Affective analysis and modeling of spoken dialogue transcripts

Summarization: At this thesis we propose affective models for the emotional analysis of lexical units in various granularity levels. We propose and evaluate the use of an affective-semantic model to expand the affective lexica of German, Greek, English, Spanish and Portuguese. Moti- vated by the assumption that semantic similarity implies affective similarity, we use word level semantic similarity scores as semantic features to estimate their corresponding af- fective scores. Various context-based semantic similarity metrics are investigated using contextual features that include both words and character n-grams. The model produces continuous affective ratings in three dimensions (valence, arousal and dominance) for all five languages, achieving consistent performance. We achieve classification accuracy (va- lence polarity task) between 85% and 91% for all five languages. For morphologically rich languages the proposed use of character n-grams is shown to improve performance. Moreover, we created the first Greek affective lexicon, translating the words of the English affective lexicon ANEW and assigning them to native speakers for affective annotation. It contains human ratings for the three continuous affective dimensions of valence, arousal and dominance for 1034 words. Motivated by recent advances in the area of Compositional Distributional Semantic Models (CDSMs), we propose a compositional approach for estimating continuous affective ratings for adjective-noun (AN) and noun-noun (NN) pairs. The ratings are computed for the three basic dimensions of continuous affective spaces, namely, valence, arousal and dominance. We propose that similarly to the semantic modification that underlies CDSMs, affective modification may occur within the framework of affective spaces, especially when the constituent words of the linguistic structures under investigation form modifier-head pairs (e.g., AN and NN). The affective content of the entire structure is determined from the interaction between the respective constituents, i.e., the affect conveyed by the head is altered by the modifier. In addition, we investigate the fusion of the proposed model with the semantic-affective model proposed in literature applied both at word- and phrase- level. The automatically computed affective ratings were evaluated against human ratings in terms of correlation. The most accurate estimates are achieved via fusion and absolute performance improvement up to 5% and 4% is reported for NN and AN, respectively. We also investigate text based models for the affective analysis of sentences that are mainly based on affective features. We investigate various datasets including news head- lines, movie subtitles, Twitter status updates and spoken dialogue transcriptions and the best (state-of-the-art) perfo