Customer intention identification method and device

Abstract

The present invention provides a method and device for identifying customer intentions, which are applicable to the field of artificial intelligence technology. The method includes: obtaining customer's question information; clustering the question information to construct a standard question set of the current question information; wherein the standard question set includes at least one standard question; and using a preset intention model to perform intention identification on the standard question set and the question information to generate an intention identification result. The present invention constructs a standard question set based on the customer's question information, and uses a preset intention model to perform intention identification on the standard question set and the question information to generate an intention identification result, thereby solving the problem of inaccurate intention identification caused by factors such as short length of customer questions, high noise, and diverse expressions in human-computer interaction.

Description

Technical Field

[0001] The present invention relates to intelligent interaction technology, and in particular to a method and device for identifying customer intention. Background Art

[0002] In recent years, with the rapid development and widespread application of natural language understanding and deep learning technologies, intelligent interaction has become an important research task in the field of natural language processing, and has attracted widespread attention from academia and the business community.

[0003] Classic human-computer interaction systems generally include three main parts: customer intent recognition, customer dialogue management, and natural language generation. Among them, customer intent recognition is a primary and basic link for whether the human-computer interaction system can conduct accurate and effective dialogue. However, due to the complexity of the dialogue environment in human-computer dialogue and the diversity of customer expressions, the dialogue system in the existing technology often makes mistakes in recognizing customer intent, which greatly affects the interaction experience. Therefore, there is an urgent need to propose a more effective solution for the task of customer intent recognition in intelligent interaction. Summary of the invention

[0004] In order to overcome a defect of the prior art and improve the accuracy of intention recognition, the present invention provides a method for identifying customer intention, the method comprising:

[0005] Get customer's question information;

[0006] Clustering the question information to construct a standard question set of the current question information; wherein the standard question set includes at least one standard question;

[0007] The preset intention model is used to perform intention recognition on the standard question set and question information to generate an intention recognition result.

[0008] In the embodiment of the present invention, clustering the question information to construct a standard question set of the current question information includes:

[0009] Generate cluster samples according to the keywords of the question information;

[0010] Determining an average silhouette coefficient of the clustered samples;

[0011] Determining the optimal number of clusters according to the determined average silhouette coefficient to determine the clusters of the question information;

[0012] The standard question sentences of the cluster group are determined according to the similarity of the cluster examples in the cluster group, so as to determine the standard question sentence set of the current question sentence information.

[0013] In the embodiment of the present invention, the step of determining the standard question sentence of the cluster group according to the similarity of the cluster examples in the cluster group to determine the standard question sentence set of the current question sentence information includes:

[0014] Determining the similarity of each cluster example within the cluster family;

[0015] The representativeness value of each clustering example to its cluster is determined according to the similarity of the clustering examples, so that the clustering example with the highest representativeness value in each clustering family is used as the standard question set of the current question information.

[0016] In the embodiment of the present invention, the method of using a preset intention model to perform intention recognition on the standard question set and question information to generate an intention recognition result includes:

[0017] Vectorizing the standard questions and customer question information in the standard question set to construct a mapping matrix between the standard questions and customer question information;

[0018] Using a bidirectional LSTM neural network to encode the mapping matrix to determine the standard question features and the semantic features of the customer question information;

[0019] The mutual attention mechanism is used to determine the matching information features between the semantic features of the standard question and the semantic features of the customer's question information;

[0020] Determine the local features in the matching information features using a CNN neural network to generate a prediction vector;

[0021] Generate intent recognition results based on the generated prediction vector.

[0022] In the embodiment of the present invention, generating the intention recognition result according to the generated prediction vector includes:

[0023] Determine the similarity between the prediction vector in the standard question set and the customer question information;

[0024] The intent recognition result is determined by using the voting method based on the determined similarity.

[0025] At the same time, the present invention also provides a customer intention recognition device, comprising:

[0026] Information acquisition module, used to obtain customer's question information;

[0027] A question set building module, used for clustering the question information to build a standard question set of the current question information; wherein the standard question set includes at least one standard question;

[0028] The intention recognition module is used to use a preset intention model to perform intention recognition on the standard question set and question information to generate an intention recognition result.

[0029] In the embodiment of the present invention, the question set construction module includes:

[0030] A clustering unit, used to generate clustering samples according to the keywords of the question information;

[0031] A silhouette coefficient determination unit, used to determine an average silhouette coefficient of the clustering samples;

[0032] A cluster group determination unit, used to determine the optimal number of cluster groups according to the determined average silhouette coefficient to determine the cluster group of the question information;

[0033] The question set determination unit is used to determine the standard question of the cluster group according to the similarity of the cluster examples in the cluster group, so as to determine the standard question set of the current question information.

[0034] In an embodiment of the present invention, the question set determination unit includes:

[0035] A similarity determination unit, used to determine the similarity of each clustering example in the clustering family;

[0036] The representative value determination unit is used to determine the representative degree value of each cluster example to its cluster according to the similarity of the cluster examples, so as to use the cluster example with the highest representative degree value in each cluster group as the standard question set of the current question information.

[0037] In an embodiment of the present invention, the intention recognition module includes:

[0038] A vectorization unit, used for performing vectorization processing on the standard questions and customer question information in the standard question set to construct a mapping matrix between the standard questions and customer question information;

[0039] A semantic feature determination unit, used to encode the mapping matrix using a bidirectional LSTM neural network to determine the semantic features of the standard question features and the customer question information;

[0040] A matching information feature determination unit, used to determine matching information features between semantic features of the standard question and semantic features of the customer's question information by using a mutual attention mechanism;

[0041] A prediction vector determination unit, used for determining the local features in the matching information features by using a CNN neural network to generate a prediction vector;

[0042] The result generating unit is used to generate the intention recognition result according to the generated prediction vector.

[0043] In the embodiment of the present invention, the result generating unit includes:

[0044] A similarity determination unit, used to determine the similarity between the prediction vector in the standard question set and the customer question information;

[0045] The voting unit is used to determine the intent recognition result by using a voting method according to the determined similarity.

[0046] At the same time, the present invention also provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, and the above method is implemented when the processor executes the computer program.

[0047] At the same time, the present invention also provides a computer-readable storage medium, which stores a computer program for executing the above method.

[0048] The customer intention recognition method provided by the present invention constructs a standard question set based on the customer's question information, and uses a preset intention model to perform intention recognition on the standard question set and the question information to generate an intention recognition result, thereby solving the problem of inaccurate intention recognition caused by factors such as short customer question length, high noise, and diverse expressions in human-computer interaction. The method overcomes the defect in the prior art that the dialogue system often makes mistakes in customer intention recognition due to the complexity of the dialogue environment in human-computer dialogue and the diversity of customer expressions.

[0049] In order to make the above and other purposes, features and advantages of the present invention more obvious and easy to understand, preferred embodiments are specifically cited below and described in detail with reference to the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0051] Figure 1 A flow chart of the customer intention identification method provided by the present invention;

[0052] Figure 2 It is a schematic diagram of an embodiment of the present invention;

[0053] Figure 3 It is a schematic diagram of an embodiment of the present invention;

[0054] Figure 4 It is a schematic diagram of an embodiment of the present invention;

[0055] Figure 5 A block diagram of a client intention recognition device provided by the present invention;

[0056] Figure 6 A schematic diagram of an electronic device embodiment provided by the present invention. DETAILED DESCRIPTION

[0057] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0058] The typical method of existing intent recognition is to convert the intent recognition task into an automatic classification task of intent. That is, given a customer intent classification system, the automatic recognition of intent is achieved by extracting features and training classification models. For example, if a customer expresses "I want to buy some stable funds", the system will classify it into the product consultation category; if the customer expresses "When will the financial product I bought expire", the system will classify it into the "after-sales consultation" category, etc. Foreign researchers have proposed using sentence similarity to identify customer intent; domestic researchers have introduced the idea of ​​feature extension into intent recognition research, that is, introducing additional information to mine the meaning expressed by the text to assist classification. These studies of the existing technology have achieved effective results on specific data sets. However, since most of these intent recognition methods are based on template matching or artificial feature sets, they are not only time-consuming and labor-intensive, but also have poor scalability. In addition, they often give irrelevant answers due to intent recognition errors, which greatly affects the interactive experience.

[0059] Due to the complexity of the dialogue environment in human-computer dialogue and the diversity of customer expressions, the dialogue system in the prior art often makes mistakes in identifying customer intentions. In order to overcome the defects of customer intention recognition in the prior art and improve the accuracy of intention recognition, the present invention provides a customer intention recognition method, such as Figure 1 As shown, the method of the present invention comprises:

[0060] Step S101, obtaining the customer's question information;

[0061] Step S102, clustering the question information to construct a standard question set of the current question information; wherein the standard question set includes at least one standard question;

[0062] Step S103: Use a preset intention model to perform intention recognition on the standard question set and question information to generate an intention recognition result.

[0063] The customer intention recognition method provided by the present invention utilizes the customer's question information to construct a standard question set, performs intention recognition on the standard question set including the standard questions and the customer's file information together, increases the diversity of questions during intention recognition, solves the problem of inaccurate intention recognition caused by factors such as short length of customer questions, high noise, and diverse expressions in human-computer interaction, and proposes a customer intention recognition method based on standard question selection and customer question fusion.

[0064] In the embodiment of the present invention, the clustering process of the question information in step S102 to construct a standard question set of the current question information includes:

[0065] Generate cluster samples according to the keywords of the question information;

[0066] Determining an average silhouette coefficient of the clustered samples;

[0067] Determining the optimal number of clusters according to the determined average silhouette coefficient to determine the clusters of the question information;

[0068] The standard question sentences of the cluster group are determined according to the similarity of the cluster examples in the cluster group, so as to determine the standard question sentence set of the current question sentence information.

[0069] In the embodiment of the present invention, customer questions in intelligent interaction are clustered, and the optimal number of clusters is selected by comparing the average silhouette coefficients under different numbers of clusters; at the same time, standard questions in each cluster are selected by similarity calculation to construct a standard question set, which helps to increase the diversity of standard questions.

[0070] Specifically, in the embodiment of the present invention, determining the standard question set by using the above steps includes:

[0071] Receive the audio file of the customer in the intelligent interaction and obtain the customer's question information, that is, use the ASR technology to convert the audio file in the intelligent interaction into text and perform word segmentation operations, and convert the corresponding text into a word sequence;

[0072] Use clustering algorithm to extract text keyword features, cluster customer questions in intelligent interaction, and select the optimal number of clusters based on the calculated average silhouette coefficient;

[0073] Through similarity calculation, the sample with the highest similarity to other samples (non-standard questions) in the cluster is selected as a reference, and the standard questions selected in each cluster are constructed into a standard question set.

[0074] like Figure 2 As shown in FIG. 1 , a schematic diagram of the construction of a standard file set in an embodiment of the present invention is shown. The customer files are clustered and the optimal number of clusters is determined by calculating the silhouette coefficient. Figure 2As shown, it determines that the optimal number of clusters is 3, and through similarity calculation, selects the samples R1-R3 with the highest similarity in the C1-C3 cluster as the standard questions of the cluster, and the samples R1-R3 construct the standard question set.

[0075] By comparing the average silhouette coefficients under different numbers of clusters, the optimal number of clusters is selected. After determining the optimal number of clusters, the standard questions in each cluster are selected through similarity calculation to construct a standard question set, which helps to increase the diversity of standard questions.

[0076] In the embodiment of the present invention, in step S103, using a preset intention model to perform intention recognition on the standard question set and question information to generate an intention recognition result includes:

[0077] Vectorizing the standard questions and customer question information in the standard question set to construct a mapping matrix between the standard questions and customer question information;

[0078] Using a bidirectional LSTM neural network to encode the mapping matrix to determine the standard question features and the semantic features of the customer question information;

[0079] The mutual attention mechanism is used to determine the matching information features between the semantic features of the standard question and the semantic features of the customer's question information;

[0080] Determine the local features in the matching information features using a CNN neural network to generate a prediction vector;

[0081] Generate intent recognition results based on the generated prediction vector.

[0082] like Figure 3 As shown, it is a schematic diagram of performing intent recognition processing on standard questions and customer questions in an embodiment of the present invention, that is, in an embodiment of the present invention, vectorization operation is performed on the words in the standard questions and customer questions, and mapping matrices of the standard questions and customer questions are respectively constructed. A bidirectional LSTM neural network is used to encode the mapping matrices to obtain semantic representation features of the text, and a mutual attention mechanism is used to calculate matching information of the two semantic features to obtain matching information features. The mutual attention mechanism is an algorithm, which refers to bidirectional attention calculation from customer questions to standard questions and from standard questions to customer questions, which can better match the information of the two. Then, a CNN neural network is used to further capture local feature information in the matching information to obtain a final prediction vector to improve the accuracy of the prediction.

[0083] In the embodiment of the present invention, generating an intention recognition result according to the generated prediction vector includes:

[0084] Determine the similarity between the prediction vector in the standard question set and the customer question information;

[0085] The intent recognition result is determined by using the voting method based on the determined similarity.

[0086] The intent recognition method of the present invention is further described in detail below in conjunction with the specific steps of the embodiment. The technical problem to be solved in this embodiment is: to solve the problem of inaccurate intent recognition caused by factors such as short customer question length, high noise, and diverse expressions in human-computer interaction. Therefore, a customer intent recognition method based on the fusion of standard question selection and attention mechanism is proposed.

[0087] like Figure 4 As shown, it is a flowchart of customer intent recognition provided by this embodiment. This embodiment provides a customer intent recognition method based on standard question selection and attention mechanism fusion, including the following steps:

[0088] Step 1: Use ASR technology to convert the audio file in the intelligent interaction into text and perform word segmentation, and convert the corresponding text into a word sequence;

[0089] Step 2: Use clustering algorithm to extract text keyword features, cluster customer questions in intelligent interaction, and select the optimal number of clusters based on the calculated average silhouette coefficient;

[0090] The average silhouette coefficient is calculated by the following formula:

[0091]

[0092] Among them, S i Represents the silhouette coefficient of the i-th sample in the cluster;

[0093] a i Represents the average distance from the i-th sample to other samples in the cluster;

[0094] b i It represents the minimum distance from the i-th sample to other samples in the cluster.

[0095] In this embodiment, the value range of the silhouette coefficient is [-1, 1]. The closer the distance between samples within a cluster is and the farther the distance between samples between clusters is, the larger the value is, and the better the clustering effect is.

[0096] Step 3: Select the sample with the highest similarity to other samples (non-standard questions) in the cluster as a reference through similarity calculation, and construct the standard questions selected in each cluster into a standard question set; Figure 4 As shown, the standard question set constructed in this embodiment includes three standard questions.

[0097] The similarity between customer questions in a cluster is calculated as follows:

[0098]

[0099] Among them, x1 and x2 represent two customer questions in the cluster respectively;

[0100] α1 and α2 are weight parameters, the sum of which is 1. In this embodiment, the specific value is 0.6;

[0101] Loverlap represents the number of overlapping words between customer questions x1 and x2;

[0102] L1 and L2 represent the number of words in x1 and x2 respectively.

[0103] After calculating the similarity of any two sentences in the cluster, we get the similarity matrix M∈R n×m , where m ij Represents the i-th question x i With the jth question x j The similarity.

[0104] Use v i Indicates question x i The degree of cluster representation, v i ∈[0,1],v i The bigger the question x i The more representative the i The largest question is used as the standard question of this cluster.

[0105]

[0106] Step 4: Vectorize the words in the standard questions and customer questions, and construct mapping matrices for each standard question and customer question respectively; Figure 4 As shown, mapping matrices between standard question 1 and customer question, standard question 2 and customer question, and standard document 3 and customer question are constructed respectively.

[0107] In this embodiment, each word in the customer's question is searched in a preset dictionary to obtain a vectorized representation of the word, and then a customer question mapping matrix is ​​constructed according to the order in which each word appears in the customer's question; if individual words in the customer's question do not appear in the dictionary, they can be regarded as stop words and discarded.

[0108] Step 5: Use a bidirectional LSTM neural network to encode the mapping matrix to obtain the semantic representation features of the text;

[0109] In this embodiment, the standard question mapping matrix R and the customer question mapping matrix C are input into a bidirectional LSTM neural network, which can effectively retain the word order information in the text.

[0110] Step 6: Use the mutual attention mechanism to calculate the matching information of the two semantic features and obtain the matching information features. The mutual attention mechanism is an algorithm that refers to the bidirectional attention calculation from the customer question to the standard question and from the standard question to the customer question, which can better match the information of the two.

[0111] Among them, the mutual attention mechanism is divided into: the attention mechanism from customer questions to standard questions and the attention mechanism from standard questions to customer questions, which can effectively match the information of the two.

[0112] Step 7: Use the CNN neural network to further capture the local feature information in the matching information and obtain the final prediction vector to improve the accuracy of the prediction.

[0113] In the embodiment of the present invention, after being processed by the CNN neural network, local features can be extracted to highlight the common information between the standard question and the customer question.

[0114] Step 8: Calculate the similarity between the customer's question and each standard question in the standard question set. All standard questions in the standard question set are taken from the final prediction vector in step 7 to form a prediction branch set. The voting method is used to calculate the final intent category.

[0115] In the embodiment of the present invention, customer questions in intelligent interaction are first clustered, and the optimal number of clusters is selected by comparing the average silhouette coefficients under different numbers of clusters; at the same time, standard questions in each cluster are selected by similarity calculation to construct a standard question set, which helps to increase the diversity of standard questions.

[0116] In addition, the present invention uses an intention classification model based on the attention mechanism, which can accurately describe the semantic similarity between customer questions and standard questions. At the same time, a voting method is used to calculate multiple intention prediction categories, and the best result is selected as the final result, thereby effectively improving the accuracy of intention recognition and reducing the phenomenon of irrelevant answers.

[0117] At the same time, the present invention also provides a customer intention recognition device, such as Figure 5 As shown, it includes:

[0118] The information acquisition module 501 is used to acquire the customer's question information;

[0119] A question set building module 502 is used to cluster the question information to build a standard question set of the current question information; wherein the standard question set includes at least one standard question;

[0120] The intention recognition module 503 is used to use a preset intention model to perform intention recognition on the standard question set and question information to generate an intention recognition result.

[0121] In the embodiment of the present invention, the question set construction module 502 includes:

[0122] A clustering unit, used to generate clustering samples according to the keywords of the question information;

[0123] A silhouette coefficient determination unit, used to determine an average silhouette coefficient of the clustering samples;

[0124] A cluster group determination unit, used to determine the optimal number of cluster groups according to the determined average silhouette coefficient to determine the cluster group of the question information;

[0125] The question set determination unit is used to determine the standard question of the cluster group according to the similarity of the cluster examples in the cluster group, so as to determine the standard question set of the current question information.

[0126] In an embodiment of the present invention, the question set determination unit includes:

[0127] A similarity determination unit, used to determine the similarity of each clustering example in the clustering family;

[0128] The representative value determination unit is used to determine the representative degree value of each cluster example to its cluster according to the similarity of the cluster examples, so as to use the cluster example with the highest representative degree value in each cluster group as the standard question set of the current question information.

[0129] In the embodiment of the present invention, the intention recognition module 503 includes:

[0130] A vectorization unit, used for performing vectorization processing on the standard questions and customer question information in the standard question set to construct a mapping matrix between the standard questions and customer question information;

[0131] A semantic feature determination unit, used to encode the mapping matrix using a bidirectional LSTM neural network to determine the semantic features of the standard question features and the customer question information;

[0132] A matching information feature determination unit, used to determine matching information features between semantic features of the standard question and semantic features of the customer's question information by using a mutual attention mechanism;

[0133] A prediction vector determination unit, used for determining the local features in the matching information features by using a CNN neural network to generate a prediction vector;

[0134] The result generating unit is used to generate the intention recognition result according to the generated prediction vector.

[0135] In the embodiment of the present invention, the result generating unit includes:

[0136] A similarity determination unit, used to determine the similarity between the prediction vector in the standard question set and the customer question information;

[0137] The voting unit is used to determine the intent recognition result by using a voting method according to the determined similarity.

[0138] For those skilled in the art, the implementation method of the customer intention recognition device provided by the present invention can be clearly known through the description of the above embodiments, which will not be repeated here.

[0139] It should be noted that the customer intention recognition method and device provided by the present invention are applicable to the field of artificial intelligence technology. The customer intention recognition method and device provided by the present invention can be used to recognize customer intention in human-computer interaction in the financial field, and can also be used to recognize customer intention in any field other than the financial field. The application field of the customer intention recognition method and device provided by the present invention is not limited.

[0140] This embodiment also provides an electronic device, which may be a desktop computer, a tablet computer, a mobile terminal, etc., but this embodiment is not limited thereto. In this embodiment, the electronic device may refer to the embodiments of the aforementioned method and device, the contents of which are incorporated herein, and the repeated parts are not repeated.

[0141] Figure 6 FIG. 6 is a schematic block diagram of a system structure of an electronic device 600 according to an embodiment of the present invention. Figure 6 As shown, the electronic device 600 may include a central processor 100 and a memory 140; the memory 140 is coupled to the central processor 100. It should be noted that this figure is exemplary; other types of structures may also be used to supplement or replace this structure to implement telecommunication functions or other functions.

[0142] In one embodiment, the customer intention recognition function may be integrated into the central processor 100. The central processor 100 may be configured to perform the following control:

[0143] Get customer's question information;

[0144] Clustering the question information to construct a standard question set of the current question information; wherein the standard question set includes at least one standard question;

[0145] The preset intention model is used to perform intention recognition on the standard question set and question information to generate an intention recognition result.

[0146] In another embodiment, the customer intention recognition device may be configured separately from the central processor 100. For example, the customer intention recognition device may be configured as a chip connected to the central processor 100, and the customer intention recognition function may be implemented under the control of the central processor.

[0147] like Figure 6As shown, the electronic device 600 may further include: a communication module 110, an input unit 120, an audio processing unit 130, a display 160, and a power supply 170. It is worth noting that the electronic device 600 does not necessarily have to include Figure 6 In addition, the electronic device 600 may also include Figure 6 For components not shown, reference may be made to the prior art.

[0148] like Figure 6 As shown, the central processor 100 is sometimes also referred to as a controller or an operation control, and may include a microprocessor or other processor devices and / or logic devices. The central processor 100 receives inputs and controls the operations of various components of the electronic device 600.

[0149] The memory 140 may be, for example, one or more of a cache, a flash memory, a hard drive, a removable medium, a volatile memory, a non-volatile memory or other suitable devices. The above-mentioned information related to the failure may be stored, and a program for executing the relevant information may also be stored. The CPU 100 may execute the program stored in the memory 140 to implement information storage or processing.

[0150] The input unit 120 provides input to the CPU 100. The input unit 120 is, for example, a key or a touch input device. The power supply 170 is used to provide power to the electronic device 600. The display 160 is used to display display objects such as images and text. The display may be, for example, an LCD display, but is not limited thereto.

[0151] The memory 140 may be a solid-state memory, such as a read-only memory (ROM), a random access memory (RAM), a SIM card, etc. It may also be a memory that saves information even when the power is off, can be selectively erased, and is provided with more data, examples of which are sometimes referred to as EPROMs, etc. The memory 140 may also be some other type of device. The memory 140 includes a buffer memory 141 (sometimes referred to as a buffer). The memory 140 may include an application / function storage unit 142, which is used to store application programs and function programs or processes for executing the operation of the electronic device 600 through the central processor 100.

[0152] The memory 140 may also include a data storage unit 143 for storing data, such as contacts, digital data, pictures, sounds, and / or any other data used by the electronic device. The driver storage unit 144 of the memory 140 may include various drivers for communication functions of the electronic device and / or for executing other functions of the electronic device (such as messaging applications, address book applications, etc.).

[0153] The communication module 110 is a transmitter / receiver 110 that transmits and receives signals via an antenna 111. The communication module (transmitter / receiver) 110 is coupled to the central processor 100 to provide input signals and receive output signals, which may be the same as the case of a conventional mobile communication terminal.

[0154] Based on different communication technologies, multiple communication modules 110 may be provided in the same electronic device, such as a cellular network module, a Bluetooth module and / or a wireless LAN module. The communication module (transmitter / receiver) 110 is also coupled to a speaker 131 and a microphone 132 via an audio processor 130 to provide an audio output via the speaker 131 and receive an audio input from the microphone 132, thereby realizing a common telecommunication function. The audio processor 130 may include any suitable buffer, decoder, amplifier, etc. In addition, the audio processor 130 is also coupled to the central processor 100, so that the sound can be recorded on the local machine through the microphone 132, and the sound stored on the local machine can be played through the speaker 131.

[0155] An embodiment of the present invention further provides a computer-readable program, wherein when the program is executed in an electronic device, the program enables a computer to execute the customer intention identification method as described in the above embodiment in the electronic device.

[0156] An embodiment of the present invention further provides a storage medium storing a computer-readable program, wherein the computer-readable program enables a computer to execute the customer intention recognition described in the above embodiment in an electronic device.

[0157] The preferred embodiments of the present invention have been described above with reference to the accompanying drawings. Many features and advantages of these embodiments are clear from this detailed description, and therefore the appended claims are intended to cover all such features and advantages of these embodiments that fall within their true spirit and scope. In addition, since many modifications and changes are readily apparent to those skilled in the art, the embodiments of the present invention are not intended to be limited to the precise structure and operation illustrated and described, but all suitable modifications and equivalents that fall within their scope may be included.

[0158] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0159] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0160] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0161] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0162] The present invention uses specific embodiments to illustrate the principles and implementation methods of the present invention. The description of the above embodiments is only used to help understand the method of the present invention and its core idea. At the same time, for those skilled in the art, according to the idea of ​​the present invention, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as a limitation on the present invention.

Claims

1. A method for identifying customer intention, characterized in that: The method includes: Get customer's question information; Clustering the question information to construct a standard question set of the current question information; wherein the standard question set includes at least one standard question; Using a preset intention model to perform intention recognition on the standard question set and question information to generate an intention recognition result; The step of clustering the question information to construct a standard question set for the current question information includes: Generate cluster samples according to the keywords of the question information; determining an average silhouette coefficient of the clustered samples; Determining the optimal number of clusters according to the determined average silhouette coefficient to determine the clusters of the question information; The standard question sentences of the cluster group are determined according to the similarity of the cluster examples in the cluster group, so as to determine the standard question sentence set of the current question sentence information.

2. The method for identifying customer intentions according to claim 1, wherein: The step of determining the standard question sentence of the cluster group according to the similarity of the cluster samples in the cluster group to determine the standard question sentence set of the current question sentence information includes: Determining the similarity of each cluster example within the cluster family; The representativeness value of each clustering example to its cluster is determined according to the similarity of the clustering examples, so that the clustering example with the highest representativeness value in each clustering family is used as the standard question set of the current question information.

3. The method for identifying customer intentions according to claim 1, wherein: The method of using a preset intention model to perform intention recognition on the standard question set and question information to generate an intention recognition result includes: Vectorizing the standard questions and customer question information in the standard question set to construct a mapping matrix between the standard questions and customer question information; Using a bidirectional LSTM neural network to encode the mapping matrix to determine the standard question features and the semantic features of the customer question information; The mutual attention mechanism is used to determine the matching information features between the semantic features of the standard question and the semantic features of the customer's question information; Determine the local features in the matching information features using a CNN neural network to generate a prediction vector; Generate intent recognition results based on the generated prediction vector.

4. The method for identifying customer intentions according to claim 3, wherein: The generating of the intention recognition result according to the generated prediction vector includes: Determine the similarity between the prediction vector in the standard question set and the customer question information; The intent recognition result is determined by using the voting method based on the determined similarity.

5. A customer intention recognition device, characterized in that: The device comprises: Information acquisition module, used to obtain customer's question information; A question set building module, used for clustering the question information to build a standard question set of the current question information; wherein the standard question set includes at least one standard question; An intention recognition module, used to perform intention recognition on the standard question set and question information using a preset intention model to generate an intention recognition result; Wherein, the question set construction module includes: A clustering unit, used to generate clustering samples according to the keywords of the question information; A silhouette coefficient determination unit, used to determine an average silhouette coefficient of the cluster samples; A cluster group determination unit, used to determine the optimal number of cluster groups according to the determined average silhouette coefficient to determine the cluster group of the question information; The question set determination unit is used to determine the standard question of the cluster group according to the similarity of the cluster examples in the cluster group, so as to determine the standard question set of the current question information.

6. The customer intention recognition device according to claim 5, characterized in that: The question set determination unit comprises: A similarity determination unit, used to determine the similarity of each cluster example in the cluster group; The representative value determination unit is used to determine the representative degree value of each cluster example to its cluster according to the similarity of the cluster examples, so as to use the cluster example with the highest representative degree value in each cluster group as the standard question set of the current question information.

7. The customer intention recognition device according to claim 5, characterized in that: The intention recognition module includes: A vectorization unit, used for performing vectorization processing on the standard questions and customer question information in the standard question set to construct a mapping matrix between the standard questions and customer question information; A semantic feature determination unit, used to encode the mapping matrix using a bidirectional LSTM neural network to determine the semantic features of the standard question features and the customer question information; A matching information feature determination unit, used to determine matching information features between semantic features of the standard question and semantic features of the customer's question information by using a mutual attention mechanism; A prediction vector determination unit, used for determining the local features in the matching information features by using a CNN neural network to generate a prediction vector; The result generating unit is used to generate the intention recognition result according to the generated prediction vector.

8. The customer intention recognition device according to claim 7, characterized in that: The result generating unit comprises: A similarity determination unit, used to determine the similarity between the prediction vector in the standard question set and the customer question information; The voting unit is used to determine the intent recognition result by using a voting method according to the determined similarity.

9. A computer device comprising a memory, a processor and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the method according to any one of claims 1 to 4 is implemented.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program for executing the method according to any one of claims 1 to 4.

Images