Abstract:
The aim of this project is developing a web application to extract software product features from a requirement document written specifically for that project using Natural Language Processing. The extracted features will be plotted in the form of a tree in hierarchical manner. A feature is defined as a user visible aspect in a software product. This project targets the functional requirements of a software product written in English, in text file, in the form of a paragraph. The purpose of this project is to aid software developers to get an initial model of a software product in terms of its features. In this project machine learning approach is also used to perform different Natural Language Processing tasks such as: sentence segmentation, lemmatization and parts of speech tagging all the before mentioned tasks are done using pre-trained models. The core functionality is identifying features out of a segmented sentence and is done by extracting set of Triple Relation out of segmented sentence. A Triple is defined in terms of Actor, Action and Object and they are interpreted in way that an Actor will perform an Action on an Object. The set of Triples are then filtered to extract only meaning full words with in the Actor, Action and Object. Next step is to remove redundancy among filtered Triples. The filtered triples are then shown in the form of a tree in hierarchal manner on GUI. For Natural Language Processing libraries like Apache Open NLP and Stanford Open IE are used and JAVA is used as the core development language and front-end is made using JSP & Servlets with HTMLlCSS.
