Wei Lu and Min-Yen Kan
AIRS 2005 (Jeju Island, Korea)
12
/
22
Evaluation on Metrics Analysis
•
Language-specific metrics are worth investigating
•
Edit distance alone is not sufficient to build a good categorizer
From the evaluation result we can observe that using standard complexity metrics alone does not work very well in the categorization task.
Certain language-specific metrics are worth investigating. Complexity metrics alone can not perform categorization task well.
Another finding here is that edit distance alone is not sufficient to build a good categorizer, as it can not outperform the baseline.
