The Speech and Natural Language Processing groups at Columbia Engineering do fundamental research in language understanding and generation with applications to a wide variety of topics, including summarization, argumentation, persuasion, sentiment (including cross-lingual), dialogue modeling, understanding figurative language, detecting deceptive, trusted, emotional and charismatic speech, text-to-speech synthesis, prosody, and analysis of social media to detect sarcasm, fake information, potential for gang violence, mental illness, expressions of emotional distress, abusive language, and radicalization. Many of these projects have been done in a multilingual setting and several have included multimodal information. Other research topics include exploring new machine learning techniques and human-machine interaction. We have also previously carried out research in TTS for Low Resource Languages, code-switching, text-to-scene generation and multimedia explanation.
Descriptions of current projects are available in the following locations:
These groups collaborate closely on many research projects with each other, with language faculty in other universities, and with Columbia faculty in other disciplines. They also mentor a very large number of master’s and undergraduate research project students who participate in their research each semester. They organize regular talks for faculty, students, and the larger New York area community and give many talks themselves at conferences and workshops.
