The NLP pipeline

The NLP pipeline consists of five major components: a parser, an interpreter, a grammar with semantic translations that drives both, an automated reasoner, and finite-state tools for morphological and phonological processing (though the latter are not currently integrated into the parser). A conversational agent packages up these components into a test-bed application.

• Conversational agent — selkie.nlp.bot
• Grammars
  • Grammars — selkie.nlp.grammar
  • Grammar development tool — selkie.nlp.gdev
  • Feature vectors — selkie.nlp.features
  • Recursive feature structures — selkie.nlp.avs
  • English vocabulary — selkie.nlp.eng
• Chart parser
  • Constituent parser — selkie.nlp.parser
  • Trees — selkie.nlp.tree
  • Dependency conversion — selkie.nlp.dep
  • Head marking rules — selkie.nlp.head
• Interpreter and reasoner
  • Semantic interpretation — selkie.nlp.interp
  • Predicate calculus expressions — selkie.nlp.expr
  • Knowledge base and automated reasoner — selkie.nlp.logic
• Morphology and phonology
  • Finite-state automata — selkie.nlp.fsa
  • Finite-state transducers — selkie.nlp.fst
  • English stemmer — selkie.nlp.stemmer
• Preprocessing
  • Converting text to bare words — selkie.nlp.pretts
  • Tokenizer for Latin scripts — selkie.nlp.tok
• GLab — selkie.nlp.glab
• Praat textgrid — selkie.nlp.textgrid

Other

• Dependency parsers
  • Parser
  • Pseudo-projective parsing
  • Evaluation — selkie.nlp.dp.eval
  • Features — selkie.nlp.dp.features
  • The Nivre parser
  • MST Parser — selkie.nlp.dp.mst
  • Machine learning support