This package contains the collection of algorithms and data structure I use for research. Hence, the name is an acronym: brian's arsenal of algorithms for language/learning.

• .crf: Conditional Random Field
  • currently, only a linear chain implementation
• .learn.bhc: Bayesian Hierarchical Clustering/Grouping
  • A port from a MATLAB port from the original C implementation
• .nlp: Natural Language Processing
  • .nlp.corpora:
    • Interfaces and handlers for corpora go here for a variety of datasets
  • .nlp.grammars:
    • Tree Insertion Grammar
  • .nlp.induce:
    • Tree enrichment rules from Michael Collins' dissertation. Mostly ported from Stanford CoreNLP
  • .nlp.lexicon:
    • Words and their properties.
  • .nlp.parse:
    • CKY Tree chart parser for the tree insertion grammar
  • .nlp.semantics:
    • a simple model of hobbsian logical form that uses derived tree structure to produce logical predicates
  • .nlp.generate:
    • a simple model of natural language generation
  • Note: the tree data structure which bears the brunt of the tree grammar work is in the utils.data_structures
• .utils
  • .general:
    • useful utilities to make algorithms nicer
  • .utils.config:
    • Store global configuration modules
  • .utils.config
    • flyweights, singletons, and trees
  • .utils.sugar
    • decorators and other syntactic sugar
  • .utils.vault
    • model storage
  • .montecarlo
  • .vocabulary (borrowed, citation in file)
• .science
  • experiment scripts for reproducability