Lexicon-based emotion analysis in Turkish

This thesis presents a new dataset and a new lexicon for emotion analysis studies in Turkish text. To gather this dataset, we conducted a survey and collected 27,350 entries from 4,709 individuals. Then, we performed a validation process in which annotators validated each entry one by one by assigning a related emotion category. As a result, we obtained two datasets, one raw and the other validated. Subsequently, we generated four versions of these two datasets using two different stemming methods and then modeled them using a vector space model. Then, we ran machine learning algorithms on the models to calculate the accuracy, precision, recall and F measure values. Based on the results we obtained, we concluded that the SVM classifier yielded the highest performance value and that the models trained with a validated dataset provide more accurate results than the models trained with a non validated dataset. In the second phase of the thesis, we propose a lexicon for the use of lexicon-based emotion analysis in Turkish text by using the dataset we constructed within the thesis. We explored the effects of stemming, term weighting, lexicon enrichment and term selection approaches for lexicon-based emotion analysis. We first pre-processed the documents (entries) to obtain stems of each term using different approaches. Afterward, we proposed two different weighting schemas based on term class frequencies and Mutual Information values. Next, we examined bi-grams and concept hierarchy for lexicon enrichment. Furthermore, we applied term selection for efficiency issues. Lastly, we evaluated the performance of the lexicon by using keyword-spotting technique on a separate Turkish dataset. The experiments showed that use of our proposed lexicon in keyword spotting technique produces a satisfactory result in emotion analysis in Turkish Text.