A method for implementing a dialogue robot that recommends reply content to human customer service

Abstract

The present invention provides a method for implementing a dialogue robot that recommends reply content to artificial customer service, comprising the following steps: Step 1, extracting features from the question through the first LSTM neural tensor network model, obtaining the word segmentation result of the customer question, and using the second An LSTM neural tensor network model encodes the word segmentation results of customer questions into vector c; step 2, extracts features from customer service replies through the second LSTM neural tensor network model, obtains the word segmentation results of customer service replies and uses the second LSTM The neural tensor network model encodes the result of the word segmentation of the customer service reply into a vector r; step 3, obtains the personalized characteristics of the store where the problem is located, and encodes it into the characteristic vector s of the merchant; step 4, converts the vector c. The vector r and the vector s are directly interactively calculated through tensors; step 5, output the selection and recommendation function options of similar answers corresponding to the customer questions. The present invention can intelligently recommend reply content to artificial customer service.

Description

technical field [0001] The invention relates to the field of computer networks, in particular to a method for realizing a customer service dialogue robot that recommends reply content to manual customer service. Background technique [0002] The customer service dialogue system is a dialogue system in a specific field. At present, this field is a relatively cutting-edge research content. The main goal is to realize the automatic content reply of customer service and improve the efficiency of solving customer problems. [0003] Community question answering system (CQA) is a platform where users can share their professional knowledge with questioners. This kind of question answering data is somewhat similar to the customer service system, and some solutions in CQA can usually be used for reference. However, the answers in the customer service dialogue system are often the answers depending on the situation, and the attributes of the goods sold by different merchants are differ...

AI Technical Summary

Problems solved by technology

[0005] The existing Convolutional Neural Tensor Network (CNTN) model uses the convolutional neural network to extract the features of the sentence. The convolutional neural network extracts mainly the spatial features through the filter and remembers the overall information of the sentence. There is a progressive sequence in the question Relationship, the previous text description is the problem background, and the last problem is the core problem, so the convolutional neural network is not suitable for dealing with long dependencies and time series problems

Images