140428-Fang Zhao

From cslt Wiki
Jump to: navigation, search

Accomplished this week

  • IDF trained on China daily news. The similar performance was achieved. TF/IDF is worse than IDF.
  • A bug fixed. 1048/2077 pattern matched. 479 within the 1029 mismatches get the same answer.
  • Add weight on tag for test. Tag IDF does not work
  • Add the words with in tags into the segmentation lexicon.

Planned for next week

  • Add weight parameter on tag for test
  • Analysis error
  • Keep modifying code