We present the IBM systems for the Rich Transcription 2007 (RT07) speaker diarization evaluation task on lecture meeting data. We first overview our baseline system that was devel...
Traditional wisdom holds that once documents are turned into bag-of-words (unigram count) vectors, word orders are completely lost. We introduce an approach that, perhaps surprisi...
Xiaojin Zhu, Andrew B. Goldberg, Michael Rabbat, R...
Abstract. Delay (or disruption) tolerant sensor networks may be modeled as Markovian evolving graphs [1]. We present experimental evidence showing that considering multiple (possib...
Michael Keane, Evangelos Kranakis, Danny Krizanc, ...
Abstract. We show several PAC-style concentration bounds for learning unigrams language model. One interesting quantity is the probability of all words appearing exactly k times in...
Even with todays hardware improvements, performance problems are still common in many software systems. An approach to tackle this problem for component-based software architectur...