Thesis
French
ID: <
10670/1.r4wh3t>
Abstract
The work that we present is part of a collaborative research project between the University Grenoble Alpes and the company Comongo. Our research focuses on the field of Natural Language Processing (NLP) and more particularly on the automatic analysis of written productions. The core business of the company is the image management of persons. In a first step, our work consisted in transposing a focus group based approach into a distant digital practice. The data we are working on are written productions (answers to open questions) that we call feelings. In a second step, we developed a Natural Language Processing tool and integrated semantic esources to this approach for data processing.This digital transition led us to make two main hypothesis : 1. semantic resources allow a better understanding of the data, however an incremental approach to improve these resources may be necessary in order to obtain optimal results 2. the digital transition has an impact on the quality of the data and the processing.The purpose of the tool is to identify and group similar ideas in productions of very low volume (micro-corpus). This discards any notion of machine learning and this is why we propose a new way to process data with an industrial need.