Retrieve abstracts or sentences in MEDLINE using deep-parsing results.
List the proteins or genes which interact with the given protein or gene.
Technical terms, parts-of-speech, and syntactic trees are annotated. TEI version of GENIA corpus version 3.0 (Dr. Tomaz Erjavec, Jozef Stefan Institute, Slovenia).
Output phrase structures and predicate argument dependencies. High parsing speed (more than 20 sentences per second) and high accuracy (88-90% accuracy of predicate argument dependencies).
A shift-reduce dependency parser that uses maximum entropy models for scoring parser actions and a best-first strategy to search for the best parse.
Specifically tuned for biomedical texts. POS tagging accuracy of 97-98%. Shallow parsing accuracy of 91-94%.
Optimized for biomedical texts. The classifier achieved an F-score of 99.7 on 200 unseen GENIA abstracts.
Parameter estimation algorithm for feature forests. Support GIS, IIS, and limited-memory BFGS
Fast parameter estimation using the BLMVM algorithm. Modelling with inequality constraints.
A logic programming language similar to Prolog. Manipulation of feature structures as builtin data structure. High-speed runtime system. C++ library support for feature structures.
Intended to work with LiLFeS.
The conversion guarantees strong equivalence of grammar.