Enhancing the Input Representation: From Complexity to Simplicity.
Year: 2017
Type: preprint
Abstract: We introduce an efficient algorithm for mining informative combinations of attribute-values for a given task. We use informative attribute-values to enhance the input representation of data. We apply our approach to coreference resolution using a simple set of attributes like syntactic roles and string match. With the enhanced representation, a simple coreference model outperforms more complex state-of-the-art models by a large margin. The use of the enhanced representation results in robust improvements in both in-domain and out-of-domain evaluations.
Source: arXiv (Cornell University)
Authors: Nafise Sadat Moosavi, Michael Strube
Language: English
Subfield: Artificial Intelligence
Field: Computer Science
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W2740075669
