My main research interests are Natural Language Processing and Representation Learning.
I mainly work on multi-modal approaches that combine natural language and video, having developed models to extract fine-grained opinions from product review videos and for temporal video grounding.
I am also interested in studying document changes, or edits, which lie at the core of version control systems such as GitHub and Wikipedia. I have worked on learning edit representations that can be useful for tasks such as automatic commit message generation in GitHub, and better quality assessment of Wikipedia articles.
My interests also extend to the study of affect in text and video as a more general issue, including some arguably more complex linguistic affective phenomena such as irony and sarcasm.