intelligent question answering system for insurance business information in the insurance field as the application scenario. The embodiments are essentially the same, and are not repeated here.
  • the terminal device can obtain the user operation data and determine the user operation.
  • the initial business question to which the data is associated ie, the initial business question entered by the user in the user interface.
  • the terminal device obtains user operation data and determines the initial business problem (Which auto insurance is better for an 18-year-old?).
  • the terminal device can perform semantic analysis on the initial business problem, determine the initial standard problem corresponding to the initial business problem in the standard problem database (the types of auto insurance that can be insured at the age of 18), and obtain the initial standard problem (can be insured at the age of 18).
  • the corresponding standard answers (the first car insurance, the second car insurance, etc. can be insured at the age of 18), and the business feature labels (“Type of car insurance” and "Insurance Age”).
  • the terminal device can input the initial standard questions and the business feature labels contained in the initial standard questions into the question-and-answer knowledge aggregation graph, and obtain the question and answer aggregation graph based on the initial standard questions (the types of auto insurance that can be insured at the age of 18) through the question-and-answer aggregation graph in the question-and-answer knowledge aggregation graph.
  • a set of first-type related questions (eg, first-type related question 1, first-type related question 2) that are semantically related to the initial standard question, that is, to obtain the questions related to the initial business question in the user's question-and-answer habit collection.
  • the first type of association problem set may include: the first type of association problem 1 (the protection scope of the first auto insurance), the first type of association problem 2 (the protection scope of the second motor insurance), and the like.
  • the set of the second type of association problem associated with the business feature label (for example, the second type of association problem 1, the second type of association problem 2), that is, in the business feature field corresponding to the business feature label, the Collection of questions.
  • the second type of related question set may include: the second type of related question 1 (the scope of auto insurance claims that can be enjoyed by the 18-year-old), and the second type of related question 2 (the method of auto insurance claims that can be enjoyed by the 18-year-old).
  • the terminal device can determine the initial standard answer of the initial standard question, the first type of related question set and the second related type of question set as the answer text of the initial business question, and output the answer text of the initial business question to the start-up target service User Interface.
  • the terminal device can acquire the user's subsequent operations on the user interface, so as to continue the smart question and answer task or end the smart question and answer task.
  • FIG. 1 is a schematic flowchart of an intelligent question answering method provided by an embodiment of the present application.
  • the method provided by this embodiment of the present application may include that the terminal device obtains, through the user interface, user operation data for triggering the start of the target service, and determines the initial service problem associated with the user operation data; Determine the initial standard question corresponding to the initial business question, the initial standard answer corresponding to the initial standard question, and the business feature label contained in the initial standard question; In the graph, a set of related questions of the first type that is semantically related to the initial standard question, and a set of related questions of the second type related to the business feature labels contained in the initial standard question are determined; the initial standard answer of the initial standard question, the first The class-related question set and the second class-related question set are determined as the answer text of the initial business question, and the answer text of the initial business question is output to the user interface for starting the target business.
  • the intelligent question answering method and device provided by the embodiments of the present application will be exemplified below by taking the intelligent question answering system of the insurance system as the intelligent question answering system and the terminal device as the execution subject of the application.
  • the terminal device obtains, through a user interface, user operation data for triggering the activation of a target service, and determines an initial service problem associated with the user operation data.
  • the terminal device when the user triggers the target service (ie, the smart question answering service) by clicking on the user interface on the terminal device, the terminal device can obtain the user operation data and determine the initial service problem associated with the user operation data (ie, the initial business question entered by the user in the user interface). For example, the terminal device can obtain the user operation data that the user clicks to trigger the smart question answering service and the input question, and determines the initial service question.
  • the initial business question is: Which auto insurance is better for an 18-year-old? Specifically, it can be determined according to the actual application scenario, and is not limited here.
  • S102 Based on the semantic analysis of the initial business question, determine an initial standard question corresponding to the initial business question, an initial standard answer corresponding to the initial standard question, and a business feature label included in the initial standard question from the standard question base.
  • the terminal device can perform semantic analysis on the initial business problem, determine the initial standard problem corresponding to the initial business problem in the standard problem database (the types of auto insurance that can be insured at the age of 18), and obtain the initial standard problem (Types of car insurance that can be insured at the age of 18) corresponding standard answers (the first car insurance, second car insurance, etc. can be insured at the age of 18), and the business feature labels included in the initial standard question (types of car insurance that can be insured at the age of 18) (“Type of auto insurance” and “Insurance age”).
  • the standard question base is constructed from multiple standard questions, and each standard question in the standard question base has its own business feature label.
  • the terminal device may acquire a plurality of service data from the service database of the target service, and determine the service problem associated with each service data. Afterwards, the terminal device can perform semantic analysis on these business questions, and cluster each business question according to the question intent, that is, classify the business questions with the same question intent but different questioning methods into one type of standard questions, so as to obtain multiple standard questions. question. After obtaining a plurality of standard questions, the terminal device can determine the standard answer corresponding to each standard question based on the business data, and at the same time, perform keyword extraction on these standard questions to obtain service feature labels contained in each standard question, thereby generating a standard question library.
  • the business feature label can be classified based on keywords corresponding to business problems associated with business data, or based on business features (insurance type, insurance age, claim scope, claim settlement method, etc.).
  • business features insurance type, insurance age, claim scope, claim settlement method, etc.
  • the terminal device can mark various business data, such as business data 1 (ask: which car insurance is better to insure at the age of 18, answer: the car insurance that can be insured at the age of 18 includes the first car insurance, the third car insurance Second auto insurance, etc.) and business data 2 (ask: the types of auto insurance that can be insured at the age of 18, A: the first auto insurance, the second auto insurance, etc.
  • business data 1 ask: which car insurance is better to insure at the age of 18, answer: the car insurance that can be insured at the age of 18 includes the first car insurance, the third car insurance Second auto insurance, etc.
  • business data 2 ask: the types of auto insurance that can be insured at the age of 18, A: the first auto insurance, the second auto insurance, etc.
  • the auto insurance can be insured at the age of 18) to mark, and get the marked business data 1 (18-year-old Which auto insurance is better to insure//Type of auto insurance/Insurance age//The auto insurance that can be insured at the age of 18 includes the first auto insurance, the second auto insurance, etc.) and the marked business data 2 (The types of auto insurance that can be insured at the age of 18/ /Types of car insurance/Age of insurance//The 18-year-old can insure the first car insurance, the second car insurance, etc.).
  • the terminal device can perform semantic recognition on the marked business data 1 and marked business data 2, and obtain the problem intention of the marked business data 1 (auto insurance that can be insured at the age of 18) and the problem intention of the marked business data 2 (Auto insurance that can be insured at the age of 18), cluster the marked business data 1 and marked business data 2 into a unified category, and select the problem part of the marked business data 2 with the highest frequency as the intention of such a problem
  • the standard question (the types of car insurance that can be insured at the age of 18) (the types of car insurance that can be insured at the age of 18)
  • the marked part of the business feature label of the business data 2 is taken as the standard question (the types of car insurance that can be insured at the age of 18) business feature label (type of auto insurance, age of insured)
  • the answer part of the marked business data 2 is used as the standard answer for the standard question (type of auto insurance that can be insured at the age of 18) (the first auto insurance that can be insured at the age of 18, second car insurance, etc.).
  • the terminal device can input the initial standard question and the business feature label contained in the initial standard question into the question-and-answer knowledge aggregation graph, and through the question-and-answer aggregation graph in the question-and-answer knowledge aggregation graph, based on the initial standard question (18 type of auto insurance that can be insured at the age of 1) to obtain the first type of related question set (for example, the first type of related question 1, the first type of related question 2) that is semantically related to the initial standard question, that is, obtained in the user's question and answer habits, A collection of questions associated with the initial business problem.
  • the first type of related question set for example, the first type of related question 1, the first type of related question 2
  • the first type of association problem set may include: the first type of association problem 1 (the protection scope of the first auto insurance), the first type of association problem 2 (the protection scope of the second motor insurance), and the like.
  • the knowledge aggregation graph in the question-and-answer knowledge aggregation graph based on the business feature labels (“auto insurance type” and “insurance age”) contained in the initial standard question (the types of auto insurance that can be insured at the age of 18), we can obtain the same data as the initial standard question.
  • the set of the second type of association problem associated with the business feature label for example, the second type of association problem 1, the second type of association problem 2), that is, in the business feature field corresponding to the business feature label, the Collection of questions.
  • the second type of related question set may include: the second type of related question 1 (the scope of auto insurance claims that can be enjoyed by the 18-year-old), and the second type of related question 2 (the method of auto insurance claims that can be enjoyed by the 18-year-old).
  • FIG. 2 is a schematic flowchart of constructing a question-and-answer knowledge aggregation graph provided by an embodiment of the present application.
  • the above-mentioned method for constructing a question-and-answer knowledge aggregation graph may include the implementation manners provided by each of the following steps S201 to S205.
  • S201 Acquire a plurality of historical business data from a business database of a target business, and determine a plurality of historical user data and a plurality of customer service reference data in the plurality of historical business data.
  • S202 According to the historical business questions associated with the multiple historical user data, determine the standard questions corresponding to the historical business questions in the standard question database as the sample business questions, and determine the questions of the sample business questions according to the questioning order of the historical business questions order, the user sample data is determined according to each sample business question and the questioning order of each sample business question.
  • S203 According to the reference business questions associated with the multiple customer service reference data, determine the standard questions corresponding to each reference business question in the standard question base as the customer service sample questions, and the business feature labels included in the customer service sample questions.
  • the service feature labels included in the sample questions determine the customer service sample data.
  • the terminal device may acquire multiple historical user data and multiple customer service reference data from the service database of the target service.
  • the terminal device may obtain sample business questions that conform to the user's questioning habits based on the historical user data (that is, the sample business questions obtained corresponding to the historical business questions in the standard question database).
  • the sample business questions have a questioning order based on the user's questioning habits. According to the questioning order, historical business questions with semantic correlation can usually form a directed closed loop.
  • the terminal device may obtain customer service sample questions that conform to domain knowledge based on the customer service reference data (that is, refer to the business questions and correspondingly obtained customer service sample questions in the standard question database).
  • customer service sample questions can be questions provided by customer service in related fields that are easily ignored by users, but users may need to know, and related customer service sample questions usually have the same or similar business feature labels.
  • S204 Determine the nodes of the Q&A aggregation graph in the Q&A knowledge aggregation graph through each sample business question in the user sample data, determine the node connection relationship of the Q&A aggregation graph through the questioning sequence of each sample business question, and combine the nodes of the Q&A aggregation graph Connect according to the node connection relationship of the question and answer aggregation graph to obtain the question and answer aggregation graph.
  • the terminal device uses the user sample data to construct a question-and-answer aggregation graph, and a directed graph with each sample business question as a node and the questioning order of each sample business question as an edge can be constructed in the question and answer aggregation graph. . That is, a sample business problem in a user sample data is taken as v i , and the next sample business problem raised by the user after v i in the user sample data is taken as a first-type related problem v i1 of v i .
  • each edge starts with the business sample question v i as the starting point, and takes the first type of related questions of the business sample question v i as the end point, so that the terminal device can obtain the first type of business sample question according to the constructed question and answer aggregation graph.
  • a set of associated questions ie, questions the user might want to ask next, in the order in which they were asked).
  • the number of times vi and vi ij appear in the user sample data in succession can be recorded as C ij , so as to obtain from the business sample problem vi
  • the question-and-answer knowledge aggregation graph obtained at this time is not friendly enough for new or unpopular knowledge, and it is difficult to find the correlation between user sample data with few or no access records.
  • the terminal device can construct a latent vector of the business sample question for each business sample question, and use the connection probability of each first type of related questions corresponding to each business sample question to construct a question-answer aggregation graph, Therefore, the first type of association problem of the unpopular business sample problem can be obtained through the latent vector of the unpopular business sample problem.
  • connection probability of ,v ij > constructs the question and answer aggregation graph, so that the loss function of the question and answer aggregation graph is minimized, and based on the hidden vector of the unpopular business sample questions, the first type of related questions of the unpopular business sample questions are obtained.
  • FIG. 3 is a schematic structural diagram of the question and answer aggregation graph provided by the embodiment of the present application. Because the sample business questions are all standard questions, in essence, all nodes in the Q&A aggregation graph can be expressed by standard questions. For the convenience of description, standard questions will be used to represent the nodes of the Q&A aggregation graph.
  • A1, A2, A3, A4, and A5 represent standard question A1, standard question A2, standard question A3, standard question A4, and standard question A5, respectively.
  • Standard question A2 and standard question A3 are the primary first type related questions of standard question A1.
  • Standard question A3 is a secondary first type related question of standard question A2, and standard question A4 is a secondary first type linked question of standard question A3.
  • Standard question A5 and standard question A1 are secondary first type related questions of standard question A4.
  • Standard question A1 is a secondary type 1 related question of standard question A5.
  • the standard question A1, the standard question A3 and the standard question A4 constitute the initial standard question closed loop 1 (including 3 standard questions).
  • Standard question A1, standard question A2, standard question A3 and standard question A4 constitute the initial standard question closed loop 2 (including 4 standard questions).
  • the standard question A1, the standard question A3, the standard question A4 and the standard question A5 constitute the initial standard question closed loop 3 (including 4 standard questions).
  • the standard question A1, the standard question A2, the standard question A3, the standard question A4 and the standard question A5 constitute the initial standard question closed loop 4 (including 5 standard questions).
  • S205 Determine the nodes of the knowledge aggregation graph in the Q&A knowledge aggregation graph by using the customer service sample questions in the customer service sample data, determine the node connection relationship of the knowledge aggregation graph by using the business feature labels included in the customer service sample questions, and combine the nodes of the Q&A aggregation graph. The nodes are connected according to the node connection relationship of the question and answer aggregation graph to obtain the knowledge aggregation graph.
  • the terminal device uses the customer service sample data to train the knowledge aggregation graph, and can construct in the knowledge aggregation graph each customer service sample question as a node and the service feature label contained in each customer service sample question as an edge. to the diagram. That is, a customer service sample question in a customer service sample data is taken as ui , and the service feature label of the customer service sample question ui is A i , where A i includes multiple sub-service feature labels a ij .
  • U i ⁇ u ij of the customer service sample questions u i , i ⁇ N,j ⁇ N ⁇ .
  • one endpoint of the edge is the customer service sample question ui
  • the other endpoint is the second type of associated problem u ij of the customer service sample question ui .
  • the terminal device can obtain the second type of related problem sets of the target customer service sample problems (according to the service feature labels contained in the customer service sample questions, questions the user might want to ask next).
  • the terminal device can construct a question-and-answer aggregation graph according to each sample business question and the questioning sequence of each sample business question, that is, construct a question-and-answer aggregate graph based on the business question data of the user's questioning habits.
  • the terminal device can construct a knowledge aggregation graph according to the customer service sample questions and the business feature labels contained in the customer service sample questions, that is, construct a knowledge aggregation graph based on the business problem data in each business feature field. It enables the terminal device to provide the first type of related questions that the user may ask next, and the second type of related questions that the user may want to know through the question-and-answer knowledge aggregation graph, thereby optimizing the user experience and improving the intelligent question answering system. question answering efficiency.
  • S104 Determine the initial standard answer of the initial standard question, the set of related questions of the first type, and the set of related questions of the second type as the answer text of the initial business question, and output the answer text of the initial business question to the user interface for starting the target business .
  • the terminal device can answer the initial standard questions of the initial standard questions (the 18-year-old can insure the first car insurance, the second car insurance, etc.), the first type of association question set (for example, the first type of association question 1 , the first type of related question 2, etc.) and the second set of related questions (for example, the second type of related question 1, the second type of related question 2, etc.) are determined as the answer text of the initial business question, and the The answer text is output to the user interface that initiates the target business.
  • the first type of association problem set includes: first type of association problem 1 (the protection scope of the first car insurance), and the first type of association problem 2 (the protection scope of the second car insurance) and the like.
  • the second-type related problem set includes: the second-type related problem 1 (the scope of auto insurance claims that can be enjoyed by the 18-year-old), and the second-type related problem 2 (the auto insurance claim method that can be enjoyed by the 18-year-old).
  • the terminal device can acquire the user's subsequent operations on the user interface, so as to continue the smart question and answer task or end the smart question and answer task.
  • the terminal device obtains the initial business problem associated with the user operation data by acquiring the user operation data, and based on the semantic analysis of the initial business problem, the initial standard problem corresponding to the initial business problem can be determined in the standard problem database , thereby standardizing the initial business question, simplifying the subsequent analysis process for the initial business question, and at the same time determining the initial standard answer corresponding to the initial standard question and the business feature label contained in the initial standard question.
  • the first type of related question sets that are semantically related to the initial standard questions can be determined from the question answering knowledge aggregation graph.
  • the first type of associated question set that is, the set of questions associated with the initial business question in the user's question-and-answer habit.
  • the second type of associated question sets associated with the business feature labels contained in the initial standard questions can be determined from the question-and-answer knowledge aggregation graph.
  • the second type of associated question set that is, a set of questions associated with the initial business question in the business feature field corresponding to the business feature label.
  • the initial standard answer of the initial standard question, the first type of related question set, and the second type of related question set are determined as the answer text of the initial business question and output to the user interface, which can be provided to the user in the initial standard answer corresponding to the initial business question.
  • it provides the first type of related question set that the user may ask next, and the second type of related question set that the user may want to know next, so as to optimize the user experience and improve the question answering efficiency of the intelligent question answering system.
  • FIG. 4 is another schematic flowchart of the intelligent question answering method provided by the embodiment of the present application.
  • the terminal device obtains, through the user interface, user operation data for triggering the activation of the target service, and determines the initial service problem associated with the user operation data.
  • the terminal device when the user triggers the target service (ie, the smart question answering service) by clicking on the user interface on the terminal device, the terminal device can obtain the user operation data and determine the initial service problem associated with the user operation data (ie, the initial business question entered by the user in the user interface). For example, the terminal device can obtain the user operation data that the user clicks to trigger the smart question answering service and the input question, and determines the initial service question.
  • the initial business question is: Which auto insurance is better for an 18-year-old? Specifically, it can be determined according to the actual application scenario, and is not limited here.
  • S302 Based on the semantic analysis of the initial business question, determine an initial standard question corresponding to the initial business question, an initial standard answer corresponding to the initial standard question, and a business feature label included in the initial standard question from the standard question base.
  • the terminal device can perform semantic analysis on the initial business problem, determine the initial standard problem corresponding to the initial business problem in the standard problem database (the types of auto insurance that can be insured at the age of 18), and obtain the initial standard problem (Types of car insurance that can be insured at the age of 18) corresponding standard answers (the first car insurance, second car insurance, etc. can be insured at the age of 18), and the business feature labels included in the initial standard question (types of car insurance that can be insured at the age of 18) (“Type of auto insurance” and “Insurance age”).
  • the standard question base is constructed from multiple standard questions, and each standard question in the standard question base has its own business feature label.
  • S303 Determine from the question-and-answer aggregation graph a primary first-type related question that is semantically related to the initial standard question.
  • S304 Determine from the question-and-answer aggregation graph the secondary first-class related questions semantically related to the primary first-related related questions, and determine from the question-answer aggregation graph the secondary first-related questions semantically related to the secondary first-class related questions Class related problems until initial standard problems appear in the secondary first related problems of any level, so as to obtain a closed loop of initial business problems composed of first related problems at all levels.
  • all nodes in the Q&A aggregation graph can be expressed by standard questions.
  • standard questions will be used to represent Q&A. Aggregate the nodes of the graph, and take the closed loop of the initial business problem as the equivalent representation of the closed loop of the initial standard problem in the previous section.
  • A1, A2, A3, A4, and A5 represent standard question A1, standard question A2, standard question A3, standard question A4, and standard question A5, respectively.
  • the terminal device may determine, from the question-and-answer aggregation graph, the primary first-type associated questions (standard question A2, standard question A3) that are semantically related to the initial standard question (standard question A1).
  • the terminal device can determine the secondary first-type related questions (standard question A3, standard question A4) that are semantically related to the primary first-type related questions (standard question A2, standard question A3) from the question-and-answer aggregation graph, and from the question and answer In the aggregated graph, determine the secondary first type association questions (standard question A4, standard question A5) that are semantically related to the secondary first type association questions (standard question A3, standard question A4), until any level of the first type association
  • the initial standard questions appear in the secondary first type related questions of the questions (standard questions A4, standard questions A5), so as to obtain a closed loop of initial business questions composed of the first related questions at all levels.
  • the initial business problem closed loop 1 (including two first-type related problems) constituted by the standard problem A1, the standard problem A3 and the standard problem A4.
  • the initial business problem closed loop 2 (including 3 first-type related problems) constituted by the standard problem A1, the standard problem A2, the standard problem A3 and the standard problem A4.
  • the initial business problem closed loop 3 (including 3 first-type related problems) constituted by the standard problem A1, the standard problem A3, the standard problem A4 and the standard problem A5.
  • Standard problem A1, standard problem A2, standard problem A3, standard problem A4, and standard problem A5 constitute an initial business problem closed loop 4 (including four first-type related problems).
  • the terminal device may determine, based on the connection probability between the first type of related questions at all levels and the initial business question included in the closed loop of the initial business question in the question-and-answer aggregation graph, the first type of related questions at all levels included in the closed loop of the initial business question.
  • the probability of connection for a class of association problems to appear after the initial standard problem According to the first type of related problems that appear after the initial standard problem in the closed loop of the initial business problem and whose connection probability is greater than the threshold, determine the set of the first type of related problems of the initial standard problem.
  • the terminal device can input the initial standard question and the business feature label contained in the initial standard question into the question-and-answer knowledge aggregation graph, and through the question-and-answer aggregation graph in the question-and-answer knowledge aggregation graph, based on the initial standard question (18 type of auto insurance that can be insured at the age of 1) to obtain the first type of related question set (for example, the first type of related question 1, the first type of related question 2) that is semantically related to the initial standard question, that is, obtained in the user's question and answer habits, A collection of questions associated with the initial business problem.
  • the first type of association problem set includes: the first type of association problem 1 (the protection scope of the first automobile insurance), the first type of association problem 2 (the protection scope of the second automobile insurance), and the like.
  • S305 Based on the number of first-type related problems at all levels included in the initial business problem closed loop, determine a first-type related problem set of the initial standard problem.
  • the terminal device may, when the number of the first type of association problems at all levels included in the closed loop of initial business problems is less than a threshold, set the first type of association problems at all levels included in the closed loop of initial business problems
  • the class association problem is determined as the first class association problem set of the initial standard problem.
  • the terminal device may set the threshold value of the number of the first type of associated problems at all levels included in the initial service problem closed loop to 3.
  • the first type of related problems included in the closed loop 1 of the initial business problem is determined as the first type of related problems of the initial standard problem set (standard Question A3, Standard Question A4).
  • the terminal device may determine, based on the connection probability between the first type of related questions at all levels and the initial business question included in the closed loop of the initial business question in the question-and-answer aggregation graph, the first type of related questions at all levels included in the closed loop of the initial business question.
  • the connection probability of a type of association problem that appears after the initial standard problem, and the first type of association problem that appears after the initial standard problem in the closed loop of the initial business problem is greater than the threshold. Determine the first type of association problem set of the initial standard problem .
  • the terminal device can determine, from the question-and-answer aggregation graph, the first type of related questions at all levels that are semantically related to the initial standard questions, and obtain a closed loop of initial business questions.
  • the terminal device can filter the first type of related problems at all levels in the closed loop of initial business problems by the number of problems contained in the closed loop of initial business problems, and can also use the connection probability to filter the first type of related problems at all levels in the closed loop of initial business problems. , to eliminate the first type of related questions that are too low semantically related to the initial standard question, avoid excessive correlation, and improve the question-answering efficiency of the intelligent question-answering system.
  • S306 According to the initial standard question and the business feature label included in the initial standard question, determine a second type of associated question set associated with the business feature label included in the initial standard question from the knowledge aggregation graph.
  • the terminal device can use the knowledge aggregation graph in the question-and-answer knowledge aggregation graph, based on the business feature labels (“auto insurance type” and “insurance insurance type”) included in the initial standard question (18-year-old insured auto insurance types). age”) to obtain the second type of association question set (for example, the second type of association question 1, the second type of association question 2) associated with the business feature labels contained in the initial standard question, that is, in the business characteristics label corresponding to the business A collection of problems associated with the initial business problem in the feature domain.
  • the business feature labels (“auto insurance type” and “insurance insurance type” included in the initial standard question (18-year-old insured auto insurance types). age”
  • age the second type of association question set associated with the business feature labels contained in the initial standard question
  • the second type of related problem set includes: the second type of related problem 1 (the scope of auto insurance claims that can be enjoyed by the 18-year-old), and the second-type related problem 2 (the method of auto insurance claims that can be enjoyed by the 18-year-old).
  • S307 Determine the initial standard answer to the initial standard question, the first type of related question set, and the second type of related question set as the answer text of the initial business question, and output the answer text of the initial business question to the user interface for starting the target business .
  • the terminal device can answer the initial standard questions of the initial standard questions (the 18-year-old can insure the first car insurance, the second car insurance, etc.), the first type of association question set (for example, the first type of association question 1 , the first type of related question 2, etc.) and the second set of related questions (for example, the second type of related question 1, the second type of related question 2, etc.) are determined as the answer text of the initial business question, and the The answer text is output to the user interface that initiates the target business.
  • the first type of association problem set includes: first type of association problem 1 (the protection scope of the first car insurance), and the first type of association problem 2 (the protection scope of the second car insurance) and the like.
  • the second-type related problem set includes: the second-type related problem 1 (the scope of auto insurance claims that can be enjoyed by the 18-year-old), and the second-type related problem 2 (the auto insurance claim method that can be enjoyed by the 18-year-old).
  • the terminal device can acquire the user's subsequent operations on the user interface, so as to continue the smart question and answer task or end the smart question and answer task.
  • the terminal device can determine the first type of related questions at all levels that are semantically related to the initial standard questions from the question-and-answer aggregation graph, and obtain a closed loop of initial business questions.
  • the terminal device can filter the first type of related problems at all levels in the closed loop of initial business problems by the number of problems contained in the closed loop of initial business problems, and can also use the connection probability to filter the first type of related problems at all levels in the closed loop of initial business problems. , to eliminate the first type of related questions that are too low semantically related to the initial standard question, avoid excessive correlation, and improve the question-answering efficiency of the intelligent question-answering system.
  • FIG. 5 is a schematic structural diagram of an intelligent question answering device provided by an embodiment of the present application.
  • the device includes:
  • the problem obtaining module 401 is configured to obtain, through a user interface, user operation data for triggering the start of the target service, and determine the initial service problem associated with the above-mentioned user operation data.
  • the question acquisition module 401 can acquire the user operation data, and determine the user operation data associated with it.
  • the initial business problem that is, the initial business problem entered by the user in the user interface.
  • the question obtaining module 401 can obtain the user operation data that the user clicks to trigger the intelligent question answering service and the input question, and determines the initial business question.
  • the initial business question is: Which auto insurance is better for an 18-year-old? Specifically, it can be determined according to the actual application scenario, and is not limited here.
  • the semantic analysis module 402 is configured to determine the initial standard question corresponding to the above-mentioned initial business question, the initial standard answer corresponding to the above-mentioned initial standard question, and the business included in the initial standard question from the standard question base based on the semantic analysis of the above-mentioned initial business question Feature label.
  • the semantic analysis module 402 can perform semantic analysis on the initial business problem, determine the initial standard problem (the types of auto insurance that can be insured at the age of 18) corresponding to the initial business problem in the standard problem database, and obtain the initial standard problem.
  • Standard questions types of auto insurance that can be insured at the age of 18
  • the standard answers the first auto insurance, second auto insurance, etc. can be insured at the age of 18
  • the business included in the initial standard question types of auto insurance that can be insured at the age of 18
  • Feature labels (“Type of auto insurance” and "Insurance age”).
  • the standard question base is constructed from multiple standard questions, and each standard question in the standard question base has its own business feature label.
  • FIG. 6 is another schematic structural diagram of an intelligent question answering device provided by an embodiment of the present application.
  • the intelligent question answering device further includes:
  • the standard question database generation module 412 is used to obtain a plurality of business data from the business database of the above-mentioned target business, and determine the business problems associated with each business data, and cluster the business problems associated with the above-mentioned business data based on semantic analysis A plurality of standard questions are obtained, and a standard question library is generated according to the above-mentioned plurality of standard questions.
  • the association aggregation module 403 is configured to, according to the above-mentioned initial standard questions and the business feature labels contained in the above-mentioned initial standard questions, from the question-and-answer knowledge aggregation graph, determine the first type of association question sets that are semantically related to the above-mentioned initial standard questions, and the set of associated questions related to the above-mentioned initial standard questions. The set of the second type of association questions associated with the business feature labels included in the above initial standard questions.
  • the association aggregation module 403 may input the initial standard questions and the business feature labels included in the initial standard questions into the question-and-answer knowledge aggregation graph, and through the question-and-answer aggregation graph in the question-and-answer knowledge aggregation graph, based on the initial standard questions (Types of auto insurance that can be insured at the age of 18) to obtain the first-type related question set (for example, the first-type related question 1, the first-type related question 2) that is semantically related to the initial standard question, that is, to obtain the user's question and answer habits , the collection of questions associated with the initial business question.
  • the initial standard questions Types of auto insurance that can be insured at the age of 18
  • the first type of association problem set may include: the first type of association problem 1 (the protection scope of the first auto insurance), the first type of association problem 2 (the protection scope of the second motor insurance), and the like.
  • the knowledge aggregation graph in the question-and-answer knowledge aggregation graph based on the business feature labels (“auto insurance type” and “insurance age”) contained in the initial standard question (the types of auto insurance that can be insured at the age of 18), we can obtain the same data as the initial standard question.
  • the set of the second type of association problem associated with the business feature label for example, the second type of association problem 1, the second type of association problem 2), that is, in the business feature field corresponding to the business feature label, the Collection of questions.
  • the second type of related question set may include: the second type of related question 1 (the scope of auto insurance claims that can be enjoyed by the 18-year-old), and the second type of related question 2 (the method of auto insurance claims that can be enjoyed by the 18-year-old).
  • the above-mentioned intelligent question answering device further includes a graph generation module 413, and the above-mentioned graph generation module 413 includes:
  • a data acquisition unit 4130 configured to acquire a plurality of historical business data from the business database of the above-mentioned target business, and determine a plurality of historical user data and a plurality of customer service reference data in the above-mentioned plurality of historical business data;
  • the user sample data generation unit 4131 is used to determine the standard questions corresponding to each historical business problem in the above-mentioned standard problem database according to the historical business problems associated with the above-mentioned multiple historical user data as sample business problems, and according to the above-mentioned historical business problems
  • the questioning order of each sample business question determines the questioning order of each sample business question
  • the user sample data is determined according to the above-mentioned each sample business question and the questioning order of each of the above-mentioned sample business questions;
  • the customer service sample data generating unit 4132 is configured to determine the standard questions corresponding to each reference business problem in the above standard question base as the customer service sample questions and the services included in the customer service sample questions according to the reference business questions associated with the above-mentioned multiple customer service reference data. Feature labels, customer service sample data is determined according to the above customer service sample questions and the business feature labels included in the above customer service sample questions.
  • the above-mentioned map generation module 413 includes:
  • the question and answer aggregation graph generation unit 4133 is used to determine the nodes of the question and answer aggregation graph in the above question and answer knowledge aggregation graph by using each sample business question in the above user sample data, and determine the above question and answer aggregation graph by the questioning sequence of each of the above sample business questions.
  • the node connection relationship of the above-mentioned question and answer aggregation map is connected according to the node connection relationship of the above-mentioned question and answer aggregation map, so as to obtain the above-mentioned question and answer aggregation map;
  • the knowledge aggregation graph generating unit 4134 is used to determine the nodes of the question and answer aggregation graph in the above question and answer knowledge aggregation graph by using each sample business question in the above user sample data, and determine the above question and answer aggregation graph through the questioning sequence of each of the above sample business questions.
  • the nodes of the above question and answer aggregation graph are connected according to the node connection relation of the above question and answer aggregation graph, so as to obtain the above question and answer aggregation graph.
  • the above-mentioned association aggregation module 403 includes:
  • the closed-loop generation unit 4031 is configured to determine, from the above question and answer aggregation graph, the primary first type of related questions that are semantically related to the above-mentioned initial standard questions, and determine from the above question and answer aggregation graph the semantically related questions of the above-mentioned primary first type of related questions. Secondary first-type related questions, and determine the secondary first-type related questions semantically related to the above-mentioned secondary first-type related questions from the above question and answer aggregation graph, until the secondary first-level related questions of any level.
  • the above initial standard problem occurs in a class of related problems, so as to obtain a closed loop of initial business problems composed of the first related problems at all levels, and based on the first type of related problems at all levels included in the closed loop of the above initial business problems, determine the above initial standard.
  • a set of related questions of the first kind of questions are provided.
  • the above-mentioned association aggregation module 403 includes:
  • the closed-loop confirmation unit 4032 is configured to determine, when the number of the first type of related problems at all levels included in the closed loop of the initial business problem is less than the threshold, the first type of related problems at all levels included in the closed loop of the initial business problem as the above-mentioned initial business problem.
  • the above-mentioned association aggregation module 403 includes:
  • the first type of related question confirmation unit 4033 is configured to determine, based on the connection probability between the first type of related questions at all levels and the above-mentioned initial business question included in the above-mentioned closed loop of the above-mentioned initial business question in the above-mentioned question-and-answer aggregation graph, to determine that the above-mentioned closed-loop of the above-mentioned initial business question contains:
  • the connection probability of the first type of association problem at all levels that appear after the above-mentioned initial standard problem according to the first type of association problem that appears after the above-mentioned initial standard problem in the closed loop of the above-mentioned initial business problem is greater than the threshold, determine the above-mentioned initial standard problem A set of related problems of the first kind.
  • the result output module 404 is configured to determine the initial standard answer of the above-mentioned initial standard question, the above-mentioned first type of related question set and the above-mentioned second type of related question set as the answer text of the above-mentioned initial business question, and the above-mentioned initial business question.
  • the answer text is output to the user interface that initiates the above target service.
  • the result output module 404 can answer the initial standard questions of the initial standard question (the first car insurance, the second car insurance, etc. can be insured at the age of eighteen), the first type of association question set (for example, the first type of association Question 1, the first type of related question 2, etc.) and the second set of related questions (for example, the second type of related question 1, the second type of related question 2, etc.) are determined as the answer text of the initial business question, and the initial business The text of the answer to the question is output to the user interface that launches the target business.
  • the first type of association problem set includes: first type of association problem 1 (the protection scope of the first car insurance), and the first type of association problem 2 (the protection scope of the second car insurance) and the like.
  • the second-type related problem set includes: the second-type related problem 1 (the scope of auto insurance claims that can be enjoyed by the 18-year-old), and the second-type related problem 2 (the auto insurance claim method that can be enjoyed by the 18-year-old).
  • the terminal device obtains the initial business problem associated with the user operation data by acquiring the user operation data, and based on the semantic analysis of the initial business problem, the initial standard problem corresponding to the initial business problem can be determined in the standard problem database , thereby standardizing the initial business question, simplifying the subsequent analysis process for the initial business question, and at the same time determining the initial standard answer corresponding to the initial standard question and the business feature label contained in the initial standard question.
  • the first type of related question sets that are semantically related to the initial standard questions can be determined from the question answering knowledge aggregation graph.
  • the first type of associated question set that is, the set of questions associated with the initial business question in the user's question-and-answer habit.
  • the second type of associated question sets associated with the business feature labels contained in the initial standard questions can be determined from the question-and-answer knowledge aggregation graph.
  • the second type of associated question set that is, a set of questions associated with the initial business question in the business feature field corresponding to the business feature label.
  • the initial standard answer of the initial standard question, the first type of related question set, and the second type of related question set are determined as the answer text of the initial business question and output to the user interface, which can be provided to the user in the initial standard answer corresponding to the initial business question.
  • it provides the first type of related question set that the user may ask next, and the second type of related question set that the user may want to know next, so as to optimize the user experience and improve the question answering efficiency of the intelligent question answering system.
  • FIG. 7 is a schematic structural diagram of a terminal device provided by an embodiment of the present application.
  • the terminal device in this embodiment may include: one or more processors 501 and a memory 502 .
  • the above-mentioned processor 501 and memory 502 are connected through a bus 503 .
  • the memory 502 is used to store a computer program, the computer program includes program instructions, and the processor 501 is used to execute the program instructions stored in the memory 502, and perform the following operations:
  • the terminal device obtains, through the user interface, user operation data for triggering the activation of the target service, and determines the initial service problem associated with the above-mentioned user operation data;
  • the initial standard questions corresponding to the above-mentioned initial business questions, the initial standard answers corresponding to the above-mentioned initial standard questions, and the business feature labels contained in the initial standard questions are determined from the standard question base;
  • the first type of related question sets semantically related to the above-mentioned initial standard questions are determined from the question-and-answer knowledge aggregation graph, and the first type of related question sets that are semantically related to the above-mentioned initial standard questions, as well as those contained in the above-mentioned initial standard questions
  • the above-mentioned processor 501 is further used for:
  • the business problems associated with the above business data are clustered to obtain a plurality of standard questions, and a standard question library is generated according to the above-mentioned plurality of standard questions.
  • the above-mentioned processor 501 is used for:
  • the standard questions corresponding to each historical business question are determined in the above standard question database as sample business questions, and the question order of each of the above historical business questions is determined.
  • the questioning order according to the above-mentioned sample business questions and the questioning order of the above-mentioned sample business questions, to determine the user sample data;
  • the standard questions corresponding to each reference business question are determined in the above standard question database as the customer service sample questions, and the business feature labels included in the customer service sample questions.
  • the service feature labels contained in the above customer service sample questions determine the customer service sample data;
  • a question and answer knowledge aggregation graph is constructed.
  • the above-mentioned processor 501 is used for:
  • the nodes of the question and answer aggregation graph in the above question and answer knowledge aggregation graph are determined by each sample business question in the above user sample data, and the node connection relationship of the above question and answer aggregation graph is determined by the question order of the above sample business questions, and the above question and answer aggregation graph is aggregated.
  • the nodes of the graph are connected according to the node connection relationship of the above question and answer aggregation graph, so as to obtain the above question and answer aggregation graph;
  • the nodes of the knowledge aggregation graph in the above question and answer knowledge aggregation graph are determined by the customer service sample questions in the above customer service sample data, and the node connection relationship of the above knowledge aggregation graph is determined by the business feature labels contained in the above customer service sample questions.
  • the nodes of the aggregation graph are connected according to the node connection relationship of the above question and answer aggregation graph, so as to obtain the above knowledge aggregation graph.
  • the above-mentioned processor 501 is used for:
  • a set of the first type of related problems of the above-mentioned initial standard problem is determined.
  • the processor 501 is configured to: when the number of the first type of related problems at all levels included in the closed loop of the initial business problem is less than a threshold, set the number of the first type of related problems at all levels included in the closed loop of the initial business problem A class of related problems is determined as a first-class related problem set of the above-mentioned initial standard problems.
  • the first type of correlation problem set for determining the above-mentioned initial standard problem based on the first type of correlation problems at all levels included in the closed loop of the above-mentioned initial business problem includes:
  • connection probability between the first type of related questions at all levels included in the closed loop of the initial business questions in the above question and answer aggregation graph and the above initial business questions it is determined that the first type of related questions at all levels included in the closed loop of the above initial business questions appear in the above the connection probability after the initial standard problem;
  • the first type of association problem whose connection probability after the above-mentioned initial standard problem in the above-mentioned closed loop of the above-mentioned initial business problem is greater than the threshold, the first type of the above-mentioned initial standard problem is determined.
  • the above-mentioned processor 501 may be a central processing unit (central processing unit, CPU), and the processor may also be other general-purpose processors, digital signal processors (digital signal processors, DSP), dedicated integrated Circuit (application specific integrated circuit, ASIC), off-the-shelf programmable gate array (field-programmable gate array, FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc.
  • a general purpose processor may be a microprocessor or the processor may be any conventional processor or the like.
  • the memory 502 may include read only memory and random access memory, and provides instructions and data to the processor 501 .
  • a portion of memory 502 may also include non-volatile random access memory.
  • memory 502 may also store device type information.
  • the above-mentioned terminal device can execute the implementation manners provided by the respective steps in the above-mentioned FIG. 1 , FIG. 2 and FIG. 4 through its built-in functional modules.
  • the above-mentioned terminal device can execute the implementation manners provided by the respective steps in the above-mentioned FIG. 1 , FIG. 2 and FIG. 4 through its built-in functional modules.
  • the terminal device obtains the initial business problem associated with the user operation data by acquiring the user operation data, and based on the semantic analysis of the initial business problem, the initial standard problem corresponding to the initial business problem can be determined in the standard problem database , thereby standardizing the initial business question, simplifying the subsequent analysis process for the initial business question, and at the same time determining the initial standard answer corresponding to the initial standard question and the business feature label contained in the initial standard question.
  • the first type of related question sets that are semantically related to the initial standard questions can be determined from the question answering knowledge aggregation graph.
  • the first type of associated question set that is, the set of questions associated with the initial business question in the user's question-and-answer habit.
  • the second type of associated question sets associated with the business feature labels contained in the initial standard questions can be determined from the question-and-answer knowledge aggregation graph.
  • the second type of associated question set that is, a set of questions associated with the initial business question in the business feature field corresponding to the business feature label.
  • the initial standard answer of the initial standard question, the first type of related question set, and the second type of related question set are determined as the answer text of the initial business question and output to the user interface, which can be provided to the user in the initial standard answer corresponding to the initial business question.
  • it provides the first type of related question set that the user may ask next, and the second type of related question set that the user may want to know next, so as to optimize the user experience and improve the question answering efficiency of the intelligent question answering system.
  • Embodiments of the present application further provide a computer-readable storage medium, where a computer program is stored in the computer-readable storage medium, and the computer program includes program instructions, and when the program instructions are executed by a processor, the programs shown in FIGS.
  • a computer program is stored in the computer-readable storage medium, and the computer program includes program instructions, and when the program instructions are executed by a processor, the programs shown in FIGS.
  • the storage medium involved in this application such as a computer-readable storage medium, may be non-volatile or volatile.
  • the above-mentioned computer-readable storage medium may be the apparatus for identifying user behavior based on the prediction model provided in any of the foregoing embodiments or an internal storage unit of the above-mentioned terminal device, such as a hard disk or memory of an electronic device.
  • the computer-readable storage medium can also be an external storage device of the electronic device, such as a pluggable hard disk, a smart media card (SMC), a secure digital (SD) card equipped on the electronic device, Flash card (flash card), etc.
  • the computer-readable storage medium may also include both an internal storage unit of the electronic device and an external storage device.
  • the computer-readable storage medium is used to store the computer program and other programs and data required by the electronic device.
  • the computer-readable storage medium can also be used to temporarily store data that has been or will be output.
  • each process and/or the schematic structural diagrams of the method flowcharts and/or structural schematic diagrams can be implemented by computer program instructions. or blocks, and combinations of processes and/or blocks in flowcharts and/or block diagrams.
  • These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, embedded processor or other programmable data processing device to produce a machine such that the instructions executed by the processor of the computer or other programmable data processing device produce a function
  • These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory result in an article of manufacture comprising instruction means, the instructions
  • the apparatus implements the functions specified in one or more of the flowcharts and/or one or more blocks of the structural diagram.
  • These computer program instructions can also be loaded on a computer or other programmable data processing device to cause a series of operational steps to be performed on the computer or other programmable device to produce a computer-implemented process such that The instructions provide steps for implementing the functions specified in the block or blocks of the flowchart and/or structural representation.

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Data Mining & Analysis (AREA)
  • Databases & Information Systems (AREA)
  • General Engineering & Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Computational Linguistics (AREA)
  • Mathematical Physics (AREA)
  • Information Retrieval, Db Structures And Fs Structures Therefor (AREA)

Abstract

一种智能问答方法、装置、设备及存储介质，包括：终端设备通过用户界面获取用于触发启动目标业务的用户操作数据，并确定初始业务问题（S101）；基于初始业务问题的语义分析，从标准问题库中确定出初始业务问题对应的初始标准问题、初始标准问题对应的初始标准回答以及初始标准问题所包含的业务特征标签（S102）；从问答知识聚合图谱中确定出与初始标准问题的语义关联的第一类关联问题集合，以及与初始标准问题所包含的业务特征标签关联的第二类关联问题集合（S103）；并将初始标准问题的初始标准回答、第一类关联问题集合以及第二关类联问题集合输出至启动目标业务的用户界面（S104）。

Description

智能问答方法、装置、设备及存储介质
本申请要求于2020年12月18日提交中国专利局、申请号为202011511700.9，发明名称为“智能问答方法、装置、设备及存储介质”的中国专利申请的优先权，其全部内容通过引用结合在本申请中。
技术领域
本申请涉及人工智能领域，尤其涉及一种智能问答方法、装置、设备及存储介质。
背景技术
随着人工智能领域技术的发展，越来越多的企业通过线上问答系统为用户提供信息，包括对用户提出的问题进行解答并提供相关问题以供用户进行选择。发明人意识到，其中，相关问题的提出通常依赖于工作人员自身的经验，有经验的工作人员可以更快意识到用户真正想要咨询的问题，及时提供相关问题并给予解答，这依赖于工作人员的经验，对于经验不足的工作人员而言，也有一定的局限性，适用性差。
发明内容
本申请实施例提供一种智能问答方法、装置、设备及存储介质，可提高智能问答系统的问答效率，优化用户体验。
第一方面，本申请实施例供了一种智能问答方法，该方法包括：
终端设备通过用户界面获取用于触发启动目标业务的用户操作数据，并确定上述用户操作数据所关联的初始业务问题；
基于上述初始业务问题的语义分析，从标准问题库中确定出上述初始业务问题对应的初始标准问题、上述初始标准问题对应的初始标准回答以及初始标准问题所包含的业务特征标签；
根据上述初始标准问题以及上述初始标准问题所包含的业务特征标签，从问答知识聚合图谱中确定出与上述初始标准问题的语义关联的第一类关联问题集合，以及与上述初始标准问题所包含的业务特征标签关联的第二类关联问题集合；
将上述初始标准问题的初始标准回答、上述第一类关联问题集合以及上述第二关类联问题集合确定为上述初始业务问题的回答文本，并将上述初始业务问题的回答文本输出至启动上述目标业务的用户界面。
第二方面，本申请实施例提供了一种智能问答装置，该装置包括：
问题获取模块，用于通过用户界面获取用于触发启动目标业务的用户操作数据，并确定上述用户操作数据所关联的初始业务问题；
语义分析模块，用于基于上述初始业务问题的语义分析，从标准问题库中确定出上述初始业务问题对应的初始标准问题、上述初始标准问题对应的初始标准回答以及初始标准问题所包含的业务特征标签；
关联聚合模块，用于根据上述初始标准问题以及上述初始标准问题所包含的业务特征标签，从问答知识聚合图谱中确定出与上述初始标准问题的语义关联的第一类关联问题集合，以及与上述初始标准问题所包含的业务特征标签关联的第二类关联问题集合；
结果输出模块，用于将上述初始标准问题的初始标准回答、上述第一类关联问题集合以及上述第二关类联问题集合确定为上述初始业务问题的回答文本，并将上述初始业务问题的回答文本输出至启动上述目标业务的用户界面。
第三方面，本申请实施例提供了一种终端设备，该终端设备包括处理器和存储器，该处理器和存储器相互连接。该存储器用于存储计算机程序，该计算机程序包括程序指令，该处理器被配置用于调用上述程序指令，执行以下方法：
通过用户界面获取用于触发启动目标业务的用户操作数据，并确定上述用户操作数据所关联的初始业务问题；
基于上述初始业务问题的语义分析，从标准问题库中确定出上述初始业务问题对应的初始标准问题、上述初始标准问题对应的初始标准回答以及初始标准问题所包含的业务特征标签；
根据上述初始标准问题以及上述初始标准问题所包含的业务特征标签，从问答知识聚合图谱中确定出与上述初始标准问题的语义关联的第一类关联问题集合，以及与上述初始标准问题所包含的业务特征标签关联的第二类关联问题集合；
将上述初始标准问题的初始标准回答、上述第一类关联问题集合以及上述第二关类联问题集合确定为上述初始业务问题的回答文本，并将上述初始业务问题的回答文本输出至启动上述目标业务的用户界面。
第四方面，本申请实施例提供了一种计算机可读存储介质，该计算机可读存储介质存储有计算机程序，该计算机程序包括程序指令，该程序指令当被处理器执行时使该处理器执行以下方法：
通过用户界面获取用于触发启动目标业务的用户操作数据，并确定上述用户操作数据所关联的初始业务问题；
基于上述初始业务问题的语义分析，从标准问题库中确定出上述初始业务问题对应的初始标准问题、上述初始标准问题对应的初始标准回答以及初始标准问题所包含的业务特征标签；
根据上述初始标准问题以及上述初始标准问题所包含的业务特征标签，从问答知识聚合图谱中确定出与上述初始标准问题的语义关联的第一类关联问题集合，以及与上述初始标准问题所包含的业务特征标签关联的第二类关联问题集合；
将上述初始标准问题的初始标准回答、上述第一类关联问题集合以及上述第二关类联问题集合确定为上述初始业务问题的回答文本，并将上述初始业务问题的回答文本输出至启动上述目标业务的用户界面。
本申请实施例可优化用户体验，提高智能问答系统的问答效率。
附图说明
图1是本申请实施例提供的智能问答方法的一流程示意图；
图2是本申请实施例提供的构建问答知识聚合图谱的流程示意图；
图3是本申请实施例提供的问答聚合图谱的结构示意图；
图4是本申请实施例提供的智能问答方法的另一流程示意图；
图5是本申请实施例提供的智能问答装置的结构示意图；
图6是本申请实施例提供的智能问答装置的另一结构示意图；
图7是本申请实施例提供的终端设备的结构示意图。
具体实施方式
下面将结合本申请实施例中的附图，对本申请实施例中的技术方案进行描述。
本申请的技术方案可涉及人工智能和/或大数据技术领域，如可具体涉及自然语言处理技术，以实现智能问答，从而推动智慧城市的建设。可选的，本申请涉及的数据如业务问题、标准问题、标准回答、标签和/或回答文本等可存储于数据库中，或者可以存储于区块链中，比如通过区块链分布式存储，本申请不做限定。
目前，利用人工智能技术对某个领域的信息进行智能问答系统的构建，可以更好地推动该领域的技术发展。例如在保险领域，对保险业务信息进行智能问答系统构建，可以帮助人们快速了解某种保险的适用人群，保护范围，理赔范围以及理赔方式等信息。智能问答系统的适用范围非常广泛，本申请仅以在保险领域对保险业务信息的智能问答系统构建作为应用场景进行说明，对其他领域或保险领域的其他信息进行智能问答系统构建与本申请提供的实施例本质相同，在此不再赘述。以保险领域的客服智能问答系统为具体应用场 景为例，当用户在终端设备上的用户界面，通过点击触发目标业务(即智能问答业务)时，终端设备可以获取用户操作数据，并确定用户操作数据所关联的初始业务问题(也即，用户在用户界面输入的初始业务问题)。例如，终端设备获取用户操作数据，并确定初始业务问题(十八岁投保哪种车险比较好？)。随后，终端设备可以对初始业务问题进行语义分析，在标准问题库中确定出初始业务问题对应的初始标准问题(十八岁可投保的车险种类)，并得到初始标准问题(十八岁可投保的车险种类)对应的标准回答(十八岁可投保第一车险、第二车险等)，以及初始标准问题(十八岁可投保的车险种类)所包含的业务特征标签(“车险种类”和“投保年龄”)。终端设备可以将初始标准问题以及初始标准问题所包含的业务特征标签输入至问答知识聚合图谱，通过问答知识聚合图谱中的问答聚合图谱，基于初始标准问题(十八岁可投保的车险种类)得到与初始标准问题的语义关联的第一类关联问题集合(例如，第一类关联问题1，第一类关联问题2)，也即，得到在用户问答习惯中，与初始业务问题相关联的问题的集合。例如，第一类关联问题集合可以包括：第一类关联问题1(第一车险的保护范围)，以及第一类关联问题2(第二车险的保护范围)等。通过问答知识聚合图谱中的知识聚合图谱，基于初始标准问题(十八岁可投保的车险种类)所包含的业务特征标签(“车险种类”和“投保年龄”)得到与初始标准问题所包含的业务特征标签关联的第二类关联问题集合(例如，第二类关联问题1，第二类关联问题2)，也即，在业务特征标签对应的业务特征领域中，与初始业务问题相关联的问题的集合。例如，第二类关联问题集合可以包括：第二类关联问题1(十八岁可享受的车险理赔范围)，以及第二类关联问题2(十八岁可享受的车险理赔方式)等。随后终端设备可以将初始标准问题的初始标准回答、第一类关联问题集合以及第二关类联问题集合确定为初始业务问题的回答文本，并将初始业务问题的回答文本输出至启动目标业务的用户界面。可选地，终端设备可以获取用户在用户界面的后续操作，从而继续智能问答任务或者结束智能问答任务。
请参阅图1，图1是本申请实施例提供的智能问答方法的一流程示意图。本申请实施例提供的方法可包括终端设备通过用户界面获取用于触发启动目标业务的用户操作数据，并确定用户操作数据所关联的初始业务问题；基于初始业务问题的语义分析，从标准问题库中确定出初始业务问题对应的初始标准问题、初始标准问题对应的初始标准回答以及初始标准问题所包含的业务特征标签；根据初始标准问题以及初始标准问题所包含的业务特征标签，从问答知识聚合图谱中确定出与初始标准问题的语义关联的第一类关联问题集合，以及与初始标准问题所包含的业务特征标签关联的第二类关联问题集合；将初始标准问题的初始标准回答、第一类关联问题集合以及第二关类联问题集合确定为初始业务问题的回答文本，并将初始业务问题的回答文本输出至启动目标业务的用户界面。在本申请实施例中，为表述方便，下面将以保险系统的智能问答系统作为智能问答系统，以终端设备为本申请的执行主体，对本申请实施例提供的智能问答方法及装置进行示例说明。
本申请实施例提供的方法可包括如下步骤：
S101：终端设备通过用户界面获取用于触发启动目标业务的用户操作数据，并确定用户操作数据所关联的初始业务问题。
在一些可行的实施方式中，当用户在终端设备上的用户界面，通过点击触发目标业务(即智能问答业务)时，终端设备可以获取用户操作数据，并确定用户操作数据所关联的初始业务问题(也即，用户在用户界面输入的初始业务问题)。例如，终端设备可以获取用户点击触发智能问答业务以及输入问题的用户操作数据，确定初始业务问题。初始业务问题为：十八岁投保哪种车险比较好？具体可根据实际应用场景确定，在此不做限制。
S102：基于初始业务问题的语义分析，从标准问题库中确定出初始业务问题对应的初始标准问题、初始标准问题对应的初始标准回答以及初始标准问题所包含的业务特征标签。
在一些可行的实施方式中，终端设备可以对初始业务问题进行语义分析，在标准问题 库中确定出初始业务问题对应的初始标准问题(十八岁可投保的车险种类)，并得到初始标准问题(十八岁可投保的车险种类)对应的标准回答(十八岁可投保第一车险、第二车险等)，以及初始标准问题(十八岁可投保的车险种类)所包含的业务特征标签(“车险种类”和“投保年龄”)。其中，标准问题库由多个标准问题构建得到，标准问题库中的每个标准问题都具有自己的业务特征标签。
在一些可行的实施方式中，终端设备可以从目标业务的业务数据库中获取多个业务数据，确定各业务数据所关联的业务问题。之后，终端设备可以对这些业务问题进行语义分析，将各业务问题按照问题意图进行聚类，也即，将问题意图相同但提问方法不同的业务问题归为一类标准问题，从而得到多个标准问题。在得到多个标准问题之后，终端设备可以基于业务数据确定各标准问题对应的标准回答，同时对这些标准问题进行关键词提取，得到各标准问题包含的业务特征标签，从而生成标准问题库。其中，业务特征标签可以基于业务数据关联的业务问题所对应的关键词归类得到，也可以基于业务数据库所属的业务领域(例如，保险领域)的业务特征(保险种类、投保年龄、理赔范围、理赔方式等)得到。从而可将业务问题规范化，简化后续问答过程中针对用户操作数据关联的初始业务问题进行分析的过程，提高问答效率。
在一些可行的实施方式中，终端设备可以将各业务数据进行标注，比如业务数据1(问：十八岁投保哪种车险比较好，答：十八岁可投保的车险包括第一车险、第二车险等)和业务数据2(问：十八岁可投保的车险种类，答：十八岁可投保第一车险、第二车险等)进行标注，得到标注后的业务数据1(十八岁投保哪种车险比较好//车险种类/投保年龄//十八岁可投保的车险包括第一车险、第二车险，等)以及标注后的业务数据2(十八岁可投保的车险种类//车险种类/投保年龄//十八岁可投保第一车险、第二车险等)。终端设备可以对标注后的业务数据1和标注后的业务数据2进行语义识别，得到标注后的业务数据1的问题意图(十八岁可投保的车险)以及标注后的业务数据2的问题意图(十八岁可投保的车险)，将标注后的业务数据1和标注后的业务数据2聚类为统一类别，并选取出现频率最高的标注后的业务数据2的问题部分作为此类问题意图(十八岁可投保的车险)的标准问题(十八岁可投保的车险种类)，同时将标注后的业务数据2的业务特征标签部分作为该标准问题(十八岁可投保的车险种类)的业务特征标签(车险种类、投保年龄)，并标注后的业务数据2的回答部分作为该标准问题(十八岁可投保的车险种类)对应的标准回答(十八岁可投保第一车险、第二车险等)。
S103：根据初始标准问题以及初始标准问题所包含的业务特征标签，从问答知识聚合图谱中确定出与初始标准问题的语义关联的第一类关联问题集合，以及与初始标准问题所包含的业务特征标签关联的第二类关联问题集合。
在一些可行的实施方式中，终端设备可以将初始标准问题以及初始标准问题所包含的业务特征标签输入至问答知识聚合图谱，通过问答知识聚合图谱中的问答聚合图谱，基于初始标准问题(十八岁可投保的车险种类)得到与初始标准问题的语义关联的第一类关联问题集合(例如，第一类关联问题1，第一类关联问题2)，也即，得到在用户问答习惯中，与初始业务问题相关联的问题的集合。例如，第一类关联问题集合可以包括：第一类关联问题1(第一车险的保护范围)，以及第一类关联问题2(第二车险的保护范围)等。通过问答知识聚合图谱中的知识聚合图谱，基于初始标准问题(十八岁可投保的车险种类)所包含的业务特征标签(“车险种类”和“投保年龄”)得到与初始标准问题所包含的业务特征标签关联的第二类关联问题集合(例如，第二类关联问题1，第二类关联问题2)，也即，在业务特征标签对应的业务特征领域中，与初始业务问题相关联的问题的集合。例如，第二类关联问题集合可以包括：第二类关联问题1(十八岁可享受的车险理赔范围)，以及第二类关联问题2(十八岁可享受的车险理赔方式)等。
在一些可行的实施方式中，请一并参阅图2，图2是本申请实施例提供的构建问答知识聚合图谱的流程示意图。上述构建问答知识聚合图谱的方法可包括如下步骤S201至S205中各个步骤所提供的实现方式。
S201：从目标业务的业务数据库中获取多个历史业务数据，并确定多个历史业务数据中的多个历史用户数据和多个客服参考数据。
S202：根据多个历史用户数据关联的历史业务问题，在标准问题库中确定出各历史业务问题对应的标准问题作为样本业务问题，并根据各历史业务问题的提问顺序确定各样本业务问题的提问顺序，根据各样本业务问题和各样本业务问题的提问顺序确定用户样本数据。
S203：根据多个客服参考数据关联的参考业务问题，在标准问题库中确定出各参考业务问题对应标准问题作为客服样本问题，以及客服样本问题所包含的业务特征标签，根据客服样本问题和客服样本问题所包含的业务特征标签确定客服样本数据。
在一些可行的实施方式中，终端设备可以在目标业务的业务数据库中获取多个历史用户数据和多个客服参考数据。终端设备可以基于历史用户数据得到符合用户提问习惯的样本业务问题(也即，历史业务问题在标准问题库中对应得到的样本业务问题)。样本业务问题有着基于用户提问习惯的提问顺序，按照提问顺序，存在语义关联的历史业务问题通常可以构成一个有向的闭环。终端设备可以基于客服参考数据得到符合领域知识的客服样本问题(也即，参考业务问题再标准问题库中对应得到的客服样本问题)。其中，客服样本问题可以是相关领域客服提供的容易被用户忽略，但用户可能需要知道的问题，存在关联的客服样本问题通常具有相同或者相似的业务特征标签。
S204：通过用户样本数据中的各样本业务问题确定出问答知识聚合图谱中的问答聚合图谱的节点，通过各样本业务问题的提问顺序确定出问答聚合图谱的节点连接关系，将问答聚合图谱的节点按照问答聚合图谱的节点连接关系进行连接，以得到问答聚合图谱。
在一些可行的实施方式中，终端设备利用用户样本数据对问答聚合图谱进行构建，可以在问答聚合图谱中构建以各样本业务问题为节点，以各样本业务问题的提问顺序为边的有向图。也即，将一个用户样本数据中的一个样本业务问题作为v i，将该用户样本数据中用户在v i之后提出的下一个样本业务问题作为v i的一个第一类关联问题v i1。将所有用户样本数据中用户在v i之后提出的下一个业务样本问题作为v i的第一类关联问题，以得到业务样本问题v i的第一类关联问题集合V i＝{v ij，i∈N,j∈N}。在问答聚合图谱中，以业务样本问题v i和V i中的所有第一类关联问题v ij为节点，以业务样本问题v i和业务样本问题v i的第一类关联问题v ij的提问顺序为边构建有向图。其中，各个边以业务样本问题v i为起点，以业务样本问题v i的各第一类关联问题为终点，从而使得终端设备可以根据构建后的问答聚合图谱，得到业务样本问题的第一类关联问题集合(也即，按照提问顺序，用户接下来可能会想要提问的问题)。
在一些可行的方式中，对于业务样本问题v i的第一类关联问题v ij，可以将用户样本数据中v i与v ij先后出现次数记录为C ij，由此得到从业务样本问题v i指向第一类关联问题v ij的边<v i,v ij>的连接概率w ij。
其中，
Figure PCTCN2021090197-appb-000001
在一些可行的方式中，此时得到的问答知识聚合图谱对于新增或者冷门的知识不够友好，难以找出较少或没有访问记录的用户样本数据之间的关联性。终端设备可以在获取多组样本信息的过程中，对每一个业务样本问题构造该业务样本问题的隐向量，利用每一个业务样本问题对应的各第一类关联问题的连接概率构建问答聚合图谱，从而可以通过冷门业务样本问题的隐向量得到冷门业务样本问题的第一类关联问题。
具体地，构造业务样本问题v i的隐向量h i，以及第一类关联问题v ij的隐向量h ij，并根据从业务样本问题v i指向第一类关联问题v ij的边<v i,v ij>的连接概率构建问答聚合图谱，使得问答聚合图谱的损失函数达到最小，并基于冷门业务样本问题的隐向量，得到冷门业务样本问题的第一类关联问题。
在一些可行的实施方式中，请一并参阅图3，图3是本申请实施例提供的问答聚合图谱的结构示意图。因为样本业务问题都属于标准问题，本质上问答聚合图谱中的所有节点都可以用标准问题进行表述，为描述方便，将以标准问题来表示问答聚合图谱的节点。
如图3所示，A1、A2、A3、A4和A5分别代表标准问题A1、标准问题A2、标准问题A3、标准问题A4以及标准问题A5。标准问题A2和标准问题A3是标准问题A1的初级第一类关联问题。标准问题A3是标准问题A2的次级第一类关联问题，标准问题A4是标准问题A3的次级第一类关联问题。标准问题A5和标准问题A1是标准问题A4的次级第一类关联问题。标准问题A1是标准问题A5的次级第一类关联问题。其中，标准问题A1、标准问题A3和标准问题A4构成了初始标准问题闭环1(包含3个标准问题)。标准问题A1、标准问题A2、标准问题A3和标准问题A4构成了初始标准问题闭环2(包含4个标准问题)。标准问题A1、标准问题A3、标准问题A4和标准问题A5构成了初始标准问题闭环3(包含4个标准问题)。标准问题A1、标准问题A2、标准问题A3、标准问题A4和标准问题A5构成了初始标准问题闭环4(包含5个标准问题)。
S205：通过客服样本数据中的客服样本问题确定出问答知识聚合图谱中的知识聚合图谱的节点，通过客服样本问题所包含的业务特征标签确定出知识聚合图谱的节点连接关系，将问答聚合图谱的节点按照问答聚合图谱的节点连接关系进行连接，以得到知识聚合图谱。
在一些可行的实施方式中，终端设备利用客服样本数据对知识聚合图谱进行训练，可以在知识聚合图谱中构建以各客服样本问题为节点，以各客服样本问题包含的业务特征标签为边的无向图。也即，将一个客服样本数据中的一个客服样本问题作为u i，客服样本问题u i的业务特征标签为A i，其中，A i包括多个子业务特征标签a ij。将所有客服样本数据中包括子业务特征标签a ij的客服样本问题作为u i的第二类关联问题u ij，以得到客服样本问题u i的第二类关联问题集合U i＝{u ij，i∈N,j∈N}。在知识聚合图谱中，以客服样本问题u i和U i中的所有第二类关联问题u ij为节点，以客服样本问题u i和客服样本问题u i的第二类关联问题u ij的连接为边构建无向图。其中，边的一个端点是客服样本问题u i，另一个端点是客服样本问题u i的第二类关联问题u ij。从而使得终端设备可以通过训练后的知识聚合图谱，基于客服样本问题以及客服样本问题包含的业务特征标签，得到目标客服样本问题的第二类关联问题集合(按照客服样本问题包含的业务特征标签，用户接下来可能会想要提问的 问题)。
由此终端设备可以根据各样本业务问题和各样本业务问题的提问顺序，构建出问答聚合图谱，也即，基于用户问答习惯的业务问题数据构建出问答聚合图谱。终端设备可以根据客服样本问题和客服样本问题所包含的业务特征标签，构建出知识聚合图谱，也即，基于各业务特征领域内的业务问题数据构建出知识聚合图谱。使得终端设备可以通过问答知识聚合图谱，提供给用户接下来可能会提问的第一类关联问题集合，以及用户接下来可能要知道的第二类关联问题集合，从而优化用户体验，提高智能问答系统的问答效率。
S104：将初始标准问题的初始标准回答、第一类关联问题集合以及第二关类联问题集合确定为初始业务问题的回答文本，并将初始业务问题的回答文本输出至启动目标业务的用户界面。
在一些可行的实施方式中，终端设备可以将初始标准问题的初始标准回答(十八岁可投保第一车险、第二车险等)、第一类关联问题集合(例如，第一类关联问题1，第一类关联问题2等)以及第二关类联问题集合(例如，第二类关联问题1，第二类关联问题2等)确定为初始业务问题的回答文本，并将初始业务问题的回答文本输出至启动目标业务的用户界面。其中，第一类关联问题集合包括：第一类关联问题1(第一车险的保护范围)，以及第一类关联问题2(第二车险的保护范围)等。第二类关联问题集合包括：第二类关联问题1(十八岁可享受的车险理赔范围)，以及第二类关联问题2(十八岁可享受的车险理赔方式)等。
可选地，终端设备可以获取用户在用户界面的后续操作，从而继续智能问答任务或者结束智能问答任务。
在本申请实施例中，终端设备通过获取用户操作数据得到用户操作数据所关联的初始业务问题，基于对初始业务问题的语义分析，可以在标准问题库中确定出初始业务问题对应的初始标准问题，从而将初始业务问题规范化，简化后续针对初始业务问题进行分析的过程，同时确定出初始标准问题对应的初始标准回答以及初始标准问题所包含的业务特征标签。根据初始标准问题，可以从问答知识聚合图谱中确定出与初始标准问题的语义关联的第一类关联问题集合。这里，第一类关联问题集合，也即，在用户问答习惯中，与初始业务问题相关联的问题的集合。根据初始标准问题以及初始标准问题所包含的业务特征标签，可以从问答知识聚合图谱中确定出与初始标准问题所包含的业务特征标签关联的第二类关联问题集合。这里，第二类关联问题集合，也即，在业务特征标签对应的业务特征领域中，与初始业务问题相关联的问题的集合。将初始标准问题的初始标准回答、第一类关联问题集合以及第二关类联问题集合确定为初始业务问题的回答文本并输出至用户界面，可以在提供给用户初始业务问题对应的初始标准回答的同时，提供给用户接下来可能会提问的第一类关联问题集合，以及用户接下来可能要知道的第二类关联问题集合，从而优化用户体验，提高智能问答系统的问答效率。
请参阅图4，图4是本申请实施例提供的智能问答方法的另一流程示意图。
S301：终端设备通过用户界面获取用于触发启动目标业务的用户操作数据，并确定用户操作数据所关联的初始业务问题。
在一些可行的实施方式中，当用户在终端设备上的用户界面，通过点击触发目标业务(即智能问答业务)时，终端设备可以获取用户操作数据，并确定用户操作数据所关联的初始业务问题(也即，用户在用户界面输入的初始业务问题)。例如，终端设备可以获取用户点击触发智能问答业务以及输入问题的用户操作数据，确定初始业务问题。初始业务问题为：十八岁投保哪种车险比较好？具体可根据实际应用场景确定，在此不做限制。
S302：基于初始业务问题的语义分析，从标准问题库中确定出初始业务问题对应的初始标准问题、初始标准问题对应的初始标准回答以及初始标准问题所包含的业务特征标签。
在一些可行的实施方式中，终端设备可以对初始业务问题进行语义分析，在标准问题库中确定出初始业务问题对应的初始标准问题(十八岁可投保的车险种类)，并得到初始标准问题(十八岁可投保的车险种类)对应的标准回答(十八岁可投保第一车险、第二车险等)，以及初始标准问题(十八岁可投保的车险种类)所包含的业务特征标签(“车险种类”和“投保年龄”)。其中，标准问题库由多个标准问题构建得到，标准问题库中的每个标准问题都具有自己的业务特征标签。
S303：从问答聚合图谱中确定出与初始标准问题的语义关联的初级第一类关联问题。
S304：从问答聚合图谱中确定出与初级第一类关联问题语义关联的次级第一类关联问题，并从问答聚合图谱中确定出与次级第一类关联问题语义关联的次级第一类关联问题，直至任一级第一类关联问题的次级第一类关联问题中出现初始标准问题，以得到由各级第一关联问题组成的初始业务问题闭环。
因为初始标准问题、以及初始标准问题的各级第一类关联问题都属于标准问题，本质上问答聚合图谱中的所有节点都可以用标准问题进行表述，为描述方便，将以标准问题来表示问答聚合图谱的节点，以初始业务问题闭环作为前文中初始标准问题闭环的等同表示。
如图3所示，A1、A2、A3、A4和A5分别代表标准问题A1、标准问题A2、标准问题A3、标准问题A4以及标准问题A5。终端设备可以从问答聚合图谱中确定出与初始标准问题(标准问题A1)的语义关联的初级第一类关联问题(标准问题A2、标准问题A3)。终端设备可以从问答聚合图谱中确定出与初级第一类关联问题(标准问题A2、标准问题A3)的语义关联的次级第一类关联问题(标准问题A3、标准问题A4)，并从问答聚合图谱中确定出与次级第一类关联问题(标准问题A3、标准问题A4)语义关联的次级第一类关联问题(标准问题A4、标准问题A5)，直至任一级第一类关联问题(标准问题A4、标准问题A5)的次级第一类关联问题中出现初始标准问题，以得到由各级第一关联问题组成的初始业务问题闭环。也即，标准问题A1、标准问题A3和标准问题A4构成的初始业务问题闭环1(包含2个第一类关联问题)。标准问题A1、标准问题A2、标准问题A3和标准问题A4构成的初始业务问题闭环2(包含3个第一类关联问题)。标准问题A1、标准问题A3、标准问题A4和标准问题A5构成的初始业务问题闭环3(包含3个第一类关联问题)。标准问题A1、标准问题A2、标准问题A3、标准问题A4和标准问题A5构成的初始业务问题闭环4(包含4个第一类关联问题)。
在一些可行的实施方式中，终端设备可以基于问答聚合图谱中初始业务问题闭环中包含的各级第一类关联问题与初始业务问题的连接概率，确定出初始业务问题闭环中包含的各级第一类关联问题出现在初始标准问题之后的连接概率。根据初始业务问题闭环中出现在初始标准问题之后的连接概率大于阈值的第一类关联问题，确定初始标准问题的第一类关联问题集合。
在一些可行的实施方式中，终端设备可以将初始标准问题以及初始标准问题所包含的业务特征标签输入至问答知识聚合图谱，通过问答知识聚合图谱中的问答聚合图谱，基于初始标准问题(十八岁可投保的车险种类)得到与初始标准问题的语义关联的第一类关联问题集合(例如，第一类关联问题1，第一类关联问题2)，也即，得到在用户问答习惯中，与初始业务问题相关联的问题的集合。例如，第一类关联问题集合包括：第一类关联问题1(第一车险的保护范围)，以及第一类关联问题2(第二车险的保护范围)等。
S305：基于初始业务问题闭环中包含的各级第一类关联问题数量，确定出初始标准问题的第一类关联问题集合。
在一些可行的实施方式中，为了避免过度关联，终端设备可以在初始业务问题闭环中包含的各级第一类关联问题的个数小于阈值时，将初始业务问题闭环中包含的各级第一类关联问题确定为初始标准问题的第一类关联问题集合。例如，终端设备可以将初始业务问 题闭环中包含的各级第一类关联问题的个数的阈值设置为3。由前文可知，只有初始业务问题闭环1中的第一类关联问题数量满足条件，故而将初始业务问题闭环1中包含的第一类关联问题确定为初始标准问题的第一类关联问题集合(标准问题A3、标准问题A4)。
在一些可行的实施方式中，终端设备可以基于问答聚合图谱中初始业务问题闭环中包含的各级第一类关联问题与初始业务问题的连接概率，确定出初始业务问题闭环中包含的各级第一类关联问题出现在初始标准问题之后的连接概率，并根据初始业务问题闭环中出现在初始标准问题之后的连接概率大于阈值的第一类关联问题，确定初始标准问题的第一类关联问题集合。
由此终端设备可以从问答聚合图谱中确定出与初始标准问题的语义关联的各级第一类关联问题，并得到初始业务问题闭环。终端设备可以通过初始问题闭环中包含的问题数量对初始业务问题闭环中的各级第一类关联问题进行筛选，也可以通过连接概率对初始业务问题闭环中的各级第一类关联问题进行筛选，将与初始标准问题语义关联程度过低的第一类关联问题剔除，避免过度关联，提高智能问答系统的问答效率。
S306：根据初始标准问题以及初始标准问题所包含的业务特征标签，从知识聚合图谱中确定出与初始标准问题所包含的业务特征标签关联的第二类关联问题集合。
在一些可行的实施方式中，终端设备可以通过问答知识聚合图谱中的知识聚合图谱，基于初始标准问题(十八岁可投保的车险种类)所包含的业务特征标签(“车险种类”和“投保年龄”)得到与初始标准问题所包含的业务特征标签关联的第二类关联问题集合(例如，第二类关联问题1，第二类关联问题2)，也即，在业务特征标签对应的业务特征领域中，与初始业务问题相关联的问题的集合。例如，第二类关联问题集合包括：第二类关联问题1(十八岁可享受的车险理赔范围)，以及第二类关联问题2(十八岁可享受的车险理赔方式)等。
S307：将初始标准问题的初始标准回答、第一类关联问题集合以及第二关类联问题集合确定为初始业务问题的回答文本，并将初始业务问题的回答文本输出至启动目标业务的用户界面。
在一些可行的实施方式中，终端设备可以将初始标准问题的初始标准回答(十八岁可投保第一车险、第二车险等)、第一类关联问题集合(例如，第一类关联问题1，第一类关联问题2等)以及第二关类联问题集合(例如，第二类关联问题1，第二类关联问题2等)确定为初始业务问题的回答文本，并将初始业务问题的回答文本输出至启动目标业务的用户界面。其中，第一类关联问题集合包括：第一类关联问题1(第一车险的保护范围)，以及第一类关联问题2(第二车险的保护范围)等。第二类关联问题集合包括：第二类关联问题1(十八岁可享受的车险理赔范围)，以及第二类关联问题2(十八岁可享受的车险理赔方式)等。
可选地，终端设备可以获取用户在用户界面的后续操作，从而继续智能问答任务或者结束智能问答任务。
在本申请实施例中，终端设备可以从问答聚合图谱中确定出与初始标准问题的语义关联的各级第一类关联问题，并得到初始业务问题闭环。终端设备可以通过初始问题闭环中包含的问题数量对初始业务问题闭环中的各级第一类关联问题进行筛选，也可以通过连接概率对初始业务问题闭环中的各级第一类关联问题进行筛选，将与初始标准问题语义关联程度过低的第一类关联问题剔除，避免过度关联，提高智能问答系统的问答效率。
请参阅图5，图5是本申请实施例提供的智能问答装置的结构示意图，该装置包括：
问题获取模块401，用于通过用户界面获取用于触发启动目标业务的用户操作数据，并确定上述用户操作数据所关联的初始业务问题。
在一些可行的实施方式中，当用户在问题获取模块401上的用户界面，通过点击触发 目标业务(即智能问答业务)时，问题获取模块401可以获取用户操作数据，并确定用户操作数据所关联的初始业务问题(也即，用户在用户界面输入的初始业务问题)。例如，问题获取模块401可以获取用户点击触发智能问答业务以及输入问题的用户操作数据，确定初始业务问题。初始业务问题为：十八岁投保哪种车险比较好？具体可根据实际应用场景确定，在此不做限制。
语义分析模块402，用于基于上述初始业务问题的语义分析，从标准问题库中确定出上述初始业务问题对应的初始标准问题、上述初始标准问题对应的初始标准回答以及初始标准问题所包含的业务特征标签。
在一些可行的实施方式中，语义分析模块402可以对初始业务问题进行语义分析，在标准问题库中确定出初始业务问题对应的初始标准问题(十八岁可投保的车险种类)，并得到初始标准问题(十八岁可投保的车险种类)对应的标准回答(十八岁可投保第一车险、第二车险等)，以及初始标准问题(十八岁可投保的车险种类)所包含的业务特征标签(“车险种类”和“投保年龄”)。其中，标准问题库由多个标准问题构建得到，标准问题库中的每个标准问题都具有自己的业务特征标签。
在一些可行的实施方式中，请参阅图6，图6是本申请实施例提供的智能问答装置的另一结构示意图，上述智能问答装置还包括：
标准问题库生成模块412，用于从上述目标业务的业务数据库中获取多个业务数据，并确定各业务数据所关联的业务问题，基于语义分析将上述各业务数据所关联的业务问题进行聚类得到多个标准问题，根据上述多个标准问题生成标准问题库。
关联聚合模块403，用于根据上述初始标准问题以及上述初始标准问题所包含的业务特征标签，从问答知识聚合图谱中确定出与上述初始标准问题的语义关联的第一类关联问题集合，以及与上述初始标准问题所包含的业务特征标签关联的第二类关联问题集合。
在一些可行的实施方式中，关联聚合模块403可以将初始标准问题以及初始标准问题所包含的业务特征标签输入至问答知识聚合图谱，通过问答知识聚合图谱中的问答聚合图谱，基于初始标准问题(十八岁可投保的车险种类)得到与初始标准问题的语义关联的第一类关联问题集合(例如，第一类关联问题1，第一类关联问题2)，也即，得到在用户问答习惯中，与初始业务问题相关联的问题的集合。例如，第一类关联问题集合可以包括：第一类关联问题1(第一车险的保护范围)，以及第一类关联问题2(第二车险的保护范围)等。通过问答知识聚合图谱中的知识聚合图谱，基于初始标准问题(十八岁可投保的车险种类)所包含的业务特征标签(“车险种类”和“投保年龄”)得到与初始标准问题所包含的业务特征标签关联的第二类关联问题集合(例如，第二类关联问题1，第二类关联问题2)，也即，在业务特征标签对应的业务特征领域中，与初始业务问题相关联的问题的集合。例如，第二类关联问题集合可以包括：第二类关联问题1(十八岁可享受的车险理赔范围)，以及第二类关联问题2(十八岁可享受的车险理赔方式)等。
在一些可行的实施方式中，如图6所示，上述智能问答装置还包括图谱生成模块413，上述图谱生成模块413包括：
数据获取单元4130，用于从上述目标业务的业务数据库中获取多个历史业务数据，并确定上述多个历史业务数据中的多个历史用户数据和多个客服参考数据；
用户样本数据生成单元4131，用于根据上述多个历史用户数据关联的历史业务问题，在上述标准问题库中确定出各历史业务问题对应的标准问题作为样本业务问题，并根据上述各历史业务问题的提问顺序确定各样本业务问题的提问顺序，根据上述各样本业务问题和上述各样本业务问题的提问顺序确定用户样本数据；
客服样本数据生成单元4132，用于根据上述多个客服参考数据关联的参考业务问题，在上述标准问题库中确定出各参考业务问题对应标准问题作为客服样本问题，以及客服样 本问题所包含的业务特征标签，根据上述客服样本问题和上述客服样本问题所包含的业务特征标签确定客服样本数据。
在一些可行的实施方式中，如图6所示，上述图谱生成模块413包括：
问答聚合图谱生成单元4133，用于通过上述用户样本数据中的各样本业务问题确定出上述问答知识聚合图谱中的问答聚合图谱的节点，通过上述各样本业务问题的提问顺序确定出上述问答聚合图谱的节点连接关系，将上述问答聚合图谱的节点按照上述问答聚合图谱的节点连接关系进行连接，以得到上述问答聚合图谱；
知识聚合图谱生成单元4134，用于通过上述用户样本数据中的各样本业务问题确定出上述问答知识聚合图谱中的问答聚合图谱的节点，通过上述各样本业务问题的提问顺序确定出上述问答聚合图谱的节点连接关系，将上述问答聚合图谱的节点按照上述问答聚合图谱的节点连接关系进行连接，以得到上述问答聚合图谱。
在一些可行的实施方式中，如图6所示，上述关联聚合模块403包括：
闭环生成单元4031，用于从上述问答聚合图谱中确定出与上述初始标准问题的语义关联的初级第一类关联问题，从上述问答聚合图谱中确定出与上述初级第一类关联问题语义关联的次级第一类关联问题，并从上述问答聚合图谱中确定出与上述次级第一类关联问题语义关联的次级第一类关联问题，直至任一级第一类关联问题的次级第一类关联问题中出现上述初始标准问题，以得到由各级第一关联问题组成的初始业务问题闭环，并基于上述初始业务问题闭环中包含的各级第一类关联问题，确定出上述初始标准问题的第一类关联问题集合。
在一些可行的实施方式中，如图6所示，上述关联聚合模块403包括：
闭环确认单元4032，用于当上述初始业务问题闭环中包含的各级第一类关联问题的个数小于阈值时，将上述初始业务问题闭环中包含的各级第一类关联问题确定为上述初始标准问题的第一类关联问题集合。
在一些可行的实施方式中，如图6所示，上述关联聚合模块403包括：
第一类关联问题确认单元4033，用于基于上述问答聚合图谱中上述初始业务问题闭环中包含的各级第一类关联问题与上述初始业务问题的连接概率，确定出上述初始业务问题闭环中包含的各级第一类关联问题出现在上述初始标准问题之后的连接概率，根据上述初始业务问题闭环中出现在上述初始标准问题之后的连接概率大于阈值的第一类关联问题，确定上述初始标准问题的第一类关联问题集合。
结果输出模块404，用于将上述初始标准问题的初始标准回答、上述第一类关联问题集合以及上述第二关类联问题集合确定为上述初始业务问题的回答文本，并将上述初始业务问题的回答文本输出至启动上述目标业务的用户界面。
在一些可行的实施方式中，结果输出模块404可以将初始标准问题的初始标准回答(十八岁可投保第一车险、第二车险等)、第一类关联问题集合(例如，第一类关联问题1，第一类关联问题2等)以及第二关类联问题集合(例如，第二类关联问题1，第二类关联问题2等)确定为初始业务问题的回答文本，并将初始业务问题的回答文本输出至启动目标业务的用户界面。其中，第一类关联问题集合包括：第一类关联问题1(第一车险的保护范围)，以及第一类关联问题2(第二车险的保护范围)等。第二类关联问题集合包括：第二类关联问题1(十八岁可享受的车险理赔范围)，以及第二类关联问题2(十八岁可享受的车险理赔方式)等。
在本申请实施例中，终端设备通过获取用户操作数据得到用户操作数据所关联的初始业务问题，基于对初始业务问题的语义分析，可以在标准问题库中确定出初始业务问题对应的初始标准问题，从而将初始业务问题规范化，简化后续针对初始业务问题进行分析的过程，同时确定出初始标准问题对应的初始标准回答以及初始标准问题所包含的业务特征 标签。根据初始标准问题，可以从问答知识聚合图谱中确定出与初始标准问题的语义关联的第一类关联问题集合。这里，第一类关联问题集合，也即，在用户问答习惯中，与初始业务问题相关联的问题的集合。根据初始标准问题以及初始标准问题所包含的业务特征标签，可以从问答知识聚合图谱中确定出与初始标准问题所包含的业务特征标签关联的第二类关联问题集合。这里，第二类关联问题集合，也即，在业务特征标签对应的业务特征领域中，与初始业务问题相关联的问题的集合。将初始标准问题的初始标准回答、第一类关联问题集合以及第二关类联问题集合确定为初始业务问题的回答文本并输出至用户界面，可以在提供给用户初始业务问题对应的初始标准回答的同时，提供给用户接下来可能会提问的第一类关联问题集合，以及用户接下来可能要知道的第二类关联问题集合，从而优化用户体验，提高智能问答系统的问答效率。
参见图7，图7是本申请实施例提供的终端设备的结构示意图。如图7所示，本实施例中的终端设备可以包括：一个或多个处理器501和存储器502。上述处理器501和存储器502通过总线503连接。存储器502用于存储计算机程序，该计算机程序包括程序指令，处理器501用于执行存储器502存储的程序指令，执行如下操作：
终端设备通过用户界面获取用于触发启动目标业务的用户操作数据，并确定上述用户操作数据所关联的初始业务问题；
基于上述初始业务问题的语义分析，从标准问题库中确定出上述初始业务问题对应的初始标准问题、上述初始标准问题对应的初始标准回答以及初始标准问题所包含的业务特征标签；
根据上述初始标准问题以及上述初始标准问题所包含的业务特征标签，从问答知识聚合图谱中确定出与上述初始标准问题的语义关联的第一类关联问题集合，以及与上述初始标准问题所包含的业务特征标签关联的第二类关联问题集合；
将上述初始标准问题的初始标准回答、上述第一类关联问题集合以及上述第二关类联问题集合确定为上述初始业务问题的回答文本，并将上述初始业务问题的回答文本输出至启动上述目标业务的用户界面。
在一些可行的实施方式中，上述处理器501还用于：
从上述目标业务的业务数据库中获取多个业务数据，并确定各业务数据所关联的业务问题；
基于语义分析将上述各业务数据所关联的业务问题进行聚类得到多个标准问题，根据上述多个标准问题生成标准问题库。
在一些可行的实施方式中，上述处理器501用于：
从上述目标业务的业务数据库中获取多个历史业务数据，并确定上述多个历史业务数据中的多个历史用户数据和多个客服参考数据；
根据上述多个历史用户数据关联的历史业务问题，在上述标准问题库中确定出各历史业务问题对应的标准问题作为样本业务问题，并根据上述各历史业务问题的提问顺序确定各样本业务问题的提问顺序，根据上述各样本业务问题和上述各样本业务问题的提问顺序确定用户样本数据；
根据上述多个客服参考数据关联的参考业务问题，在上述标准问题库中确定出各参考业务问题对应标准问题作为客服样本问题，以及客服样本问题所包含的业务特征标签，根据上述客服样本问题和上述客服样本问题所包含的业务特征标签确定客服样本数据；
根据上述用户样本数据和上述客服样本数据构建问答知识聚合图谱。
在一些可行的实施方式中，上述处理器501用于：
通过上述用户样本数据中的各样本业务问题确定出上述问答知识聚合图谱中的问答聚合图谱的节点，通过上述各样本业务问题的提问顺序确定出上述问答聚合图谱的节点连接 关系，将上述问答聚合图谱的节点按照上述问答聚合图谱的节点连接关系进行连接，以得到上述问答聚合图谱；
通过上述客服样本数据中的客服样本问题确定出上述问答知识聚合图谱中的知识聚合图谱的节点，通过上述客服样本问题所包含的业务特征标签确定出上述知识聚合图谱的节点连接关系，将上述问答聚合图谱的节点按照上述问答聚合图谱的节点连接关系进行连接，以得到上述知识聚合图谱。
在一些可行的实施方式中，上述处理器501用于：
从上述问答聚合图谱中确定出与上述初始标准问题的语义关联的初级第一类关联问题；
从上述问答聚合图谱中确定出与上述初级第一类关联问题语义关联的次级第一类关联问题，并从上述问答聚合图谱中确定出与上述次级第一类关联问题语义关联的次级第一类关联问题，直至任一级第一类关联问题的次级第一类关联问题中出现上述初始标准问题，以得到由各级第一关联问题组成的初始业务问题闭环；
基于上述初始业务问题闭环中包含的各级第一类关联问题，确定出上述初始标准问题的第一类关联问题集合。
在一些可行的实施方式中，上述处理器501用于：当上述初始业务问题闭环中包含的各级第一类关联问题的个数小于阈值时，将上述初始业务问题闭环中包含的各级第一类关联问题确定为上述初始标准问题的第一类关联问题集合。
结合第一方面，在一种可能的实施方式中，上述基于上述初始业务问题闭环中包含的各级第一类关联问题，确定出上述初始标准问题的第一类关联问题集合包括：
基于上述问答聚合图谱中上述初始业务问题闭环中包含的各级第一类关联问题与上述初始业务问题的连接概率，确定出上述初始业务问题闭环中包含的各级第一类关联问题出现在上述初始标准问题之后的连接概率；
根据上述初始业务问题闭环中出现在上述初始标准问题之后的连接概率大于阈值的第一类关联问题，确定上述初始标准问题的第一类关联问题集合。
在一些可行的实施方式中，上述处理器501可以是中央处理单元(central processing unit，CPU)，该处理器还可以是其他通用处理器、数字信号处理器(digital signal processor，DSP)、专用集成电路(application specific integrated circuit，ASIC)、现成可编程门阵列(field-programmable gate array，FPGA)或者其他可编程逻辑器件、分立门或者晶体管逻辑器件、分立硬件组件等。通用处理器可以是微处理器或者该处理器也可以是任何常规的处理器等。
该存储器502可以包括只读存储器和随机存取存储器，并向处理器501提供指令和数据。存储器502的一部分还可以包括非易失性随机存取存储器。例如，存储器502还可以存储设备类型的信息。
具体实现中，上述终端设备可通过其内置的各个功能模块执行如上述图1、图2以及图4中各个步骤所提供的实现方式，具体可参见上述各个步骤所提供的实现方式，在此不再赘述。
在本申请实施例中，终端设备通过获取用户操作数据得到用户操作数据所关联的初始业务问题，基于对初始业务问题的语义分析，可以在标准问题库中确定出初始业务问题对应的初始标准问题，从而将初始业务问题规范化，简化后续针对初始业务问题进行分析的过程，同时确定出初始标准问题对应的初始标准回答以及初始标准问题所包含的业务特征标签。根据初始标准问题，可以从问答知识聚合图谱中确定出与初始标准问题的语义关联的第一类关联问题集合。这里，第一类关联问题集合，也即，在用户问答习惯中，与初始业务问题相关联的问题的集合。根据初始标准问题以及初始标准问题所包含的业务特征标签，可以从问答知识聚合图谱中确定出与初始标准问题所包含的业务特征标签关联的第二 类关联问题集合。这里，第二类关联问题集合，也即，在业务特征标签对应的业务特征领域中，与初始业务问题相关联的问题的集合。将初始标准问题的初始标准回答、第一类关联问题集合以及第二关类联问题集合确定为初始业务问题的回答文本并输出至用户界面，可以在提供给用户初始业务问题对应的初始标准回答的同时，提供给用户接下来可能会提问的第一类关联问题集合，以及用户接下来可能要知道的第二类关联问题集合，从而优化用户体验，提高智能问答系统的问答效率。
本申请实施例还提供一种计算机可读存储介质，该计算机可读存储介质存储有计算机程序，该计算机程序包括程序指令，该程序指令被处理器执行时实现图1、图2以及图4中各个步骤所提供的方法，具体可参见上述各个步骤所提供的实现方式，在此不再赘述。
可选的，本申请涉及的存储介质如计算机可读存储介质可以是非易失性的，也可以是易失性的。
上述计算机可读存储介质可以是前述任一实施例提供的基于预测模型的用户行为识别装置或者上述终端设备的内部存储单元，例如电子设备的硬盘或内存。该计算机可读存储介质也可以是该电子设备的外部存储设备，例如该电子设备上配备的插接式硬盘，智能存储卡(smart media card,SMC)，安全数字(secure digital,SD)卡，闪存卡(flash card)等。进一步地，该计算机可读存储介质还可以既包括该电子设备的内部存储单元也包括外部存储设备。该计算机可读存储介质用于存储该计算机程序以及该电子设备所需的其他程序和数据。该计算机可读存储介质还可以用于暂时地存储已经输出或者将要输出的数据。
本申请的权利要求书和说明书及附图中的术语“第一”、“第二”、“第三”、“第四”等是用于区别不同对象，而不是用于描述特定顺序。此外，术语“包括”和“具有”以及它们任何变形，意图在于覆盖不排他的包含。例如包含了一系列步骤或单元的过程、方法、系统、产品或设备没有限定于已列出的步骤或单元，而是可选地还包括没有列出的步骤或单元，或可选地还包括对于这些过程、方法、产品或设备固有的其它步骤或单元。在本文中提及“实施例”意味着，结合实施例描述的特定特征、结构或特性可以包含在本申请的至少一个实施例中。在说明书中的各个位置展示该短语并不一定均是指相同的实施例，也不是与其它实施例互斥的独立的或备选的实施例。本领域技术人员显式地和隐式地理解的是，本文所描述的实施例可以与其它实施例相结合。在本申请说明书和所附权利要求书中使用的术语“和/或”是指相关联列出的项中的一个或多个的任何组合以及所有可能组合，并且包括这些组合。
本领域普通技术人员可以意识到，结合本文中所公开的实施例描述的各示例的单元及算法步骤，能够以电子硬件、计算机软件或者二者的结合来实现，为了清楚地说明硬件和软件的可互换性，在上述说明中已经按照功能一般性地描述了各示例的组成及步骤。这些功能究竟以硬件还是软件方式来执行，取决于技术方案的特定应用和设计约束条件。专业技术人员可以对每个特定的应用来使用不同方法来实现所描述的功能，但是这种实现不应认为超出本申请的范围。
本申请实施例提供的方法及相关装置是参照本申请实施例提供的方法流程图和/或结构示意图来描述的，具体可由计算机程序指令实现方法流程图和/或结构示意图的每一流程和/或方框、以及流程图和/或方框图中的流程和/或方框的结合。这些计算机程序指令可提供到通用计算机、专用计算机、嵌入式处理机或其他可编程数据处理设备的处理器以产生一个机器，使得通过计算机或其他可编程数据处理设备的处理器执行的指令产生用于实现在流程图一个流程或多个流程和/或结构示意图一个方框或多个方框中指定的功能的装置。这些计算机程序指令也可存储在能引导计算机或其他可编程数据处理设备以特定方式工作的计算机可读存储器中，使得存储在该计算机可读存储器中的指令产生包括指令装置的制造品，该指令装置实现在流程图一个流程或多个流程和/或结构示意图一个方框或多个方框 中指定的功能。这些计算机程序指令也可装载到计算机或其他可编程数据处理设备上，使得在计算机或其他可编程设备上执行一系列操作步骤以产生计算机实现的处理，从而在计算机或其他可编程设备上执行的指令提供用于实现在流程图一个流程或多个流程和/或结构示意一个方框或多个方框中指定的功能的步骤。

Claims (20)

  1. 一种智能问答方法，其中，所述方法包括：
    通过用户界面获取用于触发启动目标业务的用户操作数据，并确定所述用户操作数据所关联的初始业务问题；
    基于所述初始业务问题的语义分析，从标准问题库中确定出所述初始业务问题对应的初始标准问题、所述初始标准问题对应的初始标准回答以及初始标准问题所包含的业务特征标签；
    根据所述初始标准问题以及所述初始标准问题所包含的业务特征标签，从问答知识聚合图谱中确定出与所述初始标准问题的语义关联的第一类关联问题集合，以及与所述初始标准问题所包含的业务特征标签关联的第二类关联问题集合；
    将所述初始标准问题的初始标准回答、所述第一类关联问题集合以及所述第二关类联问题集合确定为所述初始业务问题的回答文本，并将所述初始业务问题的回答文本输出至启动所述目标业务的用户界面。
  2. 根据权利要求1所述的方法，其中，所述方法还包括：
    从所述目标业务的业务数据库中获取多个业务数据，并确定各业务数据所关联的业务问题；
    基于语义分析将所述各业务数据所关联的业务问题进行聚类得到多个标准问题，根据所述多个标准问题生成标准问题库。
  3. 根据权利要求2所述的方法，其中，所述方法还包括：
    从所述目标业务的业务数据库中获取多个历史业务数据，并确定所述多个历史业务数据中的多个历史用户数据和多个客服参考数据；
    根据所述多个历史用户数据关联的历史业务问题，在所述标准问题库中确定出各历史业务问题对应的标准问题作为样本业务问题，并根据所述各历史业务问题的提问顺序确定各样本业务问题的提问顺序，根据所述各样本业务问题和所述各样本业务问题的提问顺序确定用户样本数据；
    根据所述多个客服参考数据关联的参考业务问题，在所述标准问题库中确定出各参考业务问题对应标准问题作为客服样本问题，以及客服样本问题所包含的业务特征标签，根据所述客服样本问题和所述客服样本问题所包含的业务特征标签确定客服样本数据；
    根据所述用户样本数据和所述客服样本数据构建问答知识聚合图谱。
  4. 根据权利要求3所述的方法，其中，所述根据所述用户样本数据和所述客服样本数据构建问答知识聚合图谱包括：
    通过所述用户样本数据中的各样本业务问题确定出所述问答知识聚合图谱中的问答聚合图谱的节点，通过所述各样本业务问题的提问顺序确定出所述问答聚合图谱的节点连接关系，将所述问答聚合图谱的节点按照所述问答聚合图谱的节点连接关系进行连接，以得到所述问答聚合图谱；
    通过所述客服样本数据中的客服样本问题确定出所述问答知识聚合图谱中的知识聚合图谱的节点，通过所述客服样本问题所包含的业务特征标签确定出所述知识聚合图谱的节点连接关系，将所述问答聚合图谱的节点按照所述问答聚合图谱的节点连接关系进行连接，以得到所述知识聚合图谱。
  5. 根据权利要求4所述的方法，其中，所述从问答知识聚合图谱中确定出与所述初始标准问题的语义关联的第一类关联问题集合包括：
    从所述问答聚合图谱中确定出与所述初始标准问题的语义关联的初级第一类关联问题；
    从所述问答聚合图谱中确定出与所述初级第一类关联问题语义关联的次级第一类关 联问题，并从所述问答聚合图谱中确定出与所述次级第一类关联问题语义关联的次级第一类关联问题，直至任一级第一类关联问题的次级第一类关联问题中出现所述初始标准问题，以得到由各级第一关联问题组成的初始业务问题闭环；
    基于所述初始业务问题闭环中包含的各级第一类关联问题，确定出所述初始标准问题的第一类关联问题集合。
  6. 根据权利要求5所述的方法，其中，所述基于所述初始业务问题闭环中包含的各级第一类关联问题，确定出所述初始标准问题的第一类关联问题集合包括：
    当所述初始业务问题闭环中包含的各级第一类关联问题的个数小于阈值时，将所述初始业务问题闭环中包含的各级第一类关联问题确定为所述初始标准问题的第一类关联问题集合。
  7. 根据权利要求5所述的方法，其中，所述基于所述初始业务问题闭环中包含的各级第一类关联问题，确定出所述初始标准问题的第一类关联问题集合包括：
    基于所述问答聚合图谱中所述初始业务问题闭环中包含的各级第一类关联问题与所述初始业务问题的连接概率，确定出所述初始业务问题闭环中包含的各级第一类关联问题出现在所述初始标准问题之后的连接概率；
    根据所述初始业务问题闭环中出现在所述初始标准问题之后的连接概率大于阈值的第一类关联问题，确定所述初始标准问题的第一类关联问题集合。
  8. 一种智能问答装置，其中，所述装置包括：
    问题获取模块，用于通过用户界面获取用于触发启动目标业务的用户操作数据，并确定所述用户操作数据所关联的初始业务问题；
    语义分析模块，用于基于所述初始业务问题的语义分析，从标准问题库中确定出所述初始业务问题对应的初始标准问题、所述初始标准问题对应的初始标准回答以及初始标准问题所包含的业务特征标签；
    关联聚合模块，用于根据所述初始标准问题以及所述初始标准问题所包含的业务特征标签，从问答知识聚合图谱中确定出与所述初始标准问题的语义关联的第一类关联问题集合，以及与所述初始标准问题所包含的业务特征标签关联的第二类关联问题集合；
    结果输出模块，用于将所述初始标准问题的初始标准回答、所述第一类关联问题集合以及所述第二关类联问题集合确定为所述初始业务问题的回答文本，并将所述初始业务问题的回答文本输出至启动所述目标业务的用户界面。
  9. 一种终端设备，其中，包括处理器和存储器，所述处理器和存储器相互连接，其中，所述存储器用于存储计算机程序，所述计算机程序包括程序指令，所述处理器被配置用于调用所述程序指令，执行以下方法：
    通过用户界面获取用于触发启动目标业务的用户操作数据，并确定所述用户操作数据所关联的初始业务问题；
    基于所述初始业务问题的语义分析，从标准问题库中确定出所述初始业务问题对应的初始标准问题、所述初始标准问题对应的初始标准回答以及初始标准问题所包含的业务特征标签；
    根据所述初始标准问题以及所述初始标准问题所包含的业务特征标签，从问答知识聚合图谱中确定出与所述初始标准问题的语义关联的第一类关联问题集合，以及与所述初始标准问题所包含的业务特征标签关联的第二类关联问题集合；
    将所述初始标准问题的初始标准回答、所述第一类关联问题集合以及所述第二关类联问题集合确定为所述初始业务问题的回答文本，并将所述初始业务问题的回答文本输出至启动所述目标业务的用户界面。
  10. 根据权利要求9所述的终端设备，其中，所述处理器还用于执行：
    从所述目标业务的业务数据库中获取多个业务数据，并确定各业务数据所关联的业务问题；
    基于语义分析将所述各业务数据所关联的业务问题进行聚类得到多个标准问题，根据所述多个标准问题生成标准问题库。
  11. 根据权利要求10所述的终端设备，其中，所述处理器还用于执行：
    从所述目标业务的业务数据库中获取多个历史业务数据，并确定所述多个历史业务数据中的多个历史用户数据和多个客服参考数据；
    根据所述多个历史用户数据关联的历史业务问题，在所述标准问题库中确定出各历史业务问题对应的标准问题作为样本业务问题，并根据所述各历史业务问题的提问顺序确定各样本业务问题的提问顺序，根据所述各样本业务问题和所述各样本业务问题的提问顺序确定用户样本数据；
    根据所述多个客服参考数据关联的参考业务问题，在所述标准问题库中确定出各参考业务问题对应标准问题作为客服样本问题，以及客服样本问题所包含的业务特征标签，根据所述客服样本问题和所述客服样本问题所包含的业务特征标签确定客服样本数据；
    根据所述用户样本数据和所述客服样本数据构建问答知识聚合图谱。
  12. 根据权利要求11所述的终端设备，其中，执行所述根据所述用户样本数据和所述客服样本数据构建问答知识聚合图谱包括：
    通过所述用户样本数据中的各样本业务问题确定出所述问答知识聚合图谱中的问答聚合图谱的节点，通过所述各样本业务问题的提问顺序确定出所述问答聚合图谱的节点连接关系，将所述问答聚合图谱的节点按照所述问答聚合图谱的节点连接关系进行连接，以得到所述问答聚合图谱；
    通过所述客服样本数据中的客服样本问题确定出所述问答知识聚合图谱中的知识聚合图谱的节点，通过所述客服样本问题所包含的业务特征标签确定出所述知识聚合图谱的节点连接关系，将所述问答聚合图谱的节点按照所述问答聚合图谱的节点连接关系进行连接，以得到所述知识聚合图谱。
  13. 根据权利要求12所述的终端设备，其中，执行所述从问答知识聚合图谱中确定出与所述初始标准问题的语义关联的第一类关联问题集合包括：
    从所述问答聚合图谱中确定出与所述初始标准问题的语义关联的初级第一类关联问题；
    从所述问答聚合图谱中确定出与所述初级第一类关联问题语义关联的次级第一类关联问题，并从所述问答聚合图谱中确定出与所述次级第一类关联问题语义关联的次级第一类关联问题，直至任一级第一类关联问题的次级第一类关联问题中出现所述初始标准问题，以得到由各级第一关联问题组成的初始业务问题闭环；
    基于所述初始业务问题闭环中包含的各级第一类关联问题，确定出所述初始标准问题的第一类关联问题集合。
  14. 根据权利要求13所述的终端设备，其中，执行所述基于所述初始业务问题闭环中包含的各级第一类关联问题，确定出所述初始标准问题的第一类关联问题集合包括：
    当所述初始业务问题闭环中包含的各级第一类关联问题的个数小于阈值时，将所述初始业务问题闭环中包含的各级第一类关联问题确定为所述初始标准问题的第一类关联问题集合；或者，
    基于所述问答聚合图谱中所述初始业务问题闭环中包含的各级第一类关联问题与所述初始业务问题的连接概率，确定出所述初始业务问题闭环中包含的各级第一类关联问题出现在所述初始标准问题之后的连接概率；根据所述初始业务问题闭环中出现在所述初始标准问题之后的连接概率大于阈值的第一类关联问题，确定所述初始标准问题的第一类关 联问题集合。
  15. 一种计算机可读存储介质，其中，所述计算机可读存储介质存储有计算机程序，所述计算机程序包括程序指令，所述程序指令当被处理器执行时使所述处理器执行以下方法：
    通过用户界面获取用于触发启动目标业务的用户操作数据，并确定所述用户操作数据所关联的初始业务问题；
    基于所述初始业务问题的语义分析，从标准问题库中确定出所述初始业务问题对应的初始标准问题、所述初始标准问题对应的初始标准回答以及初始标准问题所包含的业务特征标签；
    根据所述初始标准问题以及所述初始标准问题所包含的业务特征标签，从问答知识聚合图谱中确定出与所述初始标准问题的语义关联的第一类关联问题集合，以及与所述初始标准问题所包含的业务特征标签关联的第二类关联问题集合；
    将所述初始标准问题的初始标准回答、所述第一类关联问题集合以及所述第二关类联问题集合确定为所述初始业务问题的回答文本，并将所述初始业务问题的回答文本输出至启动所述目标业务的用户界面。
  16. 根据权利要求15所述的计算机可读存储介质，其中，所述程序指令被处理器执行时还使所述处理器执行：
    从所述目标业务的业务数据库中获取多个业务数据，并确定各业务数据所关联的业务问题；
    基于语义分析将所述各业务数据所关联的业务问题进行聚类得到多个标准问题，根据所述多个标准问题生成标准问题库。
  17. 根据权利要求16所述的计算机可读存储介质，其中，所述程序指令被处理器执行时还使所述处理器执行：
    从所述目标业务的业务数据库中获取多个历史业务数据，并确定所述多个历史业务数据中的多个历史用户数据和多个客服参考数据；
    根据所述多个历史用户数据关联的历史业务问题，在所述标准问题库中确定出各历史业务问题对应的标准问题作为样本业务问题，并根据所述各历史业务问题的提问顺序确定各样本业务问题的提问顺序，根据所述各样本业务问题和所述各样本业务问题的提问顺序确定用户样本数据；
    根据所述多个客服参考数据关联的参考业务问题，在所述标准问题库中确定出各参考业务问题对应标准问题作为客服样本问题，以及客服样本问题所包含的业务特征标签，根据所述客服样本问题和所述客服样本问题所包含的业务特征标签确定客服样本数据；
    根据所述用户样本数据和所述客服样本数据构建问答知识聚合图谱。
  18. 根据权利要求17所述的计算机可读存储介质，其中，执行所述根据所述用户样本数据和所述客服样本数据构建问答知识聚合图谱包括：
    通过所述用户样本数据中的各样本业务问题确定出所述问答知识聚合图谱中的问答聚合图谱的节点，通过所述各样本业务问题的提问顺序确定出所述问答聚合图谱的节点连接关系，将所述问答聚合图谱的节点按照所述问答聚合图谱的节点连接关系进行连接，以得到所述问答聚合图谱；
    通过所述客服样本数据中的客服样本问题确定出所述问答知识聚合图谱中的知识聚合图谱的节点，通过所述客服样本问题所包含的业务特征标签确定出所述知识聚合图谱的节点连接关系，将所述问答聚合图谱的节点按照所述问答聚合图谱的节点连接关系进行连接，以得到所述知识聚合图谱。
  19. 根据权利要求18所述的计算机可读存储介质，其中，执行所述从问答知识聚合 图谱中确定出与所述初始标准问题的语义关联的第一类关联问题集合包括：
    从所述问答聚合图谱中确定出与所述初始标准问题的语义关联的初级第一类关联问题；
    从所述问答聚合图谱中确定出与所述初级第一类关联问题语义关联的次级第一类关联问题，并从所述问答聚合图谱中确定出与所述次级第一类关联问题语义关联的次级第一类关联问题，直至任一级第一类关联问题的次级第一类关联问题中出现所述初始标准问题，以得到由各级第一关联问题组成的初始业务问题闭环；
    基于所述初始业务问题闭环中包含的各级第一类关联问题，确定出所述初始标准问题的第一类关联问题集合。
  20. 根据权利要求19所述的计算机可读存储介质，其中，执行所述基于所述初始业务问题闭环中包含的各级第一类关联问题，确定出所述初始标准问题的第一类关联问题集合包括：
    当所述初始业务问题闭环中包含的各级第一类关联问题的个数小于阈值时，将所述初始业务问题闭环中包含的各级第一类关联问题确定为所述初始标准问题的第一类关联问题集合；或者，
    基于所述问答聚合图谱中所述初始业务问题闭环中包含的各级第一类关联问题与所述初始业务问题的连接概率，确定出所述初始业务问题闭环中包含的各级第一类关联问题出现在所述初始标准问题之后的连接概率；根据所述初始业务问题闭环中出现在所述初始标准问题之后的连接概率大于阈值的第一类关联问题，确定所述初始标准问题的第一类关联问题集合。
PCT/CN2021/090197 2020-12-18 2021-04-27 智能问答方法、装置、设备及存储介质 WO2022126965A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN202011511700.9 2020-12-18
CN202011511700.9A CN112667790B (zh) 2020-12-18 2020-12-18 智能问答方法、装置、设备及存储介质

Publications (1)

Publication Number Publication Date
WO2022126965A1 true WO2022126965A1 (zh) 2022-06-23

Family

ID=75407491

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2021/090197 WO2022126965A1 (zh) 2020-12-18 2021-04-27 智能问答方法、装置、设备及存储介质

Country Status (2)

Country Link
CN (1) CN112667790B (zh)
WO (1) WO2022126965A1 (zh)

Families Citing this family (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112667790B (zh) * 2020-12-18 2024-05-10 平安科技（深圳）有限公司 智能问答方法、装置、设备及存储介质
CN113076430B (zh) * 2021-04-19 2024-01-30 北京搜狗科技发展有限公司 一种基于医疗问题的数据处理方法及装置

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111400471A (zh) * 2020-03-13 2020-07-10 江苏满运软件科技有限公司 问题推荐方法、系统、电子设备和存储介质
CN111552880A (zh) * 2020-04-30 2020-08-18 杭州网易再顾科技有限公司 基于知识图谱的数据处理方法、装置、介质及电子设备
CN111651567A (zh) * 2020-04-16 2020-09-11 北京奇艺世纪科技有限公司 一种业务问答数据处理方法及装置
CN111767385A (zh) * 2020-07-09 2020-10-13 腾讯科技（深圳）有限公司 一种智能问答方法及装置
CN112667790A (zh) * 2020-12-18 2021-04-16 平安科技（深圳）有限公司 智能问答方法、装置、设备及存储介质

Family Cites Families (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108833595B (zh) * 2015-12-24 2021-06-29 上海智臻智能网络科技股份有限公司 用于在线客服服务的计算机可读存储介质
KR102428887B1 (ko) * 2016-11-07 2022-08-02 주식회사 오제이월드 태그 서칭 기반의 자동 질의 응답 시스템 및 방법
CN109087132B (zh) * 2018-07-18 2021-07-30 国家电网有限公司 一种基于知识图谱的用户问题推送方法及装置
CN109063152A (zh) * 2018-08-08 2018-12-21 鲸数科技（北京）有限公司 智能问答方法、装置及智能终端
CN109299476B (zh) * 2018-11-28 2023-04-18 出门问问创新科技有限公司 问答方法、装置、电子设备及存储介质
CN111241260A (zh) * 2020-01-08 2020-06-05 平安科技（深圳）有限公司 基于人机交互的数据处理方法、装置、设备及存储介质

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111400471A (zh) * 2020-03-13 2020-07-10 江苏满运软件科技有限公司 问题推荐方法、系统、电子设备和存储介质
CN111651567A (zh) * 2020-04-16 2020-09-11 北京奇艺世纪科技有限公司 一种业务问答数据处理方法及装置
CN111552880A (zh) * 2020-04-30 2020-08-18 杭州网易再顾科技有限公司 基于知识图谱的数据处理方法、装置、介质及电子设备
CN111767385A (zh) * 2020-07-09 2020-10-13 腾讯科技（深圳）有限公司 一种智能问答方法及装置
CN112667790A (zh) * 2020-12-18 2021-04-16 平安科技（深圳）有限公司 智能问答方法、装置、设备及存储介质

Also Published As

Publication number Publication date
CN112667790A (zh) 2021-04-16
CN112667790B (zh) 2024-05-10

Similar Documents

Publication Publication Date Title
US10650034B2 (en) Categorizing users based on similarity of posed questions, answers and supporting evidence
US11151175B2 (en) On-demand relation extraction from text
US11868733B2 (en) Creating a knowledge graph based on text-based knowledge corpora
US10929348B2 (en) Method and system for large scale data curation
US20190362247A1 (en) Image tagging based upon cross domain context
US10339188B2 (en) Question distribution method and a question distribution system for a Q and A platform
US10957214B2 (en) Managing answer feasibility
US20230031591A1 (en) Methods and apparatus to facilitate generation of database queries
CN106462807B (zh) 根据大规模非结构化数据学习多媒体语义
CN111566638B (zh) 向应用编程接口添加描述性元数据以供智能代理使用
US20120158791A1 (en) Feature vector construction
US8849854B2 (en) Method and system for providing detailed information in an interactive manner in a short message service (SMS) environment
US7818323B2 (en) Discovering topical structures of databases
US20170262429A1 (en) Collecting Training Data using Anomaly Detection
WO2022126965A1 (zh) 智能问答方法、装置、设备及存储介质
WO2016011904A1 (zh) 数据库访问方法及装置、数据库系统
US10326863B2 (en) Speed and accuracy of computers when resolving client queries by using graph database model
CN109214674B (zh) 客服系统管理方法、客服系统及电子设备
WO2020015171A1 (zh) 电子装置、招商引资的目标对象确定方法、系统及存储介质
JP2017097823A (ja) 分散ネットワークで使用するための検索サーバ、端末装置、及び検索方法
US9195733B2 (en) Clustering with virtual entities using associative memories
US20210133189A1 (en) Automatic Indexing of Relevant Domains in a Data Lake for Data Discovery and Integration
US20230214679A1 (en) Extracting and classifying entities from digital content items
Bu et al. Quality assessment in crowdsourced classification tasks
US10339135B2 (en) Query handling in search systems

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 21904904

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 21904904

Country of ref document: EP

Kind code of ref document: A1