Word sense disambiguation through lattice learning
Author(s)Stickgold, Eli (Eli B.)
Massachusetts Institute of Technology. Dept. of Electrical Engineering and Computer Science.
Patrick H. Winston.
MetadataShow full item record
The question of how a computer reading a text can go from a word to its meaning is an open and difficult one. The WordNet lexical database uses a system of nested supersets to allow programs to be specific as to what meaning of a word they are using, but a system that picks the correct meaning is still necessary. In an attempt to capture the human understanding of this problem and produce a system that can achieve this goal with minimal starting information, I created the DISAMBIGUATOR program. DISAMBIGUATOR uses Lattice Learning to capture the concept of contexts, which represent common situations that multiple words are found in, and uses Genesis' system of Things, Sequences, Derivative and Relations to understand some contexts as being related to others (i.e. that 'things which can fly to a tree' and 'things which can fly to Spain' are related in that they are both special cases of the context 'things which can fly'). Using this system, DISAMBIGUATOR can tell us which meaning of 'hawk' we should use if we see it in a sentence like 'the hawk flew to the tree.' DISAMBIGUATOR is implemented in Java as part of the Genesis system, and can disambiguate short stories of around ten related statements with only a single query to the user.
Thesis (M. Eng.)--Massachusetts Institute of Technology, Dept. of Electrical Engineering and Computer Science, 2011.Cataloged from PDF version of thesis.Includes bibliographical references (p. 51).
DepartmentMassachusetts Institute of Technology. Dept. of Electrical Engineering and Computer Science.
Massachusetts Institute of Technology
Electrical Engineering and Computer Science.