Muhammad Abdul-Mageed will present a research seminar titled ‘Computational Linguistics in the UBC Deep Learning & NLP Lab’.
Click to join Zoom meeting. The passcode is 635410. (The meeting ID is 647 6334 5823.)
Title: Computational Linguistics in the UBC Deep Learning & NLP Lab
Abstract: Can we find a needle in a haystack? Probably not. Can we investigate human language use in every corner of the world, compare across individuals and micro-communities, and make discoveries about human experience vis-a-vis language? Two convergent factors make this otherwise despairingly ambitious goal seem more plausible than ever: Massive-scale, geo-tagged language data and modern deep neural networks. In this talk, I will describe ongoing work in the UBC Deep Learning and NLP Lab chasing knowledge about human language and new methods for uncovering such a knowledge. In particular, I will overview a number of recent neural network innovations and their applications in a host of contexts. This includes location-aware NLP with micro-dialects at scale [1, 2], multilingual machine translation [3, 4, 5], and computational socio-pragmatics [6, 7]. I then explore generalizations and extensions in a billion-scale dataset of 100+ languages in the context of COVID-19 [8, 9].
[1] https://arxiv.org/pdf/2010.04900.pdf
[2] https://arxiv.org/pdf/2010.11334.pdf
[3] https://www.aclweb.org/anthology/2020.ngt-1.20.pdf
[4] https://www.aclweb.org/anthology/2020.sigmorphon-1.pdf#page=162
[5] https://www.aclweb.org/anthology/W19-5431.pdf
[5] https://www.aclweb.org/anthology/2020.osact-1.3.pdf
[6] https://www.aclweb.org/anthology/2020.osact-1.6.pdf