Défense de mémoire en sciences informatiques : application of document embedding for class name recommendation during UML class diagram creation
Défense de mémoire - Capuano Thibaut, à distance via teams
Date : 31/08/2020 14:30 - 31/08/2020 16:00
Lieu : https://teams.microsoft.com/l/meetup-join/19%3ameeting_YWI1MjczZTMtM2YwZS00YzNkLTg0ODUtNjlmMTM4NDkzZWFm%40thread.v2/0?context=%7b%22Tid%22%3a%225f31c5b4-f2e8-4772-8dd6-f268037b1eca%22%2c%22Oid%22%3a%22363e7449-861f-4e7f-bbfa-9e6261d7a69a%22%7d
Orateur(s) : Thibaut Capuano
Organisateur(s) : Isabelle Daelman
System quality is an important aspect during development. But, while code quality has an important place during system development, system engineering techniques are generally not fully exploited. This research propose a new approach in order to promote system engineering. It contributes to system engineering by proposing to apply information gathered in source code to help users during class diagram creation. This work uses of machine learning to recommend class names to the user. Few approaches use machine learning with class diagrams even though it has shown to be useful for recommendation systems in similar fields. Document embedding is used on the sequences of relations contained in code. Based on the partial diagram already drawn by the user, the embedding suggests similar sequences of relations from which tokens are extracted and then suggested to the user. As a next step, the system also suggests entire class names to the user based on those tokens. The class names are selected from all class names presents in the train set using a full text index. Those class names aims at guiding the user during its reflection in the class diagram creation process.
Contact :
Isabelle Daelman
-
4966
-
isabelle.daelman@unamur.be
Télecharger :
vCal