WO2021128663A1 - 一种机器人应答方法、装置、设备及存储介质 - Google Patents

一种机器人应答方法、装置、设备及存储介质 Download PDF

Info

Publication number
WO2021128663A1
WO2021128663A1 PCT/CN2020/085530 CN2020085530W WO2021128663A1 WO 2021128663 A1 WO2021128663 A1 WO 2021128663A1 CN 2020085530 W CN2020085530 W CN 2020085530W WO 2021128663 A1 WO2021128663 A1 WO 2021128663A1
Authority
WO
WIPO (PCT)
Prior art keywords
semantic information
question
query
clusters
voice
Prior art date
Application number
PCT/CN2020/085530
Other languages
English (en)
French (fr)
Inventor
郑宇宇
Original Assignee
京东数字科技控股有限公司
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 京东数字科技控股有限公司 filed Critical 京东数字科技控股有限公司
Priority to EP20906575.4A priority Critical patent/EP4083812A4/en
Priority to US17/785,369 priority patent/US20230028830A1/en
Publication of WO2021128663A1 publication Critical patent/WO2021128663A1/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
    • G06F16/3329—Natural language query formulation or dialogue systems
    • 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
    • G06F16/3331—Query processing
    • G06F16/334—Query execution
    • G06F16/3343—Query execution using phonetics
    • 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
    • G06F16/3331—Query processing
    • G06F16/334—Query execution
    • G06F16/3344—Query execution using natural language analysis
    • 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
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/004—Artificial life, i.e. computing arrangements simulating life
    • G06N3/006—Artificial life, i.e. computing arrangements simulating life based on simulated virtual individual or collective life forms, e.g. social simulations or particle swarm optimisation [PSO]
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/045—Combinations of networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/0464—Convolutional networks [CNN, ConvNet]
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • G06N3/088—Non-supervised learning, e.g. competitive learning
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N5/00—Computing arrangements using knowledge-based models
    • G06N5/02—Knowledge representation; Symbolic representation
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L51/00—User-to-user messaging in packet-switching networks, transmitted according to store-and-forward or real-time protocols, e.g. e-mail
    • H04L51/02—User-to-user messaging in packet-switching networks, transmitted according to store-and-forward or real-time protocols, e.g. e-mail using automatic reactions or user delegation, e.g. automatic replies or chatbot-generated messages

Definitions

  • This application relates to the field of communication technology, and in particular to a robot response method, device, equipment, and storage medium.
  • the chat robot system is a system that can be online all the time and communicate with people through natural language by means of communication.
  • the chat robot system stores a large number of questions and corresponding answers.
  • the chat robot Will find the corresponding answer according to the question and feedback it to the user.
  • chat bots will give a list of replies.
  • the content in the list has a certain degree of similarity with the user’s questions. This list is retrieved from the database, and the user needs to give Click on the question in the list to get the answer you need.
  • the user is less efficient in obtaining accurate answers to questions in the process of interacting with the robot.
  • This application provides a robot response method, device, equipment, and storage medium, so as to improve the efficiency of obtaining accurate answers.
  • the present application provides a robot response method, including: acquiring the current query voice; extracting the semantic information of the current query voice; matching the semantic information of the current query voice with a plurality of pre-stored semantic information clusters to obtain the matching
  • Each semantic information cluster includes: at least one question and answer instance, each question and answer instance includes: semantic information corresponding to a historical query voice and query questions selected in the query list corresponding to the historical query voice; obtain the target The number of times each query question in the semantic information cluster is selected, and the target query question corresponding to the current query voice is determined according to the number of times each query question is selected, and the query response corresponding to the target query question is output.
  • This application converts speech into semantic information, matches semantic information with pre-stored semantic information clusters, and obtains the optimal answer to the current query question based on the number of times the historical query question has been selected, and achieves the accuracy of the question without a clear answer. Reply, thereby improving the efficiency of obtaining accurate answers.
  • matching the semantic information of the current query voice with multiple pre-stored semantic information clusters includes: determining the similarity between the semantic information of the current query voice and the semantic information corresponding to each historical query voice in the multiple semantic information clusters Degree; if the similarity between the semantic information of the current query voice and the semantic information corresponding to the historical query voice in a semantic information cluster is greater than the first preset similarity, then the semantic information cluster is taken as the target semantic information cluster. That is, the matching of semantic information and semantic information clusters is realized according to the similarity of semantic information.
  • determining the target query question corresponding to the current query voice according to the number of times each query question has been selected includes: in each query question, determining a query question that has been selected more than a preset number of times as the target query question. That is, there is no need for the user to select the question, and the target query question can be determined according to the number of times the query question is selected, thereby improving the efficiency of obtaining answers.
  • it further includes: receiving multiple semantic information clusters sent by the server.
  • this application provides a robot response method, including: acquiring a plurality of semantic information clusters, each semantic information cluster includes: at least one question and answer instance, each question and answer instance includes: semantic information corresponding to a historical query voice and The selected query question in the query list corresponding to the historical query voice; send multiple semantic information clusters to the robot so that the robot can match the multiple semantic information clusters with the semantic information of the current query voice to obtain the matched target semantic information cluster , And determine the target query question corresponding to the current query voice according to the number of times each query question in the target semantic information cluster is selected, and output the query response corresponding to the target query question.
  • a robot response method including: acquiring a plurality of semantic information clusters, each semantic information cluster includes: at least one question and answer instance, each question and answer instance includes: semantic information corresponding to a historical query voice and The selected query question in the query list corresponding to the historical query voice; send multiple semantic information clusters to the robot so that the robot can match the multiple semantic information clusters with the semantic information of the current query
  • acquiring multiple semantic information clusters before acquiring multiple semantic information clusters, it further includes: acquiring multiple question and answer instances; correspondingly, acquiring multiple semantic information clusters includes: determining the similarity of the semantic information in each of the multiple question answering instances Degree; the question and answer instances with similarity greater than the second preset similarity are classified as a semantic information cluster. The semantic information cluster is classified, and whether it is a semantic information cluster is determined according to the similarity.
  • acquiring multiple semantic information clusters previously further includes: acquiring at least one basic semantic information cluster and at least one question and answer instance; correspondingly, acquiring multiple semantic information clusters includes: updating at least one basic semantic information cluster according to the at least one question and answer instance , In order to get multiple semantic information clusters.
  • similar question and answer instances can be regarded as a semantic information cluster, and similar question and answer instances will have similar or identical selected query questions, so that the robot can count the number of times each query question has been selected. Furthermore, the target query question corresponding to the current query voice can be determined according to the number of times each query question is selected, and the query response corresponding to the target query question can be output, thereby improving the efficiency of obtaining accurate answers.
  • updating at least one basic semantic information cluster according to at least one question and answer instance to obtain multiple semantic information clusters includes: determining the similarity between each semantic information in the at least one question answering instance and each semantic information in the at least one basic semantic information cluster ; For each question and answer instance in at least one question and answer instance, the question and answer instance is divided into at least one basic semantic information cluster and the question and answer instance similarity is greater than the third preset similarity of the basic semantic information cluster, so as to achieve the voice information Dynamic updates of clusters.
  • the method further includes: obtaining the selected times of each query question in the multiple semantic information clusters; and sending the selected times of each query question in the multiple semantic information clusters to the robot.
  • the robot can determine the target query question corresponding to the current query voice according to the number of times each query question is selected, and output the query response corresponding to the target query question. This improves the efficiency of obtaining accurate answers.
  • the present application also provides a robot answering device, equipment, readable storage medium, and computer program product.
  • a robot answering device equipment, readable storage medium, and computer program product.
  • the present application provides a robot response device, including: a first acquisition module, an extraction module, a matching module, a second acquisition module, a determination module, and an output module, the first acquisition module is used to acquire the current query voice; the extraction module Used to extract the semantic information of the current query voice; the matching module is used to match the semantic information of the current query voice with a plurality of pre-stored semantic information clusters to obtain the matched target semantic information cluster, each semantic information cluster includes at least: A question and answer instance, each question and answer instance includes: semantic information corresponding to a historical query voice and query questions selected in the query list corresponding to the historical query voice; the second acquisition module is used to obtain the information of each query question in the target semantic information cluster The number of times of being selected; the determining module is used to determine the target query question corresponding to the current query voice according to the selected times of each query question; the output module is used to output the query response corresponding to the target query question.
  • the matching module is specifically used to determine the similarity between the semantic information of the current query voice and the semantic information corresponding to each historical query voice in the multiple semantic information clusters; if the semantic information of the current query voice is in a semantic information cluster If the similarity of the semantic information corresponding to the historical query voice is greater than the first preset similarity, then the semantic information cluster is used as the target semantic information cluster.
  • the determining module is specifically used for determining, among each query question, a query question that has been selected more than a preset number of times as a target query question.
  • it further includes: a receiving module for receiving multiple semantic information clusters sent by the server.
  • the present application provides a robot response device, including: a first acquisition module for acquiring a plurality of semantic information clusters, each semantic information cluster includes: at least one question and answer instance, each question and answer instance includes: a historical query The semantic information corresponding to the voice and the query question selected in the query list corresponding to the historical query voice; the first sending module is used to send a plurality of semantic information clusters to the robot.
  • it further includes: a second obtaining module, configured to obtain multiple question and answer instances; correspondingly, the first obtaining module is specifically configured to: determine the similarity of semantic information in each of the multiple question and answer instances; The question and answer instances whose similarity is greater than the second preset similarity are classified into a semantic information cluster.
  • a second obtaining module configured to obtain multiple question and answer instances; correspondingly, the first obtaining module is specifically configured to: determine the similarity of semantic information in each of the multiple question and answer instances; The question and answer instances whose similarity is greater than the second preset similarity are classified into a semantic information cluster.
  • a third acquiring module configured to acquire at least one basic semantic information cluster and at least one question and answer instance; correspondingly, the first acquiring module is specifically configured to: update at least one basic semantic information cluster according to the at least one question and answer instance , In order to get multiple semantic information clusters.
  • the first obtaining module is specifically configured to: determine the similarity between each semantic information in at least one question and answer instance and each semantic information in at least one basic semantic information cluster; for each question and answer instance in at least one question and answer instance, the question and answer instance It is divided into at least one basic semantic information cluster with a similarity to the question and answer instance that is greater than a third preset similarity.
  • it further includes: a fourth obtaining module, which is used to obtain the number of times that each query question in the multiple semantic information clusters has been selected; the second sending module, which is used to send the query question of each query question in the multiple semantic information clusters to the robot. Choose the number of times.
  • the embodiments of the present application provide a non-transitory computer-readable storage medium storing computer instructions.
  • the computer instructions are used to make a computer execute the application of the first aspect or any one of the optional methods of the first aspect. Robot response method.
  • the embodiments of the present application provide a non-transitory computer-readable storage medium storing computer instructions.
  • the computer instructions are used to make the computer execute the application of the second aspect or any one of the optional methods of the second aspect. Robot response method.
  • an embodiment of the present application provides a computer program product, the product including: computer instructions, the computer instructions are used to make a computer execute any one of the first aspect or the first aspect of the optional manner applied to the robot Response method.
  • an embodiment of the present application provides a computer program product, the product including: computer instructions, the computer instructions are used to make a computer execute any one of the second aspect or the second aspect of the optional manner applied to the robot Response method.
  • the robot response method, device, equipment and storage medium provided by the present application obtain the current query voice through the robot, extract the semantic information of the current query voice, and match the semantic information of the current query voice with multiple pre-stored semantic information clusters , Get the matched target semantic information cluster, each semantic information cluster includes: at least one question and answer instance, each question and answer instance includes: semantic information corresponding to a historical query voice and a query selected in the query list corresponding to the historical query voice
  • the robot obtains the selected times of each query question in the target semantic information cluster, and determines the target query question corresponding to the current query voice according to the selected times of each query question, and outputs the query response corresponding to the target query question. Accurate answers to the answers to the questions, thereby improving the efficiency of obtaining accurate answers.
  • Figure 1 is a schematic diagram of an application scenario provided by this application.
  • Figure 2 is an interactive flow chart of a robot response method provided by this application
  • Figure 3 is a schematic diagram of the robot's query list
  • FIG. 4A is a schematic diagram of an interface provided by an embodiment of this application.
  • 4B is a schematic diagram of another interface provided by an embodiment of this application.
  • FIG. 5 is an interactive flowchart of another robot response method provided by this application.
  • FIG. 6 is an interaction flowchart of yet another robot response method provided by this application.
  • FIG. 7 is an interaction flowchart of yet another robot response method provided by this application.
  • FIG. 8 is a schematic structural diagram of a robot answering device provided by this application.
  • FIG. 9 is a schematic structural diagram of another robot answering device provided by this application.
  • FIG. 10 is a schematic diagram of the structure of the robot provided by this application.
  • FIG. 11 is a schematic diagram of the structure of the server provided by this application.
  • FIG. 1 is a schematic diagram of an application scenario provided by this application.
  • the robot 001 is used to implement interactive responses with users.
  • the robot stores multiple semantic information clusters and a big data search database (Elastic Search, ES), where each semantic information cluster includes: at least one Question and answer instances, each of the question and answer instances includes: semantic information corresponding to a historical query voice and a query question selected in the query list corresponding to the historical query voice; the query response corresponding to the query question is stored in the ES database, namely Find the answer to the question.
  • the server 002 is used to establish multiple semantic information clusters and send multiple semantic information clusters to the robot 001.
  • the server 002 can also train the neural network model and send the trained neural network model to the robot 001.
  • the robot 001 can determine the semantic information corresponding to the query voice through the neural network model.
  • the chat bot will give a list of replies.
  • the content in the list has a certain degree of similarity with the user’s question.
  • This list is retrieved from the database. The user needs to click on the question in the given list to get the required answer.
  • the user is less efficient in obtaining accurate answers to questions in the process of interacting with the robot.
  • the present application provides a robot response method, device, equipment, and storage medium.
  • the main idea of this application is that the robot determines the target voice information cluster corresponding to the current query voice, determines the target query question according to the number of times each query question in the target semantic information cluster is selected, and outputs the query response corresponding to the target query question.
  • Figure 2 is an interactive flow chart of a robot response method provided by this application.
  • the network elements involved in the method include a robot and a server, as shown in Figure 2, the method includes the following steps:
  • Step S201 The server obtains multiple semantic information clusters.
  • Each semantic information cluster includes: at least one question and answer instance, and each question and answer instance includes: semantic information corresponding to a historical query voice and query questions selected in the query list corresponding to the historical query voice.
  • Figure 3 is a schematic diagram of the robot's query list. As shown in Figure 3, when the user inputs a question with no clear answer, the robot will display the historical query questions that have a certain degree of similarity with the user’s voice to the user in the form of a list. Among them, all historical query questions in the same list contain the same or similar semantic information.
  • Step S202 The server sends multiple semantic information clusters to the robot.
  • Step S203 The robot obtains the current query voice.
  • Step S204 The robot extracts the semantic information of the current query voice.
  • Step S205 The robot matches the semantic information of the current query voice with a plurality of pre-stored semantic information clusters to obtain the matched target semantic information clusters.
  • Step S206 The robot obtains the selected times of each query question in the target semantic information cluster, determines the target query question corresponding to the current query voice according to the selected times of each query question, and outputs the query response corresponding to the target query question.
  • step S203 The following descriptions are made for step S203 and step 204:
  • the robot collects the user's current query voice through a microphone, and extracts semantic information of the current query voice through a neural network model.
  • the neural network model consists of a bidirectional encoder representation (BERT) + convolutional neural network (Convolutional Neural Network, CNN) + fully connected layers (FC) structure.
  • the server can train the neural network model and send the trained neural network model to the robot. So that the robot can use the current query voice as the input of the neural network model to obtain the semantic information of the current query voice.
  • the essence of the training process of the server on the neural network model is: the server trains the parameters involved in the neural network model.
  • the input of the above-mentioned BERT model is the query voice, and the first semantic information of the query voice is obtained.
  • CNN is a feed-forward neural network. Its artificial neurons can respond to a part of the surrounding units in the coverage area. It includes convolutional layer and pooling layer. One semantic information is convolved to obtain the output of all or part of the hidden layer to obtain the second semantic information, and the second semantic information is reduced by using the pooling layer to obtain the third semantic information of the query voice. It should be noted that any of the above-mentioned semantic information can be called a semantic vector or a semantic matrix.
  • the FC layer functions as a "classifier" in the entire neural network model, weighting and summing the third semantic information and the second semantic information, that is, through the fully connected layer, the final semantic information corresponding to the query voice is output.
  • the robot after extracting the semantic information of the current query voice, the robot first determines whether the semantic information corresponds to a unique query question. If it exists, steps S205 and S206 are not executed. If the semantic information corresponds to multiple query questions , Then step S205 and step S206 are executed.
  • the foregoing multiple semantic information clusters may be pre-stored in the robot's local storage space, or may also be stored in the cloud storage space, which is not limited in this application.
  • robot matches the semantic information of the current query voice with multiple pre-stored semantic information clusters.
  • Step 206 The robot obtains the selected times of each query question in the target semantic information cluster, determines the target query question corresponding to the current query voice according to the selected times of each query question, and outputs the query response corresponding to the target query question.
  • the robot obtains the selected times of each query question in multiple semantic information clusters from the server. Or, the robot counts the number of times each query question in multiple semantic information clusters is selected.
  • each query question it is determined that a query question that has been selected more than a preset number of times is determined as the target query question.
  • each query question has a priority, and the robot can select the query question with the highest priority as the target query question.
  • the above-mentioned preset times can be set according to actual conditions, such as 100 or 1000, which is not limited in this application.
  • the robot queries the query response corresponding to the target query question in the ES database and outputs it to the user.
  • ES database is a distributed, high-expansion, high-real-time search and data analysis engine. It can easily make a large amount of data have the ability to search, analyze and explore.
  • Fig. 4A is a schematic diagram of the interface provided by an embodiment of the application. As shown in Fig. 4A, the robot displays the target query question "What are the preferential policies for using QuickPass?" The corresponding query response "Preferential activities are updated irregularly," The details are subject to the page display".
  • Figure 4B is a schematic diagram of the interface provided by another embodiment of the application.
  • the robot displays the target query question "What preferential policies does the platform have?”
  • the details are subject to the page display.
  • FIG. 5 is an interactive flowchart diagram of another robot response method provided by this application.
  • the above step S205 specifically includes:
  • S205A The robot determines the similarity between the semantic information of the current query voice and the semantic information corresponding to each historical query voice in the multiple semantic information clusters.
  • any piece of semantic information can also be called a semantic vector. Therefore, the robot determines the similarity between the semantic vector of the current query voice and the semantic vector corresponding to any historical query voice in the following way: the two semantic vectors are multiplied to obtain the similarity of the two semantic vectors. The larger the multiplication result, the higher the similarity between the two semantic vectors. Alternatively, the distance between two semantic vectors is calculated. The larger the distance, the lower the similarity of the two semantic vectors.
  • the above-mentioned first preset similarity may be set according to actual conditions, for example, it may be 0.8, 0.9, etc., which is not limited in this application.
  • the robot converts speech into semantic information, and determines the target semantic information cluster according to the similarity between the semantic information of the current query voice and the semantic information corresponding to the historical query voice in the multiple semantic information clusters, and determines the target semantic information cluster through the The number of times that the historical query question is selected obtains the best answer to the current query question, and an accurate answer to the question without a clear answer is achieved, thereby improving the efficiency of obtaining accurate answers.
  • FIG. 6 is an interaction flowchart of yet another robot response method provided by this application. Further, before step S201, it further includes:
  • Step S200 The server obtains multiple question and answer instances.
  • step S201 includes:
  • Step S201A The server determines the similarity of semantic information in every two question and answer instances in the plurality of question and answer instances.
  • Step S201B The server classifies the question and answer instances whose similarity is greater than the second preset similarity into a semantic information cluster.
  • a method for obtaining semantic information clusters is provided. This method is based on a density-based clustering algorithm (Density-Based Spatial Clustering of Applications with Noise, DBScan), and DBSCAN defines clusters as densely connected points The largest set of, can divide the area with high enough density into clusters, and can find clusters of arbitrary shape in the spatial database of noise.
  • DBScan Density-Based Spatial Clustering of Applications with Noise
  • the semantic information may be referred to as a semantic vector
  • the similarity of two semantic information may be the dot product of the two semantic vectors.
  • the server first obtains multiple question and answer instances and determines the semantic similarity of each of the two question and answer instances. If the similarity of the two question and answer instances is greater than the second preset similarity, it is determined that the two question and answer instances belong to the same semantic information cluster, and then Obtain multiple semantic information clusters, and the problem instances in each semantic information cluster have a sufficiently high degree of similarity.
  • the aforementioned second preset similarity can be set according to actual conditions, for example, it can be 0.8, 0.9, etc., which is not limited in this application.
  • the server may periodically obtain the semantic information clusters. For example, at 12 o'clock every night, the server periodically traces back the question and answer instances in the past week to establish and obtain multiple semantic information clusters.
  • FIG. 7 is an interaction flowchart of another robot response method provided by this application. Further, before step S201, it further includes:
  • Step S200A The server obtains at least one basic semantic information cluster and at least one question and answer instance.
  • step S201 includes:
  • Step S201C The server determines the similarity between each semantic information in at least one question and answer instance and each semantic information in at least one basic semantic information cluster.
  • Step S201D For each question and answer instance in the at least one question and answer instance, the server divides the question and answer instance into the basic semantic information clusters whose similarity with the question and answer instance in the at least one basic semantic information cluster is greater than the third preset similarity.
  • the so-called basic semantic information cluster refers to the currently established semantic information cluster, and the server can generate a new semantic information cluster on the basis of the semantic information cluster.
  • the aforementioned third preset similarity can be set according to actual conditions, for example, it can be 0.8, 0.9, etc., which is not limited in this application.
  • the server may periodically obtain at least one question instance, for example: every one minute back to the past one minute question and answer instance to update the basic semantic information cluster.
  • the server can obtain multiple semantic information clusters through the above two optional methods, and send the multiple semantic information clusters to the robot, so that the robot can determine the current query voice through the multiple semantic information clusters.
  • the corresponding target semantic information cluster the server can periodically obtain multiple semantic information clusters, that is, multiple semantic information clusters are dynamically changing, so that it can better provide users with accurate answers.
  • Fig. 8 is a schematic structural diagram of a robot answering device provided by this application. As shown in Figure 8, the robot response device includes:
  • the first obtaining module 801 is used to obtain the current query voice.
  • the extraction module 802 is used to extract the semantic information of the current query voice.
  • the matching module 803 is used to match the semantic information of the current query voice with multiple pre-stored semantic information clusters to obtain the matched target semantic information clusters.
  • Each semantic information cluster includes: at least one question and answer instance, and each question and answer instance Including: semantic information corresponding to a historical query voice and query questions selected in the query list corresponding to the historical query voice.
  • the second obtaining module 804 is used to obtain the selected times of each query question in the target semantic information cluster.
  • the determining module 805 is configured to determine the target query question corresponding to the current query voice according to the number of times each query question is selected.
  • the output module 806 is used to output the query response corresponding to the target query question.
  • the matching module 803 is specifically configured to: determine the similarity between the semantic information of the current query voice and the semantic information corresponding to each historical query voice in the multiple semantic information clusters; if the semantic information of the current query voice and a semantic information cluster If the similarity of the semantic information corresponding to the historical query voice in is greater than the first preset similarity, then the semantic information cluster is taken as the target semantic information cluster.
  • the determining module 805 is specifically configured to: in each query question, determine a query question that has been selected more than a preset number of times as a target query question.
  • it also includes:
  • the receiving module 807 is used to receive multiple semantic information clusters sent by the server.
  • the robot answering device provided in the present application can execute the corresponding robot answering method on the robot side.
  • Fig. 9 is a schematic structural diagram of another robot answering device provided by this application. As shown in Figure 9, the robot response device includes:
  • the first acquisition module 901 is used to acquire multiple semantic information clusters, each semantic information cluster includes: at least one question and answer instance, and each question and answer instance includes: semantic information corresponding to a historical query voice and a query list corresponding to the historical query voice The query question selected in.
  • the first sending module 902 is used to send multiple semantic information clusters to the robot.
  • it also includes:
  • the second obtaining module 903 is configured to obtain multiple question and answer instances; correspondingly, the second obtaining module is specifically used to: determine the similarity of semantic information in each of the multiple question and answer instances; and make the similarity greater than the second
  • the question and answer instances with preset similarity are grouped into a semantic information cluster.
  • it also includes:
  • the third obtaining module 904 is configured to obtain at least one basic semantic information cluster and at least one question and answer instance; correspondingly, the second obtaining module is specifically configured to: update at least one basic semantic information cluster according to the at least one question and answer instance to obtain multiple semantic information Information cluster.
  • the first obtaining module 901 is specifically configured to: determine the similarity between each semantic information in at least one question and answer instance and each semantic information in at least one basic semantic information cluster; The instances are divided into basic semantic information clusters that have a similarity with the question answering instance in at least one basic semantic information cluster that is greater than a third preset similarity.
  • it also includes:
  • the fourth obtaining module 905 is used to obtain the selected times of each query question in a plurality of semantic information clusters.
  • the second sending module 906 is used to send the selected times of each query question in the multiple semantic information clusters to the robot.
  • the robot answering device provided in the present application can execute the above-mentioned server-side corresponding robot answering method.
  • the present application also provides a robot and a readable storage medium.
  • FIG. 10 is a schematic diagram of the structure of the robot provided by this application.
  • the components shown herein, their connections and relationships, and their functions are merely examples, and do not limit the implementation of the application described and/or required herein.
  • the robot includes a processor 1001 and a memory 1002, and each component is connected to each other by using different buses, and can be installed on a common motherboard or installed in other ways as needed.
  • the processor 1001 may process instructions executed in the robot, including instructions stored in or on the memory to display graphic information on an external input/output device (such as a display device coupled to an interface).
  • an external input/output device such as a display device coupled to an interface.
  • multiple processors and/or multiple buses can be used with multiple memories and multiple memories.
  • a processor 1001 is taken as an example.
  • the memory 1002 can be used to store non-transitory software programs, non-transitory computer-executable programs, and modules, such as program instructions/modules corresponding to the robot response method in the embodiment of the present application (for example, The first acquisition module 801, the extraction module 802 and the matching module 803 shown in FIG. 8).
  • the processor 1001 executes various functional applications and data processing of the server by running non-transitory software programs, instructions, and modules stored in the memory 1002, that is, realizing the robot response method in the foregoing method embodiment.
  • the robot may also include: an input device 1003 and an output device 1004.
  • the processor 1001, the memory 1002, the input device 1003, and the output device 1004 may be connected through a bus or other methods. In FIG. 10, the connection through a bus is taken as an example.
  • the input device 1003 can receive input digital or character information, and generate key signal input related to the user settings and function control of the robot, such as a touch screen, a keypad, a mouse, or multiple mouse buttons, trackballs, joysticks and other input devices .
  • the output device 1004 may be an output device such as a display device of a robot.
  • the display device may include, but is not limited to, a liquid crystal display (LCD), a light emitting diode (LED) display, and a plasma display. In some embodiments, the display device may be a touch screen.
  • the robots in the embodiments of the present application can be used to implement the technical solutions in the foregoing method embodiments of the present application.
  • the implementation principles and technical effects of the robots are similar, and will not be repeated here.
  • FIG. 11 is a schematic diagram of the structure of the server provided by this application.
  • the components shown herein, their connections and relationships, and their functions are merely examples, and do not limit the implementation of the application described and/or required herein.
  • the server includes a processor 1101 and a memory 1102.
  • the various components are connected to each other by using different buses, and can be installed on a common motherboard or installed in other ways as required.
  • the processor 1101 may process instructions executed in the server, including instructions stored in or on the memory to display graphic information on an external input/output device (such as a display device coupled to an interface).
  • an external input/output device such as a display device coupled to an interface.
  • multiple processors and/or multiple buses can be used with multiple memories and multiple memories.
  • a processor 1101 is taken as an example.
  • the memory 1102 can be used to store non-transitory software programs, non-transitory computer-executable programs, and modules, such as program instructions/modules corresponding to the robot response method in the embodiment of the present application (for example, The first acquiring module 901, the first sending module 902, and the second acquiring module 903 shown in FIG. 9).
  • the processor 1101 executes various functional applications and data processing of the server by running non-transitory software programs, instructions, and modules stored in the memory 1102, that is, realizing the robot response method in the foregoing method embodiment.
  • the server may also include: an input device 1103 and an output device 1104.
  • the processor 1101, the memory 1102, the input device 1103, and the output device 1104 may be connected by a bus or in other ways. In FIG. 11, the connection by a bus is taken as an example.
  • the input device 1103 can receive input digital or character information, and generate key signal input related to the user settings and function control of the server, such as a touch screen, a small keyboard, a mouse, or multiple mouse buttons, trackballs, joysticks and other input devices .
  • the output device 1104 may be an output device such as a display device of a server.
  • the display device may include, but is not limited to, a liquid crystal display (LCD), a light emitting diode (LED) display, and a plasma display. In some embodiments, the display device may be a touch screen.
  • the server in the embodiment of the present application may be used to execute the technical solutions in the foregoing method embodiments of the present application, and the implementation principles and technical effects are similar, and will not be repeated here.
  • An embodiment of the present application also provides a computer-readable storage medium that stores computer-executable instructions, and the computer-executable instructions are used to implement any one of the aforementioned robot response methods when executed by a processor.
  • the embodiment of the present application also provides a computer program product, the program product includes a computer-executable instruction, and the computer-executable instruction is used to implement any one of the above-mentioned robot response methods when executed by a processor.
  • the disclosed system, device, and method can be implemented in other ways.
  • the device embodiments described above are merely illustrative, for example, the division of the units is only a logical function division, and there may be other divisions in actual implementation, for example, multiple units or components may be combined or It can be integrated into another system, or some features can be ignored or not implemented.
  • the displayed or discussed mutual coupling or direct coupling or communication connection may be indirect coupling or communication connection through some interfaces, devices or units, and may be in electrical, mechanical or other forms.
  • the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the objectives of the solutions of the embodiments.
  • the functional units in the various embodiments of the present application may be integrated into one processing unit, or each unit may exist alone physically, or two or more units may be integrated into one unit.
  • the above-mentioned integrated unit can be implemented in the form of hardware or software functional unit.

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Data Mining & Analysis (AREA)
  • General Engineering & Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Computational Linguistics (AREA)
  • Mathematical Physics (AREA)
  • Artificial Intelligence (AREA)
  • Databases & Information Systems (AREA)
  • General Health & Medical Sciences (AREA)
  • Computing Systems (AREA)
  • Software Systems (AREA)
  • Health & Medical Sciences (AREA)
  • Evolutionary Computation (AREA)
  • Biomedical Technology (AREA)
  • Biophysics (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Molecular Biology (AREA)
  • Human Computer Interaction (AREA)
  • Computer Networks & Wireless Communication (AREA)
  • Signal Processing (AREA)
  • Acoustics & Sound (AREA)
  • Audiology, Speech & Language Pathology (AREA)
  • Information Retrieval, Db Structures And Fs Structures Therefor (AREA)

Abstract

本申请提供一种机器人应答方法、装置、设备及存储介质。该方法包括：机器人获取当前查询语音，提取当前查询语音的语义信息，对当前查询语音的语义信息和预先存储的多个语义信息簇进行匹配，得到匹配到的目标语义信息簇，每个语义信息簇包括：至少一个问答实例，每个问答实例包括：一个历史查询语音对应的语义信息和在历史查询语音对应的查询列表中被选择的查询问题，机器人获取目标语义信息簇中各个查询问题的被选择次数，并根据各个查询问题的被选择次数确定当前查询语音对应的目标查询问题，输出目标查询问题对应的查询响应。本申请的方法，达到了对没有明确答案的问题的准确答复，从而提高获取准确答案的效率。

Description

一种机器人应答方法、装置、设备及存储介质
本申请要求于2019年12月27日提交中国专利局、申请号为2019113765184、申请名称为“一种机器人应答方法、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
本申请涉及通信技术领域，尤其涉及一种机器人应答方法、装置、设备及存储介质。
背景技术
聊天机器人系统是一种借助于通讯手段能够时时刻刻在线、并通过自然语言与人进行沟通交流的系统，聊天机器人系统内部存储有大量的问题和对应的答案，当用户输入问题后，聊天机器人会根据问题寻找相应的回答反馈给用户。
然而聊天机器人系统内部存储的问答是有限的，有些用户输入的问题系统内部并没有明确的答案。现有技术中针对没有明确答案的用户输入的问题，聊天机器人会给出一个列表回复，列表中的内容跟用户的问题具有一定的相似度，这个列表是从数据库检索出的，用户需要在给出的列表中点选问题以获得需要的答案。
采用现有技术的方法，用户在与机器人交互的过程中获得问题准确答案的效率较低。
发明内容
本申请提供一种机器人应答方法、装置、设备及存储介质，从而提高获取准确答案的效率。
第一方面，本申请提供一种机器人应答方法，包括：获取当前查询语音；提取当前查询语音的语义信息；对当前查询语音的语义信息和预先存储的多个语义信息簇进行匹配，得到匹配到的目标语义信息簇，每个语义信息簇包括：至少一个问答实例，每个问答实例包括：一个历史查询语音对应的语义 信息和在历史查询语音对应的查询列表中被选择的查询问题；获取目标语义信息簇中各个查询问题的被选择次数，并根据各个查询问题的被选择次数确定当前查询语音对应的目标查询问题，输出目标查询问题对应的查询响应。本申请通过将语音转化为语义信息，将语义信息与预先存储的语义信息簇进行匹配，通过历史查询问题的被选择次数获得当前查询问题的最优答案，达到了对没有明确答案的问题的准确答复，从而提高获取准确答案的效率。
可选的，对当前查询语音的语义信息和预先存储的多个语义信息簇进行匹配，包括：确定当前查询语音的语义信息和多个语义信息簇中的各个历史查询语音对应的语义信息的相似度；若当前查询语音的语义信息和一个语义信息簇中的历史查询语音对应的语义信息的相似度大于第一预设相似度，则将该语义信息簇作为目标语义信息簇。即实现了根据语义信息的相似度对语义信息和语义信息簇的匹配。
可选的，根据各个查询问题的被选择次数确定当前查询语音对应的目标查询问题，包括：在各个查询问题中，将被选择次数大于预设次数的查询问题确定为目标查询问题。即无需用户选择问题，仅需要根据查询问题的被选择次数即可确定目标查询问题，从而提高了获取答案的效率。
可选的，还包括：接收服务器发送的多个语义信息簇。
第二方面，本申请提供一种机器人应答方法，包括：获取多个语义信息簇，每个语义信息簇包括：至少一个问答实例，每个问答实例包括：一个历史查询语音对应的语义信息和在历史查询语音对应的查询列表中被选择的查询问题；向机器人发送多个语义信息簇，以使机器人对多个语义信息簇和当前查询语音的语义信息进行匹配，得到匹配到的目标语义信息簇，并根据目标语义信息簇中各个查询问题的被选择次数确定当前查询语音对应的目标查询问题，输出目标查询问题对应的查询响应。从而达到了对没有明确答案的问题的准确答复，进而提高了获取准确答案的效率。
可选的，获取多个语义信息簇之前，还包括：获取多个问答实例；相应的，获取多个语义信息簇，包括：确定多个问答实例中每两个问答实例中的语义信息的相似度；将相似度大于第二预设相似度的问答实例归为一个语义信息簇。将语义信息簇进行分类，根据相似度确定是否为一个语义信息簇。或者，获取多个语义信息簇，之前还包括：获取至少一个基础语义信息簇以及至少一个问答实例；相应的，获取多个语义信息簇，包括：根据至少一个 问答实例更新至少一个基础语义信息簇，以得到多个语义信息簇。即通过这两种方法均可以将相似的问答实例作为一个语义信息簇，而相似的问答实例中会存在相似或相同的被选择的查询问题，以使机器人可以统计各个查询问题的被选择次数，进而可以根据各个查询问题的被选择次数确定当前查询语音对应的目标查询问题，输出目标查询问题对应的查询响应，从而提高了获取准确答案的效率。
可选的，根据至少一个问答实例更新至少一个基础语义信息簇，以得到多个语义信息簇，包括：确定至少一个问答实例中各个语义信息和至少一个基础语义信息簇中各个语义信息的相似度；针对至少一个问答实例中每一个问答实例，将问答实例划分至在至少一个基础语义信息簇中与问答实例的相似度大于第三预设相似度的基础语义信息簇中，从而实现对语音信息簇的动态更新。
可选的，还包括：获取多个语义信息簇中各个查询问题的被选择次数；向机器人发送多个语义信息簇中各个查询问题的被选择次数。以使机器人可以根据各个查询问题的被选择次数，确定当前查询语音对应的目标查询问题，输出目标查询问题对应的查询响应。进而提高了获取准确答案的效率。
本申请还提供一种机器人应答装置、设备、可读存储介质以及计算机程序产品，其效果可参考上述方法部分对应的效果，下面对此不再赘述。
第三方面，本申请提供一种机器人应答装置，包括：第一获取模块、提取模块、匹配模块、第二获取模块、确定模块和输出模块，第一获取模块用于获取当前查询语音；提取模块用于提取当前查询语音的语义信息；匹配模块用于对当前查询语音的语义信息和预先存储的多个语义信息簇进行匹配，得到匹配到的目标语义信息簇，每个语义信息簇包括：至少一个问答实例，每个问答实例包括：一个历史查询语音对应的语义信息和在历史查询语音对应的查询列表中被选择的查询问题；第二获取模块用于获取目标语义信息簇中各个查询问题的被选择次数；确定模块用于根据各个查询问题的被选择次数确定当前查询语音对应的目标查询问题；输出模块用于输出目标查询问题对应的查询响应。
可选的，匹配模块具体用于：确定当前查询语音的语义信息和多个语义信息簇中的各个历史查询语音对应的语义信息的相似度；若当前查询语音的语义信息和一个语义信息簇中的历史查询语音对应的语义信息的相似度大于 第一预设相似度，则将该语义信息簇作为目标语义信息簇。
可选的，确定模块具体用于：在各个查询问题中，将被选择次数大于预设次数的查询问题确定为目标查询问题。
可选的，还包括：接收模块，用于接收服务器发送的多个语义信息簇。
第四方面，本申请提供一种机器人应答装置，包括：第一获取模块，用于获取多个语义信息簇，每个语义信息簇包括：至少一个问答实例，每个问答实例包括：一个历史查询语音对应的语义信息和在历史查询语音对应的查询列表中被选择的查询问题；第一发送模块，用于向机器人发送多个语义信息簇。
可选的，还包括：第二获取模块，用于获取多个问答实例；相应的，第一获取模块具体用于：确定多个问答实例中每两个问答实例中的语义信息的相似度；将相似度大于第二预设相似度的问答实例归为一个语义信息簇。
可选的，还包括：第三获取模块，用于获取至少一个基础语义信息簇以及至少一个问答实例；相应的，第一获取模块具体用于：根据至少一个问答实例更新至少一个基础语义信息簇，以得到多个语义信息簇。
可选的，第一获取模块具体用于：确定至少一个问答实例中各个语义信息和至少一个基础语义信息簇中各个语义信息的相似度；针对至少一个问答实例中每一个问答实例，将问答实例划分至在至少一个基础语义信息簇中与问答实例的相似度大于第三预设相似度的基础语义信息簇中。
可选的，还包括：第四获取模块，用于获取多个语义信息簇中各个查询问题的被选择次数；第二发送模块，用于向机器人发送多个语义信息簇中各个查询问题的被选择次数。
第五方面，本申请实施例提供一种存储有计算机指令的非瞬时计算机可读存储介质，计算机指令用于使计算机执行如第一方面或第一方面的可选方式的任一项的应用于机器人应答方法。
第六方面，本申请实施例提供一种存储有计算机指令的非瞬时计算机可读存储介质，计算机指令用于使计算机执行如第二方面或第二方面的可选方式的任一项的应用于机器人应答方法。
第七方面，本申请实施例提供一种计算机程序产品，该产品包括：计算机指令，该计算机指令用于使计算机执行如第一方面或第一方面的可选方式的任一项的应用于机器人应答方法。
第八方面，本申请实施例提供一种计算机程序产品，该产品包括：计算机指令，该计算机指令用于使计算机执行如第二方面或第二方面的可选方式的任一项的应用于机器人应答方法。
本申请提供的一种机器人应答方法、装置、设备及存储介质，通过机器人获取当前查询语音，提取当前查询语音的语义信息，对当前查询语音的语义信息和预先存储的多个语义信息簇进行匹配，得到匹配到的目标语义信息簇，每个语义信息簇包括：至少一个问答实例，每个问答实例包括：一个历史查询语音对应的语义信息和在历史查询语音对应的查询列表中被选择的查询问题，机器人获取目标语义信息簇中各个查询问题的被选择次数，并根据各个查询问题的被选择次数确定当前查询语音对应的目标查询问题，输出目标查询问题对应的查询响应，达到了对没有明确答案的问题的准确答复，从而提高了获取准确答案的效率。
附图说明
此处的附图被并入说明书中并构成本说明书的一部分，示出了符合本公开的实施例，并与说明书一起用于解释本公开的原理。
图1为本申请提供的一种应用场景的示意图；
图2为本申请提供的一种机器人应答方法的交互流程图；
图3为机器人的查询列表示意图；
图4A为本申请一实施例提供的一种界面示意图；
图4B为本申请一实施例提供的另一种界面示意图；
图5为本申请提供的另一种机器人应答方法的交互流程图；
图6为本申请提供的再一种机器人应答方法的交互流程图；
图7为本申请提供的又一种机器人应答方法的交互流程图；
图8为本申请提供的一种机器人应答装置的结构示意图；
图9为本申请提供的另一种机器人应答装置的结构示意图；
图10为本申请提供的机器人的结构示意图；
图11为本申请提供的服务器的结构示意图。
通过上述附图，已示出本公开明确的实施例，后文中将有更详细的描述。这些附图和文字描述并不是为了通过任何方式限制本公开构思的范围，而是通过参考特定实施例为本领域技术人员说明本公开的概念。
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清楚、完整地描述，显然，所描述的实施例仅仅是本申请一部分实施例，而不是全部的实施例。基于本申请中的实施例，本领域普通技术人员在没有做出创造性劳动前提下所获得的所有其他实施例，都属于本申请保护的范围。
示例性地，图1为本申请提供的一种应用场景示意图。如图1所示，机器人001用以实现与用户的交互应答，该机器人中存储有多个语义信息簇和大数据搜索数据库(Elastic Search，ES)，其中，每个语义信息簇包括：至少一个问答实例，每个所述问答实例包括：一个历史查询语音对应的语义信息和在所述历史查询语音对应的查询列表中被选择的查询问题；ES数据库中存储有查询问题对应的查询响应，即查询问题对应的答案。服务器002用来建立多个语义信息簇，并将多个语义信息簇发送给机器人001，可选的，服务器002还可以训练神经网络模型，并将训练好的神经网络模型发送给机器人001，以使机器人001可以通过该神经网络模型确定查询语音对应的语义信息。
如上所述，现有技术中针对没有明确答案的用户输入的问题，聊天机器人会给出一个列表回复，列表中的内容跟用户的问题具有一定的相似度，这个列表是从数据库检索出的，用户需要在给出的列表中点选问题以获得需要的答案。采用现有技术的方法，用户在与机器人交互的过程中获得问题准确答案的效率较低。
为了解决上述技术问题，本申请提供一种机器人应答方法、装置、设备及存储介质。本申请的主旨思想是：机器人确定当前查询语音对应的目标语音信息簇，根据目标语义信息簇中各个查询问题的被选择次数，确定目标查询问题，并输出目标查询问题对应的查询响应。
下面对本申请技术方案进行详细阐述：
图2为本申请提供的一种机器人应答方法的交互流程图。该方法涉及的网元包括：机器人和服务器，如图2所示，该方法包括如下步骤：
步骤S201：服务器获取多个语义信息簇。
每个语义信息簇包括：至少一个问答实例，每个问答实例包括：一个历史查询语音对应的语义信息和在历史查询语音对应的查询列表中被选择 的查询问题。
例如：图3为机器人的查询列表示意图，如图3所示，当用户的输入没有明确答案的问题后，机器人将与用户的语音具有一定相似度的历史查询问题以列表的形式显示给用户，其中，同一列表中的所有历史查询问题包含有相同或相似的语义信息。
步骤S202：服务器向机器人发送多个语义信息簇。
步骤S203：机器人获取当前查询语音。
步骤S204：机器人提取当前查询语音的语义信息。
步骤S205：机器人对当前查询语音的语义信息和预先存储的多个语义信息簇进行匹配，得到匹配到的目标语义信息簇。
步骤S206：机器人获取目标语义信息簇中各个查询问题的被选择次数，并根据各个查询问题的被选择次数确定当前查询语音对应的目标查询问题，输出目标查询问题对应的查询响应。
针对步骤S203和步骤204进行如下说明：
可选的，机器人通过麦克风采集获取用户的当前查询语音，并通过神经网络模型提取当前查询语音的语义信息。其中该神经网络模型由双向编码器表征(BERT)+卷积神经网络(Convolutional Neural Network，CNN)+全连接层(fully connected layers，FC)结构组成。服务器可以对该神经网络模型进行训练，并将训练好的神经网络模型发送给机器人。以使机器人可以将当前查询语音作为该神经网络模型的输入，以得到当前查询语音的语义信息。服务器对神经网络模型的训练过程实质是：服务器对神经网络模型中涉及的参数进行训练。
上述BERT模型的输入是查询语音，得到该查询语音的第一语义信息。而CNN，是一种前馈神经网络，它的人工神经元可以响应一部分覆盖范围内的周围单元，它包括卷积层(convolutional layer)和池化层(pooling layer)，利用卷积层对第一语义信息进行卷积运算，获得所有或部分隐含层的输出，得到第二语义信息，利用池化层对第二语义信息进行降维，得到查询语音的第三语义信息。需要说明的是，上述任一语义信息都可以被称为语义向量或者语义矩阵。FC层在整个神经网络模型中起到“分类器”的作用，将第三语义信息和第二语义信息进行加权求和，即经过全连接层，输出查询语音对应的最终的语义信息。
针对步骤S205和步骤S206进行如下说明：
可选的，机器人在提取到当前查询语音的语义信息之后，先确定该语义信息是否对应有唯一的查询问题，如果存在，不执行步骤S205和步骤S206，如果该语义信息对应有多个查询问题，则执行步骤S205和步骤S206。
可选的，上述多个语义信息簇可以预先存储在机器人本地存储空间中，也可以存储在云端存储空间中，本申请对此不做限制。
所谓“机器人对当前查询语音的语义信息和预先存储的多个语义信息簇进行匹配”指的是机器人对当前查询语音的语义信息和每个语义信息簇中的语义信息进行匹配。
步骤206：机器人获取目标语义信息簇中各个查询问题的被选择次数，并根据各个查询问题的被选择次数确定当前查询语音对应的目标查询问题，输出目标查询问题对应的查询响应。
可选的，机器人从服务器获取多个语义信息簇中各个查询问题的被选择次数。或者，机器人统计多个语义信息簇中各个查询问题的被选择次数。
可选的，在各个查询问题中，确定将被选择次数大于预设次数的查询问题确定为目标查询问题。或者，每个查询问题对应有优先级，机器人可以选择优先级最高的查询问题，作为目标查询问题。上述预设次数可以根据实际情况设置，比如是100或者1000等，本申请对此不作限制。
机器人在ES数据库中查询目标查询问题对应的查询响应，并输出给用户。ES数据库是一个分布式、高扩展、高实时的搜索与数据分析引擎。它能很方便的使大量数据具有搜索、分析和探索的能力。
上述机器人根据各个查询问题的被选择次数确定当前查询语音对应的目标查询问题可以是一个或者是多个。而每个查询问题对应一个查询响应。因此，机器人最终可以呈现一个或者多个查询响应。例如：图4A为本申请一实施例提供的界面示意图，如图4A所示，机器人显示的是关于目标查询问题“使用闪付有哪些优惠政策？”对应的查询响应“优惠活动不定期更新，具体以页面显示为准”。图4B为本申请另一实施例提供的界面示意图，如图4B所示，机器人显示的是关于目标查询问题“平台都有哪些优惠政策？”对应的查询响应“A1：优惠活动不定期更新，具体以页面显示为准。A2：白条激活时会为新用户赠送新手优惠券礼包。A3：通过不同渠道激活白条会有不同的优惠，具体获得的优惠请以实际情况为准。”
图5为本申请提供的另一种机器人应答方法的交互流程图图，上述步骤S205具体包括：
S205A：机器人确定当前查询语音的语义信息和多个语义信息簇中的各个历史查询语音对应的语义信息的相似度。
S205B：若当前查询语音的语义信息和一个语义信息簇中的历史查询语音对应的语义信息的相似度大于第一预设相似度，则机器人将该语义信息簇作为目标语义信息簇。
其中，如上所述，任一个语义信息也可以被称为语义向量。因此，机器人通过如下方式确定当前查询语音的语义向量和任一个历史查询语音对应的语义向量的相似度：对两个语义向量采用点乘方式，以得到这两个语义向量的相似度，其中点乘结果越大，则表示两个语义向量的相似度越高。或者，计算两个语义向量的距离，该距离越大，则表示两个语义向量的相似度越低。
可选的，上述第一预设相似度可以根据实际情况设置，比如可以是0.8、0.9等，本申请对此不做限制。
本实施例机器人通过将语音转化为语义信息，根据当前查询语音的语义信息和多个语义信息簇中的历史查询语音对应的语义信息的相似度确定目标语义信息簇，通过目标语义信息簇中的历史查询问题的被选择次数获得当前查询问题的最优答案，达到了对没有明确答案的问题的准确答复，从而提高获取准确答案的效率。
下面将介绍服务器获取多个语义信息簇的方法。
可选方式一：图6为本申请提供的再一种机器人应答方法的交互流程图，进一步的，在步骤S201之前，还包括：
步骤S200：服务器获取多个问答实例。
相应的，步骤S201包括：
步骤S201A：服务器确定多个问答实例中每两个问答实例中的语义信息的相似度。
步骤S201B：服务器将相似度大于第二预设相似度的问答实例归为一个语义信息簇。
针对步骤S200-S201B进行如下说明：
本申请实施例中，提供了一种获取语义信息簇的方式，这种方式基于 基于密度的聚类算法(Density-Based Spatial Clustering of Applications with Noise，DBScan)，DBSCAN将簇定义为密度相连的点的最大集合，能够把具有足够高密度的区域划分为簇，并可在噪声的空间数据库中发现任意形状的聚类。
本实施例中，语义信息可以被称为语义向量，而两个语义信息的相似度可以是这两个语义向量的点乘结果。服务器首先获取多个问答实例，确定每两个问答实例的语义相似度，如果两个问答实例的相似度大于第二预设相似度，就确定这两个问答实例归属于同一语义信息簇，进而获得多个语义信息簇，每个语义信息簇中的问题实例具有足够高的相似度。
可选的，上述第二预设相似度可以根据实际情况设置，比如可以是0.8、0.9等，本申请对此不做限制。
可选的，服务器可以周期性的获取语义信息簇，例如：在每天夜里12点，服务器定时回溯过去一个星期的问答实例，以建立得到多个语义信息簇。
可选方式二：图7为本申请提供的又一种机器人应答方法的交互流程图，进一步的，在步骤S201之前，还包括：
步骤S200A：服务器获取至少一个基础语义信息簇以及至少一个问答实例。
相应的，步骤S201包括：
步骤S201C：服务器确定至少一个问答实例中各个语义信息和至少一个基础语义信息簇中各个语义信息的相似度。
步骤S201D：服务器针对至少一个问答实例中每一个问答实例，将问答实例划分至在至少一个基础语义信息簇中与问答实例的相似度大于第三预设相似度的基础语义信息簇中。
所谓基础语义信息簇指的是当前已建立的语义信息簇，而服务器可以在该语义信息簇的基础上，生成新的语义信息簇。
可选的，上述第三预设相似度可以根据实际情况设置，比如可以是0.8、0.9等，本申请对此不做限制。
可选的，服务器可以周期性的获取至少一个问题实例，例如：每隔一分钟回溯过往一分钟的问答实例，以对基础语义信息簇进行更新。
在本申请中，对问答实例和基础语义信息簇的获取频率和时间，对此 不做具体限制。
综上，在本申请中，服务器可以通过上述两种可选方式获取多个语义信息簇，并将多个语义信息簇发送给机器人，以使机器人通过该多个语义信息簇，确定当前查询语音对应的目标语义信息簇。其中，服务器可以周期性的获取多个语义信息簇，即多个语义信息簇是动态变化的，从而可以更好的为用户提供准确的答案。
图8为本申请提供的一种机器人应答装置的结构示意图。如图8所示，该机器人应答装置包括：
第一获取模块801，用于获取当前查询语音。
提取模块802，用于提取当前查询语音的语义信息。
匹配模块803，用于对当前查询语音的语义信息和预先存储的多个语义信息簇进行匹配，得到匹配到的目标语义信息簇，每个语义信息簇包括：至少一个问答实例，每个问答实例包括：一个历史查询语音对应的语义信息和在历史查询语音对应的查询列表中被选择的查询问题。
第二获取模块804，用于获取目标语义信息簇中各个查询问题的被选择次数。
确定模块805，用于根据各个查询问题的被选择次数确定当前查询语音对应的目标查询问题。
输出模块806，用于输出目标查询问题对应的查询响应。
可选的，匹配模块803具体用于：确定当前查询语音的语义信息和多个语义信息簇中的各个历史查询语音对应的语义信息的相似度；若当前查询语音的语义信息和一个语义信息簇中的历史查询语音对应的语义信息的相似度大于第一预设相似度，则将该语义信息簇作为目标语义信息簇。
可选的，确定模块805具体用于：在各个查询问题中，将被选择次数大于预设次数的查询问题确定为目标查询问题。
可选的，还包括：
接收模块807，用于接收服务器发送的多个语义信息簇。
本申请提供的机器人应答装置，可以执行上述机器人侧对应的机器人应答方法，其内容和效果可参考方法实施例部分，对此不再赘述。
图9为本申请提供的另一种机器人应答装置的结构示意图。如图9所示，该机器人应答装置包括：
第一获取模块901，用于获取多个语义信息簇，每个语义信息簇包括：至少一个问答实例，每个问答实例包括：一个历史查询语音对应的语义信息和在历史查询语音对应的查询列表中被选择的查询问题。
第一发送模块902，用于向机器人发送多个语义信息簇。
可选的，还包括：
第二获取模块903，用于获取多个问答实例；相应的，第二获取模块具体用于：确定多个问答实例中每两个问答实例中的语义信息的相似度；将相似度大于第二预设相似度的问答实例归为一个语义信息簇。
可选的，还包括：
第三获取模块904，用于获取至少一个基础语义信息簇以及至少一个问答实例；相应的，第二获取模块具体用于：根据至少一个问答实例更新至少一个基础语义信息簇，以得到多个语义信息簇。
可选的，第一获取模块901具体用于：确定至少一个问答实例中各个语义信息和至少一个基础语义信息簇中各个语义信息的相似度；针对至少一个问答实例中每一个问答实例，将问答实例划分至在至少一个基础语义信息簇中与问答实例的相似度大于第三预设相似度的基础语义信息簇中。
可选的，还包括：
第四获取模块905，用于获取多个语义信息簇中各个查询问题的被选择次数。
第二发送模块906，用于向机器人发送多个语义信息簇中各个查询问题的被选择次数。
本申请提供的机器人应答装置，可以执行上述服务器侧对应的机器人应答方法，其内容和效果可参考方法实施例部分，对此不再赘述。
根据本申请的实施例，本申请还提供了一种机器人和一种可读存储介质。
图10为本申请提供的机器人的结构示意图。本文所示的部件、它们的连接和关系、以及它们的功能仅仅作为示例，并且不限制本文中描述的和/或者要求的本申请的实现。
如图10所示，该机器人包括：处理器1001和存储器1002，各个部件利用不同的总线互相连接，并且可以被安装在公共主板上或者根据需要以其它方式安装。处理器1001可以对在机器人内执行的指令进行处理，包括 存储在存储器中或者存储器上以在外部输入/输出装置(诸如，耦合至接口的显示设备)上显示的图形信息的指令。在其它实施方式中，若需要，可以将多个处理器和/或多条总线与多个存储器和多个存储器一起使用。图10中以一个处理器1001为例。
存储器1002作为一种非瞬时计算机可读存储介质，可用于存储非瞬时软件程序、非瞬时计算机可执行程序以及模块，如本申请实施例中的机器人应答的方法对应的程序指令/模块(例如，附图8所示的第一获取模块801、提取模块802和匹配模块803)。处理器1001通过运行存储在存储器1002中的非瞬时软件程序、指令以及模块，从而执行服务器的各种功能应用以及数据处理，即实现上述方法实施例中的机器人应答的方法。
机器人还可以包括：输入装置1003和输出装置1004。处理器1001、存储器1002、输入装置1003和输出装置1004可以通过总线或者其他方式连接，图10中以通过总线连接为例。
输入装置1003可接收输入的数字或字符信息，以及产生与机器人的用户设置以及功能控制有关的键信号输入，例如触摸屏、小键盘、鼠标、或者多个鼠标按钮、轨迹球、操纵杆等输入装置。输出装置1004可以是机器人的显示设备等输出设备。该显示设备可以包括但不限于，液晶显示器(LCD)、发光二极管(LED)显示器和等离子体显示器。在一些实施方式中，显示设备可以是触摸屏。
本申请实施例的机器人，可以用于执行本申请上述各方法实施例中的技术方案，其实现原理和技术效果类似，此处不再赘述。
图11为本申请提供的服务器的结构示意图。本文所示的部件、它们的连接和关系、以及它们的功能仅仅作为示例，并且不限制本文中描述的和/或者要求的本申请的实现。
如图11所示，该服务器包括：处理器1101和存储器1102，各个部件利用不同的总线互相连接，并且可以被安装在公共主板上或者根据需要以其它方式安装。处理器1101可以对在服务器内执行的指令进行处理，包括存储在存储器中或者存储器上以在外部输入/输出装置(诸如，耦合至接口的显示设备)上显示的图形信息的指令。在其它实施方式中，若需要，可以将多个处理器和/或多条总线与多个存储器和多个存储器一起使用。图11中以一个处理器1101为例。
存储器1102作为一种非瞬时计算机可读存储介质，可用于存储非瞬时软件程序、非瞬时计算机可执行程序以及模块，如本申请实施例中的机器人应答的方法对应的程序指令/模块(例如，附图9所示的第一获取模块901、第一发送模块902和第二获取模块903)。处理器1101通过运行存储在存储器1102中的非瞬时软件程序、指令以及模块，从而执行服务器的各种功能应用以及数据处理，即实现上述方法实施例中的机器人应答的方法。
服务器还可以包括：输入装置1103和输出装置1104。处理器1101、存储器1102、输入装置1103和输出装置1104可以通过总线或者其他方式连接，图11中以通过总线连接为例。
输入装置1103可接收输入的数字或字符信息，以及产生与服务器的用户设置以及功能控制有关的键信号输入，例如触摸屏、小键盘、鼠标、或者多个鼠标按钮、轨迹球、操纵杆等输入装置。输出装置1104可以是服务器的显示设备等输出设备。该显示设备可以包括但不限于，液晶显示器(LCD)、发光二极管(LED)显示器和等离子体显示器。在一些实施方式中，显示设备可以是触摸屏。
本申请实施例的服务器，可以用于执行本申请上述各方法实施例中的技术方案，其实现原理和技术效果类似，此处不再赘述。
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中存储有计算机执行指令，计算机执行指令被处理器执行时用于实现上述任一所述的机器人应答方法。
本申请实施例还提供一种计算机程序产品，该程序产品包括计算机执行指令，计算机执行指令被处理器执行时用于实现上述任一所述的机器人应答方法。
在本申请所提供的几个实施例中，应该理解到，所揭露的系统，装置和方法，可以通过其它的方式实现。例如，以上所描述的装置实施例仅仅是示意性的，例如，所述单元的划分，仅仅为一种逻辑功能划分，实际实现时可以有另外的划分方式，例如多个单元或组件可以结合或者可以集成到另一个系统，或一些特征可以忽略，或不执行。另一点，所显示或讨论的相互之间的耦合或直接耦合或通信连接可以是通过一些接口，装置或单元的间接耦合或通信连接，可以是电性，机械或其它的形式。
所述作为分离部件说明的单元可以是或者也可以不是物理上分开的， 作为单元显示的部件可以是或者也可以不是物理单元，即可以位于一个地方，或者也可以分布到多个网络单元上。可以根据实际的需要选择其中的部分或者全部单元来实现本实施例方案的目的。
另外，在本申请各个实施例中的各功能单元可以集成在一个处理单元中，也可以是各个单元单独物理存在，也可以两个或两个以上单元集成在一个单元中。上述集成的单元既可以采用硬件的形式实现，也可以采用软件功能单元的形式实现。
本领域技术人员在考虑说明书及实践这里公开的发明后，将容易想到本公开的其它实施方案。本申请旨在涵盖本公开的任何变型、用途或者适应性变化，这些变型、用途或者适应性变化遵循本公开的一般性原理并包括本公开未公开的本技术领域中的公知常识或惯用技术手段。说明书和实施例仅被视为示例性的，本公开的真正范围和精神由下面的权利要求书指出。
应当理解的是，本公开并不局限于上面已经描述并在附图中示出的精确结构，并且可以在不脱离其范围进行各种修改和改变。本公开的范围仅由所附的权利要求书来限制。

Claims (22)

  1. 一种机器人应答方法，其特征在于，包括：
    获取当前查询语音；
    提取所述当前查询语音的语义信息；
    对所述当前查询语音的语义信息和预先存储的多个语义信息簇进行匹配，得到匹配到的目标语义信息簇，每个所述语义信息簇包括：至少一个问答实例，每个所述问答实例包括：一个历史查询语音对应的语义信息和在所述历史查询语音对应的查询列表中被选择的查询问题；
    获取所述目标语义信息簇中各个查询问题的被选择次数，并根据所述各个查询问题的被选择次数确定所述当前查询语音对应的目标查询问题，输出所述目标查询问题对应的查询响应。
  2. 根据权利要求1所述的方法，其特征在于，所述对所述当前查询语音的语义信息和预先存储的多个语义信息簇进行匹配，包括：
    确定所述当前查询语音的语义信息和所述多个语义信息簇中的各个历史查询语音对应的语义信息的相似度；
    若所述当前查询语音的语义信息和一个语义信息簇中的历史查询语音对应的语义信息的相似度大于第一预设相似度，则将该语义信息簇作为所述目标语义信息簇。
  3. 根据权利要求1或2所述的方法，其特征在于，所述根据所述各个查询问题的被选择次数确定所述当前查询语音对应的目标查询问题，包括：
    在所述各个查询问题中，将被选择次数大于预设次数的查询问题确定为所述目标查询问题。
  4. 根据权利要求1或2所述的方法，其特征在于，还包括：
    接收服务器发送的所述多个语义信息簇。
  5. 一种机器人应答方法，其特征在于，包括：
    获取多个语义信息簇，每个所述语义信息簇包括：至少一个问答实例，每个所述问答实例包括：一个历史查询语音对应的语义信息和在所述历史查询语音对应的查询列表中被选择的查询问题；
    向机器人发送所述多个语义信息簇。
  6. 根据权利要求5所述的方法，其特征在于，所述获取多个语义信息簇 之前，还包括：
    获取多个所述问答实例；
    相应的，所述获取多个语义信息簇，包括：
    确定多个所述问答实例中每两个问答实例中的语义信息的相似度；
    将所述相似度大于第二预设相似度的问答实例归为一个语义信息簇。
  7. 根据权利要求5所述的方法，其特征在于，所述获取多个语义信息簇之前，还包括：
    获取至少一个基础语义信息簇以及至少一个所述问答实例；
    相应的，所述获取多个语义信息簇，包括：
    根据至少一个所述问答实例更新所述至少一个基础语义信息簇，以得到所述多个语义信息簇。
  8. 根据权利要求7所述的方法，其特征在于，所述根据至少一个所述问答实例更新所述至少一个基础语义信息簇，以得到所述多个语义信息簇，包括：
    确定至少一个所述问答实例中各个语义信息和所述至少一个基础语义信息簇中各个语义信息的相似度；
    针对至少一个所述问答实例中每一个问答实例，将所述问答实例划分至在所述至少一个基础语义信息簇中与所述问答实例的相似度大于第三预设相似度的基础语义信息簇中。
  9. 根据权利要求5-8任一项所述的方法，其特征在于，还包括：
    获取所述多个语义信息簇中各个查询问题的被选择次数；
    向所述机器人发送所述多个语义信息簇中各个查询问题的被选择次数。
  10. 一种机器人应答装置，其特征在于，包括：
    第一获取模块，用于获取当前查询语音；
    提取模块，用于提取所述当前查询语音的语义信息；
    匹配模块，用于对所述当前查询语音的语义信息和预先存储的多个语义信息簇进行匹配，得到匹配到的目标语义信息簇，每个所述语义信息簇包括：至少一个问答实例，每个所述问答实例包括：一个历史查询语音对应的语义信息和在所述历史查询语音对应的查询列表中被选择的查询问题；
    第二获取模块，用于获取所述目标语义信息簇中各个查询问题的被选择次数；
    确定模块，用于根据所述各个查询问题的被选择次数确定所述当前查询语音对应的目标查询问题；
    输出模块，用于输出所述目标查询问题对应的查询响应。
  11. 根据权利要求10所述的装置，其特征在于，所述匹配模块具体用于：
    确定所述当前查询语音的语义信息和所述多个语义信息簇中的各个历史查询语音对应的语义信息的相似度；
    若所述当前查询语音的语义信息和一个语义信息簇中的历史查询语音对应的语义信息的相似度大于第一预设相似度，则将该语义信息簇作为所述目标语义信息簇。
  12. 根据权利要求10或11所述的装置，其特征在于，所述确定模块具体用于：
    在所述各个查询问题中，将被选择次数大于预设次数的查询问题确定为所述目标查询问题。
  13. 根据权利要求10或11所述的装置，其特征在于，还包括：
    接收模块，用于接收服务器发送的所述多个语义信息簇。
  14. 一种机器人应答装置，其特征在于，包括：
    第一获取模块，用于获取多个语义信息簇，每个所述语义信息簇包括：至少一个问答实例，每个所述问答实例包括：一个历史查询语音对应的语义信息和在所述历史查询语音对应的查询列表中被选择的查询问题；
    第一发送模块，用于向机器人发送所述多个语义信息簇。
  15. 根据权利要求14所述的装置，其特征在于，还包括：
    第二获取模块，用于获取多个所述问答实例；
    相应的，第一获取模块具体用于：
    确定多个所述问答实例中每两个问答实例中的语义信息的相似度；
    将所述相似度大于第二预设相似度的问答实例归为一个语义信息簇。
  16. 根据权利要求14所述的装置，其特征在于，还包括：
    第三获取模块，用于获取至少一个基础语义信息簇以及至少一个所述问答实例；
    相应的，第一获取模块具体用于：
    根据至少一个所述问答实例更新所述至少一个基础语义信息簇，以得到所述多个语义信息簇。
  17. 根据权利要求16所述的装置，其特征在于，第一获取模块具体用于：
    确定至少一个所述问答实例中各个语义信息和所述至少一个基础语义信息簇中各个语义信息的相似度；
    针对至少一个所述问答实例中每一个问答实例，将所述问答实例划分至在所述至少一个基础语义信息簇中与所述问答实例的相似度大于第三预设相似度的基础语义信息簇中。
  18. 根据权利要求14-17任一项所述的装置，其特征在于，还包括：
    第四获取模块，用于获取所述多个语义信息簇中各个查询问题的被选择次数；
    第二发送模块，用于向所述机器人发送所述多个语义信息簇中各个查询问题的被选择次数。
  19. 一种机器人，其特征在于，包括：
    至少一个处理器；以及
    与所述至少一个处理器通信连接的存储器；其中，
    所述存储器存储有可被所述至少一个处理器执行的指令，所述指令被所述至少一个处理器执行，以使所述至少一个处理器能够执行权利要求1-4中任一项所述的方法。
  20. 一种服务器，其特征在于，包括：
    至少一个处理器；以及
    与所述至少一个处理器通信连接的存储器；其中，
    所述存储器存储有可被所述至少一个处理器执行的指令，所述指令被所述至少一个处理器执行，以使所述至少一个处理器能够执行权利要求5-9中任一项所述的方法。
  21. 一种计算机可读存储介质，其特征在于，所述计算机可读存储介质中存储有计算机执行指令，所述计算机执行指令被处理器执行时用于实现如权利要求1至4任一项所述的机器人应答方法。
  22. 一种计算机可读存储介质，其特征在于，所述计算机可读存储介质中存储有计算机执行指令，所述计算机执行指令被处理器执行时用于实现如权利要求5至9任一项所述的机器人应答方法。
PCT/CN2020/085530 2019-12-27 2020-04-20 一种机器人应答方法、装置、设备及存储介质 WO2021128663A1 (zh)

Priority Applications (2)

Application Number Priority Date Filing Date Title
EP20906575.4A EP4083812A4 (en) 2019-12-27 2020-04-20 ROBOT RESPONSE METHOD AND APPARATUS, DEVICE AND RECORDING MEDIUM
US17/785,369 US20230028830A1 (en) 2019-12-27 2020-04-20 Robot response method, apparatus, device and storage medium

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN201911376518.4A CN111159344A (zh) 2019-12-27 2019-12-27 机器人应答方法、装置、设备及存储介质
CN201911376518.4 2019-12-27

Publications (1)

Publication Number Publication Date
WO2021128663A1 true WO2021128663A1 (zh) 2021-07-01

Family

ID=70558447

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2020/085530 WO2021128663A1 (zh) 2019-12-27 2020-04-20 一种机器人应答方法、装置、设备及存储介质

Country Status (4)

Country Link
US (1) US20230028830A1 (zh)
EP (1) EP4083812A4 (zh)
CN (1) CN111159344A (zh)
WO (1) WO2021128663A1 (zh)

Families Citing this family (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112118311B (zh) * 2020-09-17 2023-10-27 阿波罗智联(北京)科技有限公司 信息车载交互方法、装置、设备以及存储介质
CN113360769B (zh) * 2021-06-28 2024-02-09 北京百度网讯科技有限公司 信息查询方法、装置、电子设备以及存储介质
CN113886435B (zh) * 2021-12-08 2022-03-18 国能大渡河大数据服务有限公司 一种基于循环神经网络的信息查询方法及系统

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108090174A (zh) * 2017-12-14 2018-05-29 北京邮电大学 一种基于系统功能语法的机器人应答方法及装置
CN109145099A (zh) * 2018-08-17 2019-01-04 百度在线网络技术（北京）有限公司 基于人工智能的问答方法和装置
CN110334200A (zh) * 2019-07-11 2019-10-15 深圳市仝智科技有限公司 一种问题答案库的生成方法、装置及存储介质
US20190377824A1 (en) * 2018-06-11 2019-12-12 Stratifyd, Inc. Schemaless systems and methods for automatically building and utilizing a chatbot knowledge base or the like

Family Cites Families (13)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US8645122B1 (en) * 2002-12-19 2014-02-04 At&T Intellectual Property Ii, L.P. Method of handling frequently asked questions in a natural language dialog service
CN101431573B (zh) * 2007-11-08 2013-02-20 上海撼世网络科技有限公司 通过人机交互技术实现自动客户服务的方法和设备
CN103810218B (zh) * 2012-11-14 2018-06-08 北京百度网讯科技有限公司 一种基于问题簇的自动问答方法和装置
US10691698B2 (en) * 2014-11-06 2020-06-23 International Business Machines Corporation Automatic near-real-time prediction, classification, and notification of events in natural language systems
US10692006B1 (en) * 2016-06-30 2020-06-23 Facebook, Inc. Crowdsourced chatbot answers
CN106709018A (zh) * 2016-12-27 2017-05-24 羽意互动咨询（北京）有限公司 命中预设问题的方法及装置
US11151992B2 (en) * 2017-04-06 2021-10-19 AIBrain Corporation Context aware interactive robot
US11200506B2 (en) * 2017-12-15 2021-12-14 Microsoft Technology Licensing, Llc Chatbot integrating derived user intent
WO2019150583A1 (ja) * 2018-02-05 2019-08-08 日本電気株式会社 質問群抽出方法、質問群抽出装置および記録媒体
US11645459B2 (en) * 2018-07-02 2023-05-09 Oracle International Corporation Social autonomous agent implementation using lattice queries and relevancy detection
US11380305B2 (en) * 2019-01-14 2022-07-05 Accenture Global Solutions Limited System and method for using a question and answer engine
CN109829048B (zh) * 2019-01-23 2023-06-23 平安科技（深圳）有限公司 电子装置、访谈辅助方法和计算机可读存储介质
CN110569356B (zh) * 2019-08-14 2022-05-20 中国平安人寿保险股份有限公司 基于智能面试交互系统的面试方法、装置和计算机设备

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108090174A (zh) * 2017-12-14 2018-05-29 北京邮电大学 一种基于系统功能语法的机器人应答方法及装置
US20190377824A1 (en) * 2018-06-11 2019-12-12 Stratifyd, Inc. Schemaless systems and methods for automatically building and utilizing a chatbot knowledge base or the like
CN109145099A (zh) * 2018-08-17 2019-01-04 百度在线网络技术（北京）有限公司 基于人工智能的问答方法和装置
CN110334200A (zh) * 2019-07-11 2019-10-15 深圳市仝智科技有限公司 一种问题答案库的生成方法、装置及存储介质

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
See also references of EP4083812A4 *

Also Published As

Publication number Publication date
EP4083812A4 (en) 2023-12-20
EP4083812A1 (en) 2022-11-02
US20230028830A1 (en) 2023-01-26
CN111159344A (zh) 2020-05-15

Similar Documents

Publication Publication Date Title
JP7108122B2 (ja) コンピュータによるエージェントのための合成音声の選択
WO2021128663A1 (zh) 一种机器人应答方法、装置、设备及存储介质
US10949660B2 (en) Electronic handwriting processor with convolutional neural networks
JP2020034897A (ja) 自然言語会話に関連する情報の視覚的提示
US20160019280A1 (en) Identifying question answerers in a question asking system
CN111353299B (zh) 基于人工智能的对话场景确定方法和相关装置
CN106570181B (zh) 基于上下文管理的智能交互方法及系统
CN115062627B (zh) 用于基于人工智能的计算机辅助说服系统的方法和装置
CN112307166B (zh) 一种智能问答方法、装置、存储介质以及计算机设备
CN112115244A (zh) 对话交互方法、装置、存储介质及电子设备
CN111967599A (zh) 用于训练模型的方法、装置、电子设备及可读存储介质
CN109275036A (zh) 一种用于教学直播的消息提醒方法、装置及设备
CN111444321B (zh) 问答方法、装置、电子设备和存储介质
US12079629B2 (en) Score prediction using hierarchical attention
CN109190116A (zh) 语义解析方法、系统、电子设备及存储介质
CN114373098B (zh) 一种图像分类方法、装置、计算机设备及存储介质
CN109002498A (zh) 人机对话方法、装置、设备及存储介质
CN112426726A (zh) 游戏事件抽取方法、装置、存储介质及服务器
US11934974B2 (en) Intelligent integrated remote reporting system
JP7466143B2 (ja) 情報処理装置、学習提案プログラム、及び学習提案方法
CN112036208B (zh) 一种基于交互式学习系统的人工智能小精灵
Goh et al. Embodied conversational agents for H5N1 pandemic crisis
Onsorodi Application of Genetic Algorithm to Keyboard Arrangement Problem
CN115617992A (zh) 标签生成方法、装置、计算机可读存储介质及计算机设备
CN116821295A (zh) 智能问答方法、装置、计算机设备及介质

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: 20906575

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

ENP Entry into the national phase

Ref document number: 2020906575

Country of ref document: EP

Effective date: 20220727