When: Thursday, March 01, 1:30 p.m.

Where: Mauldin Auditorium
1305 Newell-Simon Hall

Benjamin van Durme,
Assistant Research Professor
Department of Computer Science
Johns Hopkins University

Machine Learning Special Seminar

Abstract:
Simple Randomized Algorithms for Large-scale HLT

There has been a surge of interest in tackling problems of large scale language data with streaming and randomized algorithms. I will give an overview of two well known methods: Bloom filters and Locality Sensitive Hashing, and provide examples of how they've been used within HLT, such as for counting over large key sets in limited memory (as when constructing language models), in building distributional "semantic" similarity functions over the Google n-gram collection, and for streaming topic detection and tracking (TDT) on social media.

About the Speaker

Poster

<< Back