Towards Cross-Language Word Sense Disambiguation for Quechua

In this paper we present initial work on cross-language word sense disambiguation for translating adjectives from Spanish to Quechua and situate CLWSD as part of the translation task. While there are many available resources for training Spanish-language NLP systems, linguistic resources for Quechua, especially Spanish-Quechua bitext, are quite limited, so some ingenuity is required in developing Spanish-Quechua systems. This work makes use of only freely available resources and compares a few different techniques for CLWSD, including classifiers with simple word context features, features from a Spanish-language dependency parser, a multilingual version of the Lesk algorithm, and a distance metric based on the Spanish wordnet.