Shuki Cohen New York University, Department of Psychology, Doctoral Candidate

shukic@yahoo.com


Identifying Speaker's Emotional Processing Patterns through Computerized Text-Analysis of Personal Narratives

Time:Thursday, April 29, 11:30-12:30

Place:CS Conference Room in MUDD

Abstract:
Lately, the social sciences have witnessed a resurgence of interest in language as a marker for personality traits. Most notably, James Pennebaker has shown that language use is a stable and reliable predictor of various indices of physical and mental health. However, contrary to the emphasis current research in personality places on emotional processing, the correlation between the use of negative and positive affect words and external measures of negative emotionality is relatively low, and at times fails to reach statistical significance.

The premise of this study is that the meaning of emotion words is more sensitive to their immediate context than other linguistic features found to be related to mental distress. Thus, emotion words can be used in a non-emotional way (e.g the filler "like"), can be negated, used in a hypothetical manner, or to bear contradictive meanings (e.g. "serious"). To test this hypothesis, a corpus of 500 personal narratives was used, in which psychological information was available for each speaker.

Using corpus-linguistic methodology, inclusion and exclusion criteria were constructed for each marker to ensure the detection of the emotion word in its intended sense. The resulting dictionaries have higher correlations with indices of mental health compared to existing text-analytic dictionaries.

About the speaker: Shuki Cohen is doing a Ph.D. in clinical psychology at NYU. Shuki has a bachelor's degree in Biochemistry and Master's in brain electrophysiology from the Weizmann Institute of Science in Israel. His current research revolves around the relationship between language, personality and emotional instability. Lately he has been interested in the development of word-choice measures to monitor mental states in real time. These measures can be applied to psychotherapy transcripts for a non-invasive (and potentially retrospective) study of the therapeutic process.