Project abstract for group pedersen

Finding Similar and Related Words in High-Dimensional Spaces and in Ontologies

One of the most fundamental operations in both human processing and computer processing of language is to identify which words are similar or related to each other, and what relations exist between words in a particular sentence. The goal of this research is to develop computational methods that automatically identify sets of related words, and easily adapt to the variations in word meaning that accompany changes in the subject matter and intended audience of a text. The researchers rely on automatic discovery of relations between words in large corpora of text, and also upon information that can be obtained from human curated ontologies. Word meanings are central to language understanding, and success in this research will improve the ability of computer systems to perform translation, retrieve information from the web, and summarize documents. 

A bibliography of this group’s publications acknowledging MSI is attached.