dirk hovy

2015/07/30, 07:21PM
NEWS: Our paper “The Rating Game: Sentiment Rating Reproducibility from Text”, written with two of our students, was accepted at EMNLP! We question how reproducible ratings are from text, for both humans and statistical models (spoiler: not very).

2015/07/30, 07:17PM
NEWS: Our paper “Personality Traits on Twitter—or—How to Get 1500 Personality Tests in a Week” was accepted at WASSA.


“We don’t make mistakes here — we have only happy accidents.”
Bob Ross

home

Hi, glad you found this site! I work in Natural Language Processing (NLP), a subfield of artificial intelligence. My work lies at the intersection of NLP and Machine Learning. I am interested in enhancing our understanding of language with respect to lexical semantics (modeling what words mean in context), non-standard language, and co-text (the influence of extra-linguistic factors on language use). I am currently a postdoc at CST at the University of Copenhagen.
I received my PhD from the University of Southern California in Los Angeles, where I was working as research assistant at the Information Sciences Institute. My thesis topic was how to learn semantic types and relations from plain text.
You will find here

If you have questions or comments, feel free to send me a mail.
Cheers,

Dirk
2015/07/30, 07:21PM

NEWS: EMNLP Paper

Our paper “The Rating Game: Sentiment Rating Reproducibility from Text”, written with two of our students, was accepted at EMNLP! We question how reproducible ratings are from text, for both humans and statistical models (spoiler: not very).
2015/07/30, 07:17PM

NEWS: WASSA Paper

Our paper “Personality Traits on Twitter—or—How to Get 1500 Personality Tests in a Week” was accepted at WASSA.
2015/07/10, 04:07AM

BLOG: Are our models ageist?

We don’t expect 60-year-olds to speak like teenager―but we taught our models to. As a result, they perform worse on analyzing data from different demographic groups. As NLP is used more widely, this can have serious repercussions. However, we show that we can a) include demographics to learn better models, and b) use NLP to discover new facts about language.
more…

Möchten Sie diese Website lieber auf deutsch lesen?

Valid XHTML 1.0 Transitional Valid CSS!