Abstract
Evaluation of text chunking is revisited. The proposed method tries to analyze the errors made by a chunker and formulates an evaluation strategy that brings out the strength and weakness of a chunker in a better way than the existing precision, recall and F score based methods or their variants do. A tree-matching based algorithm of linear time complexity is designed, analyzed, and illustrated by giving examples. Correctness of the algorithm is checked by using a chunker and a set of test sentences.
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have