Loading...
Please wait, while we are loading the content...
Similar Documents
Personalized Emotion Classification with Latent Dirichlet Allocation
| Content Provider | Semantic Scholar |
|---|---|
| Author | Tang, Kui |
| Copyright Year | 2012 |
| Abstract | The inability for machines to infer emotion handicaps user interaction in all applications and limits accuracy of applications that monitor human behavior, such as anti-fraud systems. Related work centers on one-dimensional representations of emotion as well as time-series analysis of aggregate sentiment of many people, both of which offer limited utility to a designer of emotionally-aware applications. In this project, we derive a personal, probabilistic emotional lexicon from Tumblr blogs. This lexicon can then be used to infer future emotional state. We take a semi-supervised approach with latent Dirichlet allocation (LDA) and a derivative, SubjLDA, initializing our model with labeled emotional words with a small nondomain-specific lexicon and allowing the model to classify the remaining words and documents into emotional categories. |
| File Format | PDF HTM / HTML |
| Alternate Webpage(s) | http://kui-tang.com/pdf/2012_Personalized.pdf |
| Language | English |
| Access Restriction | Open |
| Content Type | Text |
| Resource Type | Article |
