CN-117669595-B - Customer service result prediction model establishment method, device and equipment

CN117669595BCN 117669595 BCN117669595 BCN 117669595BCN-117669595-B

Abstract

The invention provides a method, a device and equipment for establishing a customer service result prediction model, and relates to the technical field of artificial intelligence; the method comprises the steps of carrying out quantum measurement processing on target sentence text training data in a training text data set to obtain semantic features of target sentences, obtaining semantic features of target conversations according to the semantic features of the target sentences by utilizing a bidirectional long-short-term memory BiLSTM model and a self-attention mechanism, and obtaining a customer service result prediction model according to the semantic features of the target conversations and historical customer service results corresponding to the target conversations. According to the scheme of the invention, the customer service result prediction model obtained according to the semantic features between the contexts in the target dialogue fully models the semantic features of the target sentences and the semantic features between the contexts in the target dialogue, so that the accuracy of predicting the customer service result by the customer service result prediction model is improved.

Inventors

ZHAO DONGMING
HUANG KUN

Assignees

中国移动通信集团天津有限公司
中国移动通信集团有限公司

Dates

Publication Date: 20260505
Application Date: 20220809

Claims (8)

1. The method for establishing the customer service result prediction model is characterized by comprising the following steps of: Obtaining a training text data set according to the historical customer service voice data; Carrying out quantum measurement processing on the target sentence text training data in the training text data set to obtain semantic features of the target sentence; obtaining semantic features of a target dialogue according to the semantic features of target sentences by using a bidirectional long-short-time memory BiLSTM model and a self-attention mechanism, wherein the target dialogue is any dialogue in the training text data set; Obtaining a customer service result prediction model according to semantic features of a target dialogue and a history customer service result corresponding to the target dialogue; the quantum measurement processing is performed on the target sentence text training data in the training text data set to obtain the semantic features of the target sentence, including: word vector embedding processing is carried out on the target sentence text training data, so that word vectors of words in the target sentence are obtained; processing word vectors of words in the target sentence by utilizing BiLSTM models to obtain hidden state vectors of words in the target sentence; Coding the hidden state vector of the words in the target sentence to obtain a quasi-density matrix of the target sentence; Performing projection measurement on the quasi-density matrix of the target sentence to obtain the semantic feature of the target sentence; the encoding the hidden state vector of the word in the target sentence to obtain the quasi-density matrix of the target sentence includes: normalizing the hidden state vector of the words in the target sentence to obtain a quantum state vector of the words in the target sentence; Carrying out outer product calculation on the quantum state vector of the word in the target sentence to obtain a density matrix of the word in the target sentence; and obtaining a quasi-density matrix of the target sentence according to the density matrix of the words in the target sentence and the weight coefficient corresponding to each word in the target sentence.
2. The customer service outcome prediction model building method according to claim 1, wherein the method further comprises: And obtaining a customer service result corresponding to the voice data to be predicted according to the voice data to be predicted by using the customer service result prediction model.
3. The method for building a prediction model of a customer service result according to claim 1, wherein the processing, with the BiLSTM model, the word vector of the word in the target sentence to obtain the hidden state vector of the word in the target sentence includes: inputting the target sentence into the BiLSTM model; Acquiring a hidden state vector of a word in the target sentence output by the BiLSTM model; the number of hidden layers in the BiLSTM model is the same as the number of words in the target sentence.
4. The method for building a customer service result prediction model according to claim 1, wherein the performing projection measurement on the pseudo-density matrix of the target sentence to obtain the semantic feature of the target sentence comprises: And performing projection calculation for preset times on the quasi-density matrix of the target sentence on a projection operator to obtain the semantic features of the target sentence.
5. The method for building a customer service result prediction model according to claim 1, wherein the obtaining semantic features of the target dialogue according to the semantic features of the target sentence by using the bidirectional long-short-time memory BiLSTM model and the self-attention mechanism comprises: processing semantic features of a plurality of target sentences in the target dialogue by using the BiLSTM model to obtain semantic features among the target sentences in the target dialogue, wherein the number of hidden layers in the BiLSTM model is equal to the number of the target sentences in the target dialogue; And fusing semantic features among the target sentences through a self-attention mechanism to obtain abstract semantic features of the target dialogue.
6. A customer service result prediction model building apparatus, comprising: The first processing module is used for obtaining a training text data set according to the historical customer service voice data; the second processing module is used for carrying out quantum measurement processing on the target sentence text training data in the training text data set to obtain semantic features of the target sentence; The third processing module is used for obtaining semantic features of a target dialogue according to the semantic features of the target sentence by utilizing a bidirectional long-short-term memory BiLSTM model and a self-attention mechanism, wherein the target dialogue is any dialogue in the training text data set; The model generation module is used for obtaining a client service result prediction model according to semantic features of a target dialogue and a history client service result corresponding to the target dialogue; Wherein the second processing module comprises: the first processing unit is used for carrying out word vector embedding processing on the target sentence text training data to obtain word vectors of words in the target sentence; the second processing unit is used for processing word vectors of words in the target sentence by utilizing a BiLSTM model to obtain hidden state vectors of the words in the target sentence; the coding unit is used for coding the hidden state vector of the words in the target sentence to obtain a quasi-density matrix of the target sentence; the measuring unit is used for carrying out projection measurement on the quasi-density matrix of the target sentence to obtain the semantic features of the target sentence; wherein, the coding unit is specifically configured to: normalizing the hidden state vector of the words in the target sentence to obtain a quantum state vector of the words in the target sentence; Carrying out outer product calculation on the quantum state vector of the word in the target sentence to obtain a density matrix of the word in the target sentence; and obtaining a quasi-density matrix of the target sentence according to the density matrix of the words in the target sentence and the weight coefficient corresponding to each word in the target sentence.
7. A customer service outcome prediction model building apparatus comprising a processor, a memory and a program stored on the memory and executable on the processor, the program when executed by the processor implementing the steps of the customer service outcome prediction model building method according to any of claims 1 to 5.
8. A readable storage medium, wherein a program is stored on the readable storage medium, which when executed by a processor, implements the steps in the customer service outcome prediction model building method according to any one of claims 1 to 5.

Description

Customer service result prediction model establishment method, device and equipment Technical Field The present invention relates to the field of artificial intelligence technologies, and in particular, to a method, an apparatus, and a device for establishing a customer service result prediction model. Background Communication operators accumulate a large amount of customer service data such as customer service voice, outbound voice, complaint work orders, online service records, service labels, etc., but currently have shortcomings in terms of complaint data value mining. The call voice data of the hotline customer service contain rich semantic information, but the part of the information is not effectively utilized, and potential dissatisfaction or complaint tendency of customers can be known in advance through mining the rich semantic information in the complaint voice data, so that operators can communicate and pacify potential complaint users in time, and service experience is continuously improved. The prior method has been studied by a series of customer complaint prediction models (customer service result prediction models), but the prior method still has the problems that firstly, complaint scenes are not fully combined and data are fully utilized. For example, currently, most of the customer complaint prediction models stay in customer feature engineering, and the hotline speech data is ignored. The voice data contains the contents of user requirements, user satisfaction conditions, complaint trends and the like, and contains rich semantic information. Therefore, the hotline voice data has great value for researching potential complaints of users, and is also helpful for improving the accuracy and reliability of the prediction result. Secondly, the modeling capability of the upper part and the lower part Wen Yuyi of the existing complaint prediction model is insufficient, and the performance is still improved. Although some studies of user complaint predictive models in non-telecommunications fields are currently performed by text data, the models employed do not mine the semantics of the words in the context of the context, i.e. do not adequately model the semantic interactions between words. However, according to current research, word context semantics tend to aid in understanding text content. That is, the existing service prediction model models the problem that semantic features in voice data are not considered, resulting in inaccurate prediction of service results. Disclosure of Invention The embodiment of the invention provides a method, a device and equipment for establishing a customer service result prediction model, which are used for solving the problem that in the prior art, semantic features in voice data are not considered when the customer service result prediction model is modeled, so that the prediction of a customer service result is inaccurate. In order to solve the technical problems, the embodiment of the invention provides the following technical scheme: the embodiment of the invention provides a method for establishing a customer service result prediction model, which comprises the following steps: Obtaining a training text data set according to the historical customer service voice data; Carrying out quantum measurement processing on the target sentence text training data in the training text data set to obtain semantic features of the target sentence; obtaining semantic features of the target dialogue according to the semantic features of the target sentence by using a bidirectional long-short-term memory BiLSTM model and a self-attention mechanism; Obtaining a customer service result prediction model according to semantic features of a target dialogue and a history customer service result corresponding to the target dialogue; Wherein the target dialogue is any dialogue in the training text data set; The target sentence is any sentence in the target dialogue. Optionally, the method further comprises: And obtaining a customer service result corresponding to the voice data to be predicted according to the voice data to be predicted by using the customer service result prediction model. Optionally, the quantum measurement processing is performed on the target sentence text training data in the training text data set to obtain semantic features of the target sentence, including: Word vector embedding processing is carried out on the text data of the target sentence, so that word vectors of words in the target sentence are obtained; processing word vectors of words in the target sentence by utilizing BiLSTM models to obtain hidden state vectors of words in the target sentence; Coding the hidden state vector of the words in the target sentence to obtain a quasi-density matrix of the target sentence; and carrying out projection measurement on the quasi-density matrix of the target sentence to obtain the semantic feature of the target sentence. Optionally, the processing, by usi