- Berkeley Entity Resolution System
Entity resolution system accompanying our TACL 2014 paper. Jointly predicts coreference information (which spans of text in the document refer to the same entity), semantic types (what kind of entity each span of text refers to), and entity links to Wikipedia (which exact entity a span refers to).
- Berkeley Document Summarizer
Single-document summarization system released along with our ACL 2016 paper. Features learned on a large corpus of New York Times documents paired with summaries allow us to identify important document content, and constraints in generation ensure grammaticality and coherence of references even in the presence of compression.
- Berkeley Coreference Resolution System
Stand-alone coreference component of the Berkeley Entity Resolution System from our EMNLP 2013 work.
- Neural CRF Parser
Syntactic constituency parser for a range of languages based on our ACL 2015 paper.
- Morphological Inflecter (1MB tgz)
Morphological inflection tool as described in our NAACL 2013 paper. Trains on data from Wiktionary (or other lexical resources) for a particular language and can then predict how to inflect unseen words in that language (how to conjugate verbs, decline nouns, etc.). Download the accompanying dataset (10MB tgz).