open-gram

open-gram is a project tries to collect lexicon and build n-gram dataset for NLP in Chinese. This project tries to leverage existing open source resources like crfpp and CC-CEDICT.

open-gram includes 4 parts

• corpus collection
• segmentation
• (new) word extraction
• n-gram info counting

corpus collection

1. crawl Chinese web sites using scrapy, grab the body HTML pages of them
2. proprocess the pages - detect the encoding - remove HTML tags and other stuff we are not interested in - split the text into sentences

segmentation

there two ways to segment tokens into words

• tagging
• matching

word extraction

n-gram info counting