Abstract:
We present our sentence encoding approach that uses local contexts constructed from pq-gram representations of a sentence’s dependency tree. The context localization scop...Show MoreMetadata
Abstract:
We present our sentence encoding approach that uses local contexts constructed from pq-gram representations of a sentence’s dependency tree. The context localization scope can be adjusted through parameters p, the dependency depth, and q, the dependency width, which allows controlling context-sensitivity. We show competitive results of using our sentence encoding approach for sentence-pair modeling tasks.
Date of Conference: 01-03 November 2018
Date Added to IEEE Xplore: 13 December 2018
ISBN Information: