Date of Award

Spring 1-1-2015

Document Type


Degree Name

Doctor of Philosophy (PhD)



First Advisor

Martha Palmer

Second Advisor

Bhuvana Narasimhan

Third Advisor

Laura Michaelis

Fourth Advisor

James Martin

Fifth Advisor

Miriam Butt


Complex predicates that comprise of a noun and verb e.g. yaad kar 'memory do; remember' are a productive class of multi-words in Hindi. In this thesis, we examine the challenges of identification and representation for these complex predicates in Hindi. We design and implement their representation in a lexical semantic resource as well as in lexicalized computational grammars. As productive multi-word predicates, their accurate identification is a necessity for natural language processing applications. We use a combination of linguistic and computational approaches to address these challenges. We use these methods to demonstrate the semi-automatic creation of subcategorization frames for Hindi and the development of classes for nominal predicates. Finally, we demonstrate how linguistic features and computational tools can be used in tandem to automatically identify complex predicates from unseen text.