CONTENTS

FEATURES

Corpus Frequency Features

  • AvgNewsBodyTfIdf
  • AvgNewsTitleTfIdf
  • AvgNewsFirstParagraphTfIdf
  • AvgWikipediaTfIdf
  • NewsBodyTfIdfSum
  • NewsTitleTfIdfSum
  • NewsFirstParagraphTfIdfSum
  • WikipediaTfIdfSum

Context Features

We use Jensen-Shannon divergence and Cosine Similarity as distance measures.

  • AvgNewsJS
  • AvgBlogJS
  • AvgNewsCosine
  • AvgBlogCosine

Query-Only Features

Overhead of Feature Extraction

LEARNING MODEL

Multiple Additive Regression-Trees(MART)