Joining Hands: Exploiting Monolingual Treebanks for Parsing of Code-mixing Data

摘要

In this paper, we propose efficient and less resource-intensive strategies for pars ing of code-mixed data. These strategies are not constrained by in-domain anno tations, rather they leverage pre-existing monolingual annotated resources for train ing. We show that these methods can pro duce significantly better results as com pared to an informed baseline. Besides, we also present a data set of 450 Hindi and English code-mixed tweets of Hindi mul tilingual speakers for evaluation. The data set is manually annotated with Universal Dependencies.