Emotions from Text: Machine Learning for Text-based Emotion Prediction

Proceedings of Human Language Technology Conference and Conference on Empirical Methods in Natural Language Processing (HLT/EMNLP), pages 579–586, Vancouver, October 2005. c⃝2005 Association for Computational Linguistics Emotions from text: machine learning for text-based emotion prediction Cecilia Ovesdotter Alm∗ Dept. of Linguistics UIUC Illinois, USA ebbaalm@uiuc.edu Dan Roth Dept. of Computer Science UIUC Illinois, USA danr@uiuc.edu Richard Sproat Dept. of Linguistics Dept. of Electrical Eng. UIUC Illinois, USA rws@uiuc.edu Abstract In addition to information, text con- tains attitudinal, and more specifically, emotional content. This paper explores the text-based emotion prediction prob- lem empirically, using supervised machine learning with the SNoW learning archi- tecture. The goal is to classify the emo- tional affinity of sentences in the narra- tive domain of children’s fairy tales, for subsequent usage in appropriate expres- sive rendering of text-to-speech synthe- sis. Initial experiments on a preliminary data set of 22 fairy tales show encourag- ing results over a na¨ıve baseline and BOW approach for classification of emotional versus non-emotional contents, with some dependency on parameter tuning. We also discuss results for a tripartite model which covers emotional valence, as well as feature set alternations. In addition, we present plans for a more cognitively sound sequential model, taking into considera- tion a larger set of basic emotions. 1 Introduction Text does not only communicate informative con- tents, but also attitudinal information, including emotional states. The following reports on an em- pirical study of text-based emotion prediction. Section 2 gives a brief overview of the intended application area, whereas section 3 summarizes re- lated work. Next, section 4 explains the empirical study, including the machine learning model, the corpus, the feature set, parameter tuning, etc. Sec- tion 5 presents experimental results from two classi- fication tasks and feature set modifications. Section 6 describes the agenda for refining the model, before presenting concluding remarks in 7. 2 Application area: Text-to-speech Narrative text is often especially prone to having emotional contents. In the literary genre of fairy tales, emotions such as HAPPINESS and ANGER and related cognitive states, e.g. LOVE or HATE, become integral parts of the story plot, and thus are of par- ticular importance. Moreover, the story teller read- ing the story interprets emotions in order to orally convey the story in a fashion which makes the story come alive and catches the listeners’ attention. In speech, speakers effectively express emotions by modifying prosody, including pitch, intensity, and durational cues in the speech signal. Thus, in order to make text-to-speech synthesis sound as nat- ural and engaging as possible, it is important to con- vey the emotional stance in the text. However, this implies first having identified the appropriate emo- tional meaning of the corresponding text passage. Thus, an application for emotional text-to-speech synthesis has to solve two basic problems. First, what emotion or emotions most appropriately de- scribe a certain text passage, and second, given a text passage and a specified emotional mark-up, how to render the prosodic contour in order to convey the emotional content, (Cahn, 1990). …

Content truncated to 500 words