mihai surdeanu
mihai AT surdeanu DOT name


Here you can download some of the NLP software I have written. I am currently using most of it, which might indicate that it has some degree of usefulness.

  • Ensemble: Linearly-interpolated Dependency Parsers

    This code implements a linear interpolation of several linear-time parsing models (all based on MaltParser). Each individual parser runs in its own thread, which means that, if a sufficient number of cores are available, the overall runtime is essentially similar to a single Malt parser. The resulting parser has state of the art performance yet it remains very fast.

  • SwiRL: The Semantic Role Labeler

    SwiRL is a Semantic Role Labeling (SRL) system for English constructed on top of the full syntactic analysis of text. Achieved state-of-the-art performance in the CoNLL 2005 SRL evaluation.

  • Bios: Suite of Syntactico-Semantic Analyzers

    Includes a named-entity recognizer, a syntactic chunker, a POS tagger, and a "smart" tokenizer. All processors are learned using the MiLL machine learning library (see below).

  • MiLL: Machine Learning Library

    Includes SVM, Maximum Entropy and Perceptron classifiers under a unique and simple interface. All algorithms support mult-class problems. MiLL includes the novel Perceptron algorithm with dynamic uneven margins I designed for my ACE Information Extraction system (see the publication page). MiLL is distributed together with BIOS but it can be used independently of BIOS, for any ML task.

  • Spear: Syntactic Parser

    Syntactic parser heavily based on Michael Collins' Model 1 parser. The Spear package includes also a corpus of parsed questions I created from the TREC 8 - 12 evaluations. This corpus was crucial in improving the parser performance on questions.


last updated on june 8, 2010