A5045037642- Natural Language Processing Techniques
- Speech and dialogue systems
- Speech Recognition and Synthesis
- Phonetics and Phonology Research
- Topic Modeling
- Language, Discourse, Communication Strategies
- Language, Metaphor, and Cognition
- Linguistic Variation and Morphology
- Music and Audio Processing
- Deception detection and forensic psychology
- Digital Communication and Language
- Speech and Audio Processing
- Video Analysis and Summarization
- Personal Information Management and User Behavior
- Hate Speech and Cyberbullying Detection
- Authorship Attribution and Profiling
- Multi-Agent Systems and Negotiation
- Text Readability and Simplification
- Syntax, Semantics, Linguistic Variation
- Usability and User Interface Design
- Sentiment Analysis and Opinion Mining
- Advanced Text Analysis Techniques
- Psychopathy, Forensic Psychiatry, Sexual Offending
- Emotion and Mood Recognition
- Semantic Web and Ontologies
Columbia University
2014-2024
Carnegie Mellon University
2023
Amazon (United States)
2022-2023
Laboratoire d'Informatique de Paris-Nord
2020-2021
Center for Applied Linguistics
2018
Association for Computational Linguistics
2018
City University of New York
2011
IBM (United States)
2010
Swedish e-Science Research Centre
2009
AT&T (United States)
1995-2007
Cue phrases are linguistic expressions such as now and well that function explicit indicators of the structure a discourse. For example, may signal beginning subtopic or return to previous topic, while mark subsequent material response prior material, an explanatory comment. However, cue convey discourse structure, each also has one more alternate uses. While incidentally be used sententially adverbial, for use initiates digression. Although distinguishing sentential uses is critical...
The INTERSPEECH 2016 Computational Paralinguistics Challenge addresses three different problems for the first time in research competition under well-defined conditions: classification of deceptive vs. non-deceptive speech, estimation degree sincerity, and identification native language out eleven L1 classes English L2 speakers.In this paper, we describe these sub-challenges, their conditions, baseline feature extraction classifiers, resulting baselines, as provided to participants.
In September 2016, Stanford's "One Hundred Year Study on Artificial Intelligence" project (AI100) issued the first report of its planned long-term periodic assessment artificial intelligence (AI) and impact society. It was written by a panel 17 study authors, each whom is deeply rooted in AI research, chaired Peter Stone University Texas at Austin. The report, entitled "Artificial Intelligence Life 2030," examines eight domains typical urban settings which likely to have over coming years:...
We explored general issues concerning personal information management by investigating the characteristics of office workers' paper-based information, in an industrial research environment. we examined reasons people collect paper, types data they collect, problems encountered handling and strategies used for processing it. tested three specific hypotheses course move. The greater availability public digital along with changes people's jobs or interests should lead to wholescale discarding...
The absence of intonational prominence on a referring expression ( deaccentuation) is commonly explained as consequence the GIVENness discourse entity referred to - fact that it represents old information in discourse. However, speakers sometimes use accented expressions refer such GIVEN entities, so not sufficient explanation for deaccentuation. It has also been suggested tend express entities grammatical subjects and mention them early utterance. present work investigates contributions...
We describe a statistical approach for modeling agreements and disagreements in conversational interaction. Our first identifies adjacency pairs using maximum entropy ranking based on set of lexical, durational, structural features that look both forward backward the discourse. then classify utterances as agreement or disagreement these represent various pragmatic influences previous current utterance. achieves 86.9% accuracy, 4.9% increase over work.
The occurrence of disfluencies in fully natural speech poses difficult challenges for spoken language understanding systems. For example, although self-repairs occur about 10% spontaneous utterances, they are often unmodeled recognition This is partly due to the fact that little known extent which cues signal may facilitate automatic repair processing. In this paper, acoustic and prosodic identified, based on an analysis a corpus taken from ARPA Air Travel Information System database,...
We propose a mapping between prosodic phenomena and semantico-pragmatic effects based upon the hypothesis that intonation conveys information about intentional as well attentional structure of discourse. In particular, we discuss how variations in pitch range choice accent tune can help to convey such as: discourse segmentation topic structure, appropriate referent, distinction 'given' 'new' information, conceptual contrast or parallelism mentioned items, subordination relationships...
In conversation, speakers become more like each other in various dimensions.This phenomenon, commonly called entrainment, coordination, or alignment, is widely believed to be crucial the success and naturalness of human interactions.We investigate entrainment four acoustic prosodic dimensions.We explore whether coordinate with these dimensions over conversation as a whole well on turn-by-turn basis both relative absolute terms, this coordination improves course conversation.
Cognitive theories of dialogue hold that entrainment, the automatic alignment between partners at many levels linguistic representation, is key to facilitating both production and comprehension in dialogue. In this paper we examine novel types entrainment two corpora---Switchboard Columbia Games corpus. We use high-frequency words (the most common corpus), its association with naturalness flow, as well task success. Our results show such predictive perceived dialogues significantly...