On lexical level matching
Shaoshi Ling · Illinois Digital Environment for Access to Learning and Scholarship (University of Illinois at Urbana-Champaign) · 2017
In many natural language understanding applications, text processing requires comparing lexical units: words, phrases, name entities and sentences. A significant amount of research has taken place in studying evaluating similarity metrics between those units. In this thesis, we summarize some research work in computing lexical similarity. We describe a new approach to compute similarity between two spans of text, using multiple semantic-units level comparison measures to compute sentence-level similarity scores.