Dr. Douglas O'Shaughnessy

Professor, INRS-EMT

University of Quebec, Canada

 

Title:Progress and Challenges in Automatic Speech Recognition

Abstract: Speech coding has found great success in today’s widespread usage of cellphones. In addition, people are increasingly accustomed to hearing and accepting synthetic voices when they access information by phone. The third major application for speech, automatic speech recognition (ASR), is also seeing increased usage, but still has major limitations, falling short of what human listeners can do. This talk will examine modern ASR techniques, looking at why the methods were chosen, and their advantages and weaknesses. We will examine ways to extract relevant parameters from speech, while ignoring channel distortions and extraneous sounds that may also be present in ASR input. A brief history of ASR development will show the evolution of usage of Fourier analysis, linear prediction, cepstrum, and neural networks. We will discuss mel-frequency cepstral coefficients (MFCC), hidden Markov Models (HMM), language models and deep learning. A special attention will be paid to the differences among languages for ASR, e.g., usage of tones.

Biography: Douglas O'Shaughnessy is professor and program director at INRS-EMT (formerly, INRS-Telecommunications), a constituent of the University of Quebec, in Montreal, Canada. He is also an adjunct professor at McGill University in the Department of Electrical and Computer Engineering. He was educated at MIT (B.Sc. and M.Sc. in 1972; Ph.D. in 1976). His interests and research include automatic speech synthesis, analysis, coding, enhancement, and recognition. He has served as an Associate Editor for IEEE Transactions on Speech and Audio Processing, and the Journal of the Acoustical Society of America. He was the General Chair of the 2004 International Conference on Acoustics, Speech and Signal Processing (ICASSP) in Montreal, Canada. He is the author of the textbook Speech Communications: Human and Machine (2nd edition in 2000 by IEEE press). He is a Fellow of the Acoustical Society of America and of IEEE.

 

Title:Affect Associations: The Building Blocks of Sentiment Analysis

Abstract: Beyond literal meaning, words have associations with sentiment, emotion, colour, and even music. Identifying such associations is of substantial benefit for information visualization, data sonification, and sentiment analysis, which in turn have applications in commerce, education, art, and health. I will present methods that capture such associations accurately and reliably. I will show how these resources can be used for analyzing emotions in text, detecting stance from tweets, determining sentiment composition, and generating music from novels. Finally, I will show how we created the top-ranking systems in recent SemEval competitions on the sentiment analysis of tweets.

Biography: Dr. Saif Mohammad is Senior Research Officer at the National Research Council Canada (NRC). He received his Ph.D. in Computer Science from the University of Toronto. His research interests are in Computational Linguistics and Natural Language Processing, especially Lexical Semantics, Sentiment Analysis, Social Media Analytics, and Information Visualization. His team has developed a sentiment analysis system which ranked first in recent shared task competitions. He has organized several SemEval shared tasks, and is co-chair of SemEval for 2017 and 2018. His word-emotion association resource, the NRC Emotion Lexicon, is widely used for text analysis and information visualization. His work on detecting emotions in social media and on generating music from text have garnered widespread media attention, including articles in Time, SlashDot, LiveScience, io9, The Physics arXiv Blog, PC World, and Popular Science.

Title: Deception Detection: Case Studies on Fake Web Reviews, Illegal Advertisements, and Cyber Army’s Behaviours for Elect Campaign

Abstract: As online commerce and advertising keep growing, more and more consumers depend on information on the Internet to make purchasing decisions. This kind of information includes online advertisements posted by businesses, and discussions or reviews written by users. To affect customers’ buying decisions, fake opinions are generated for purpose to promote special targets and/or denounce their competitors. Besides, false, misleading or overstated advertisements also appear on the web. According to Weekly Food and Drug Safety, food, cosmetic and drug are three major sources of false advertisements. Moreover, misinformation happens in elect campaign. Camps employ various tools and methods to convince voters to vote their supporting candidates. Cyber army hired by unscrupulous camps aim at extolling specific candidates and criticizing their opponents. They play the similar roles of opinion spammers, and try to disseminate fake postings to mislead voters and influence the electoral process and outcome. In this talk, I will give three case studies on deception detection, including 2013 Samsung Taiwan’s event, TFDA illegal advertisement administration, and 2014 Taipei mayoral race. I will analyze the users’ behaviors, address linguistic and non-linguistic features, and present models to detect deception.

Biography: Hsin-Hsi Chen is a professor in Department of Computer Science and Information Engineering, National Taiwan University. His research interests are computational linguistics, Chinese language processing, information retrieval and extraction, and web mining. He served as President (2007-2009) and Advisory Board Chair (2009-2011) of Association for Computational Linguistics and Chinese Language Processing (ACLCLP), and Chair (2011-2013) of ACL SIGHAN. He was also Editorial Board member (2000-2008) of International Journal of Computational Linguistics and Chinese Language Processing, an Associate Editor (2007-2010) of ACM Transactions on Asian Language Information Processing, and an Editor of Journal of Information Science and Engineering (2010-). He was system demonstration co-chair of ACL-IJCNLP 2015, conference chair of IJCNLP 2013, program co-chair of ACM SIGIR 2010, program co-chair of ICADL 2012, served as steering committee members of AIRS, senior PC members of ACM SIGIR 2006, 2007, 2008 and 2009, area/track chairs of ACL 2012, ACL-IJCNLP 2009 and ACM CIKM 2008, and PC members of many conferences (IJCAI, SIGIR, AIRS, ACL, COLING, EMNLP, NAACL, EACL, IJCNLP, WWW, and so on). He won Google research awards in 2007 and 2012, awards of Microsoft Research Asia in 2008 and 2009, NTU EECS Academic Award in 2011, and NTU Award for Outstanding Service in 2011.