Industrial Strength Dependency Parsing System

Date

2018-05-10

Authors

Journal Title

Journal ISSN

Volume Title

Publisher

Abstract

Dependency parsing is a useful task to help computer understand human language. By parsing the dependency grammar of a sentence automatically, dependency parser produces dependency-based syntactic representations which enhance performance of many language applications, such as machine translation, question answering and information extraction. Recently dependency parsing has attracted considerable interest from researchers and developers in the Natural Language Processing field, and many state-of-art works have achieved high accuracies. But not all of them are applicable for industry applications in terms of runtime speed and memory efficiency. We implemented and evaluated various dependency parsing algorithms, finding out the most practical algorithm in consideration of tradeoff between accuracy and runtime speed. The final achievement is a practically usable dependency parser, which can parse raw sentences to grammar trees. Our parser has been released as open source software and live demonstrated on http://iparser.hankcs.com/.

Description

Keywords

Dependency Parsing, Natural Language Processing, Deep Learning

Citation