Term Spotting: A Quick-and-dirty Method for Extracting Typological Features of Language from Grammatical Descriptions

Starting from a large collection of digitized raw-text descriptions languages the world, we address problem extracting information interest to linguists these. We describe general technique extract properties described associated with specific term. The is simple implement, explain, requires no training data or annotation, and manual tuning thresholds. results are evaluated on gold standard database classifiers accuracy that match supersede human inter-coder agreement similar tasks. Although competitive, method may still be enhanced by more rigorous probabilistic background theory usage extant NLP tools for morphological variants, collocations vector-space semantics.