Output details
11 - Computer Science and Informatics
Open University
A methodology for automatic identification of nocuous ambiguity
<22>This paper describes the creation and analysis of an original dataset, which demonstrates for the first time that nocuous ambiguity occurs in pronoun resolution ambiguity. Several original mechanisms are used to improve the computational performance of the system: the use of a range of machine learning techniques, and an in-depth analysis of how weak thresholds (an original idea in this paper) can be used to accurately identify which sentences display nocuous ambiguity. The work's significance is in showing that nocuous ambiguity occurs in the heavily-studied area of anaphora resolution, and that it can be automatically identified.