NER_corpus_chinese Dataset Overview

Main Corpora

1. People's Daily 1998 Edition:

   • Used for word segmentation training.
   • Entity tags include /t, /nr, /ns, /nt.

2. MSRA Corpus:

   • Annotated in BIO format.
   • Contains three entity types: person, location, organization.

3. Boson NLP Corpus:

   • Contains 2,000 paragraphs.
   • Annotated with six entity types, including time, company, and product names.
   • Small scale, about 1 MB.

Additional Research Corpora

1. People's Daily 2014 Edition:

   • Annotation format differs significantly from the 1998 edition.
   • More fine‑grained POS tags and supports nested entity annotations.
   • Size about 17.5 million characters, requiring complex preprocessing.

2. Unnamed Corpus:

   • Annotated in BIO format.
   • Contains three entity types: person, location, organization.
   • Size about 1.3 million characters.