Loading...
Please wait, while we are loading the content...
Similar Documents
An Information Theoretic Approach to Quantifying Text Interestingness
| Content Provider | Semantic Scholar |
|---|---|
| Copyright Year | 2014 |
| Abstract | We study the problem of automatic prediction of text interestingness and present an information theoretic approach for quantifying it in terms of topic diversity. Our hypothesis is, in many text domains, often an interesting concept is generated by mixing a diverse set of topics. Given a word distributional model, we present an approach that leverages Jensen-Shannon divergence for measuring text diversity and demonstrate how such a measure correlates with text interestingness. We describe several different base-line algorithms and present results over two different data sets: a collection of e-commerce products from eBay, and a corpus of NSF proposals. |
| File Format | PDF HTM / HTML |
| Alternate Webpage(s) | https://users.soe.ucsc.edu/~mderezin/nips2014.pdf |
| Alternate Webpage(s) | https://users.soe.ucsc.edu/~mderezin/interestingness.pdf |
| Language | English |
| Access Restriction | Open |
| Content Type | Text |
| Resource Type | Article |
