A One-to-Many Requirements Analysis and Recognition Method Based on Deep Learning

Abstract

The invention discloses a one-to-many demand analysis and identification method based on deep learning. The method is: 1) constructing a one-to-many deep learning neural network model and training it; wherein, the deep learning neural network model includes an encoder and m decoders; the encoder is used to calculate the hidden layer of the required text The state vector is sent to each decoder; the decoder processes the input data sequentially to generate a hidden layer state vector; then a global alignment is generated based on the hidden layer state vector generated by the encoder and the hidden layer state vector generated by the decoder weight vector, and then connect the context vector and the hidden layer state vector to the Softmax layer for prediction and output a function item; 2) use the deep learning neural network model to predict the text to be processed, and generate m function items; 3) merge m Similar function items in one function item, and obtain the function item identification result of the pending requirement text.

Description

technical field [0001] The invention belongs to the field of computer technology, and relates to requirements engineering, natural language processing and other technologies, especially requirements understanding technology, which is used to solve the problem of automatically extracting multiple function items from a requirement text, and at the same time, a requirement text corresponds to multiple function items in this scene A solution to the one-to-many problem based on deep learning is proposed, which can also be used to solve the one-to-many sequence generation problem in other similar fields. Background technique [0002] Requirements documents are the beginning of software development. These requirements documents usually use natural language to describe the wishes of various stakeholders. Afterwards, developers implement these requirements into working software through analysis, design and coding. Requirements analysis is an essential process to understand software f...

AI Technical Summary

Problems solved by technology

[0004] The performance of the traditional function item identification method is low in this scenario. The main reasons are as follows: it is difficult to find a common pattern (pattern), and the rules are too loose when using a common pattern, resulting in the extraction of too many functional items, and the recall rate is low. High, but the accuracy rate is very low; some function items cannot be directly extracted using NLP methods such as syntax and syntax analysis on the basis of the story text, and it is necessary to understand the business rules of the underlying system; the correspondence between stories and function points is one-to-many, Traditional s2s model doesn't work

Images