Emotion in Text page of the Sail Emotion Research Group
The purpose of this page is to (1) describe our research in textual emotions, (2) distribute papers and programs we have written, and (3) put a web front-end on the emotion in text algorithms we’ve been working on.
Our research into emotion in text focuses on ways to detect
and classify when text has emotional features. Below in the
"Applications" section we describe some ways of doing this. For these,
the display of the results is an interesting aspect of
emotion research that attempts to use the potential of the web as an
emotional text interface . Underlying the display is an
emotional markup language for text . Furthermore, the web is an interactive medium that may collect human response data that
could be used to inform the algorithms.
Today is May 25, 2013. The last time this page has been updated was January 10, 2011.
Using Whissell’s Dictionary of Affect
Cynthia Whissell’s Dictionary of Affect in Language (DAL) is a dictionary that for each of approximately 4000 English words there is a 3-dimensional value that represents that word’s position in an emotion space. These dimensions are 1) Evaluation/valence/pleasantness, 2) activation/arousal, and 3) imagery. For more info, c.f. Plutchik and Kellerman v.4 .
More info...
Fuzzy
Logic Experiment 1 This
experiment is a survey that collects human judgements and encodes it
into interval type-2 fuzzy sets. It will take about 20-40 minutes to complete
Fuzzy Logic Experiment 2
This experiment is a survey that collects human judgements and encodes it into interval type-2 fuzzy sets. It extends the previous experiment to a bigger vocabulary
Fuzzy Logic Experiment 3: Iemocap Vocab
This experiment is a survey that collects human judgements and encodes it into interval type-2 fuzzy sets. It extends the previous experiment to the emotional vocabulary used in the IEMOCAP database . It should take approximately 10-15 minutes. (Released Dec 10, 2008)
Fuzzy Logic Experiment 4: Iemocap Vocab With Modfiers
This experiment is a survey that collects human judgements and encodes it into interval type-2 fuzzy sets. It uses the emotional vocabulary from the IEMOCAP corpus but extends them with the modifiers "very" and "sort of". It should take approximately 20-30 minutes. (Released Dec 10, 2008)
TF/IDF Approach
Coming soon..................................................................................................................
3rd word net distance
Coming soon
Latent Semantic Analysis
Coming soon