About

Libraries

• Murphy is a library of learning algorithms for structured prediction.

Parsing

• The Berkeley Parser infers syntactic annotations. We have an online demo!
• Puck is a lightning-fast version of the Berkeley Parser that uses GPUs.
• Epic is a discriminative parser using many kinds of annotations.
• The neural CRF parser effectively leverages distributed representations of words by scoring anchored rule productions with feedforward neural networks.
• The Berkeley Parser Analyser performs error analysis on parse trees.
• The Berkeley CCG to PST tool converts CCG Derivations into Penn Treebank style trees.

Coreference Resolution

• The Berkeley Entity Resolution System jointly models and predicts named entity chunks, coreference, and entity links.
• The Berkeley Coreference Resolution System is a high-performing English coreference resolution system.
• The Berkeley Coreference Analyser performs error analysis on coreference resolution output.

Summarization

• The Berkeley Document Summarizer is an extractive and compressive single-document summarization system learned on a large newswire corpus.

Historical Document Recognition

• Ocular is a state-of-the-art historical document recognition system.

Word Alignment

• The Berkeley Aligner aligns the words in multilingual parallel texts.

Language Modeling

• The Berkeley LM provides a library for storing large n-gram language models efficiently in memory.

Education

• The Pac-Man projects are a set of class projects that teach foundational AI concepts, such as informed state-space search, probabilistic inference, and reinforcement learning.