Sentimenduen Analisia Ikasketa Automatikoaren laguntzaz

Abstract

Proiektu honetan Sentimenduen Analisia lantzea izan da helburua. Analisi hori, makina bat idatzizko testuak gizakien antzera interpretatzeko gai izatean datza. Datu horiek adierazten dituzten sentimenduak detektatu, hala nola, poztasuna, haserrea, tristura eta halakoak, eta elkarrengandik bereizteko gai izatea da gakoa. Proiektu hau, ordea, sentimendu orokor batzuetara mugatu da, testu bat ea positiboa, neutroa ala negatiboa den jakitera zehazki. Makina bat hori egiteko gai izan dadin Ikasketa Automatikoa aplikatu behar zaio, eta horretarako WEKA softwarea erabili da.

WEKAren bidez datu jakin batzuk entrenatu dira, eta horien ikasketa burutu eta ebaluatu da. Entrenamendua egiteko hainbat metodo desberdin aplikatu dira, eta exekuzio bakoitzarekin emaitza batzuk lortu dira. Emaitza horiek sakonki aztertuz hainbat ondorio atera dira, eta guztiak lanaren memoria honetan ahalik eta ongien azaldu dira.

The aim of this project was to work on the Sentiment Analysis. This analysis consists of a machine being able to interpret written texts in a human-like way. The key is to be able to detect the feelings expressed by these data, such as joy, anger, sadness, and so on, and to be able to distinguish them from each other. This project, however, was limited to some general sentiments, such as whether a text is positive, neutral, or negative. A machine, in order to be able to do that, must learn through Machine Learning, which can be applied using the WEKA software. Using WEKA, certain data have been trained, studied and evaluated. Several different methods of training have been applied, and some results have been achieved with each execution. An in-depth analysis of these results has led to a number of conclusions, all of which have been explained as best as possible in this written memory.