259,104 annotated tokens across syntactic chunk categories including Noun Phrases (NP), Verb Phrases (VP), and Prepositional Phrases (PP) from the Wall Street Journal corpus. It provides 211,727 training tokens and 47,377 test tokens derived from specific WSJ sections for the CoNLL-2000 shared task.
Use Cases
- Train a sequence labeling model to predict NP, VP, and PP chunk tags for individual tokens.
- Evaluate the performance of shallow parsing algorithms using the standardized section 20 test data.
- Perform error analysis on syntactic boundary detection using the provided phrase-level annotations.
Strengths
- Contains 211,727 training tokens and 47,377 test tokens.
- Includes syntactic chunk labels such as NP (Noun Phrase), VP (Verb Phrase), and PP (Prepositional Phrase).
- Derived from sections 15-18 and section 20 of the Wall Street Journal (WSJ) corpus.
- Features annotations generated by a specialized program from Tilburg University for the CoNLL-2000 shared task.