際際滷

際際滷Share a Scribd company logo

Wei Xu - Innovative Applications of AI Panel

?
?
R
Rehgan Avon

2018 Women in Analytics Conference https://www.womeninanalytics.org/ The goal of this panel is to discuss where individual industries or areas of focus has been successful and unique in utilizing AI as a technology, if/where it is struggling with the adoption and why, and what complexities exist around fully utilizing AI as a solution.

1 of 17
Download to read offline
How AI understands language Wei Xu Department of Computer Science and Engineering March-15-2018 @ Women in Analytics Conference
The Turing Test Alan Turing (1912-1954) mathematician, world-class?Marathon?runner, forced to chemical castration for homosexuality conviction. . , , . .
Natural language processing (NLP) 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 865 770 749 660 860 863 930 894 880 823 871 845 732 887 931 887 1,190 1,104 1,620 1,597 1,315 1,583 1,641 1,562 2,123 2,104 1,602 2,151 2,136 2,028 2,064 2,093 1,990 2,536 2,901 2,949 Number of Members Association for Computational Linguistics is an organization of world leading Natural Language Processing researchers since 1962. The goal is to enable computers to understand and generate human language. USA Worldwide
The Holy Grail of AI / NLP Human language is ambiguous, creative, in?nite, and ever evolving.
The Holy Grail of AI / NLP Human language is ambiguous, creative, in?nite, and ever evolving. . . . . Photo Credit: Kristy Wigglesworth / AP
The Holy Grail of AI / NLP word ＾sel?e￣ in Google Trends Human language is ambiguous, creative, in?nite, and ever evolving. Images from http://www.ufunk.net/en/gadgets/sel?e-arm/
The Holy Grail of AI / NLP . word , phrase . . sentence. . Solution: learning large-scale paraphrases Human language is ambiguous, creative, in?nite, and ever evolving. Wei Xu. ＾Data-driven Approaches for Paraphrasing Across Language Variations￣ PhD Thesis (2014)
Natural language understanding match . . , . , Question Answering ., . . , . . , . . .
Natural language generation Text Simpli?cation (adapting machine translation techniques) Wei Xu, Courtney Napoles, Ellie Pavlick, Chris Callison-Burch. ＾Optimizing Statistical Machine Translation for Simpli?cation￣ in TACL (2016)
Designing Various [Machine learning] models ? Multi-instance Learning Models [Xu et al. 2014; Tabassum et al. 2016] Wei Xu, Alan Ritter, Chris Callison-Burch, Bill Dolan, Yangfeng Ji ＾Extracting Lexically Divergent Paraphrases from Twitter￣ in TACL (2014) (Viterbi approximation + online learning)
Designing Various [Machine learning] models Wuwei Lan, Wei Xu. ＾The Important Role of Subword Embeddings in Sentence Pair Modeling￣ in NAACL (2018) ? Deep Pairwise Neural Networks [He et al. 2015; Lan et al. 2017; Lan and Xu 2018] Bi-LSTM Decompose sentence input into word context to reduce modeling dif?culty GloVe or subword CNN ? Highway Network
Multiple vector similarity measurements used to capture word pair relationship Designing Various [Machine learning] models ? Deep Pairwise Neural Networks [He et al. 2015; Lan et al. 2017; Lan and Xu 2018] Wuwei Lan, Siyu Qiu, Hua He, Wei Xu ＾A Continuously Growing Dataset of Sentential Paraphrases￣ in EMNLP (2017)
Designing Various [Machine learning] models Hua He and Jimmy Lin. ＾Pairwise Word Interaction Modeling with Deep Neural Networks for Semantic Similarity Measurement￣ in NAACL (2016) ? Deep Pairwise Neural Networks [He et al. 2015; Lan et al. 2017; Lan and Xu 2018] More attention added to top ranked word pairs.
Designing Various [Machine learning] models ? Deep Pairwise Neural Networks [He et al. 2015; Lan et al. 2017; Lan and Xu 2018] Sentence pair relationship identi?ed by pattern recognition through ConvNet. Hua He and Jimmy Lin. ＾Pairwise Word Interaction Modeling with Deep Neural Networks for Semantic Similarity Measurement￣ in NAACL (2016)
[data science] for linguistic styles she says he says wonderfully delightfully beautifully ?ne well good nicely superbly Daniel Preotiuc, Wei Xu, Lyle Ungar. ＾Discovering User Attribute Stylistic Differences via Paraphrasing￣ in AAAI (2016) (also studied age & income)
Machine reading of instructions Chaitanya Kulkarni, Wei Xu, Alan Ritter, Raghu Machiraju. ＾An Annotated Corpus for Machine Reading of Instructions in Wet Lab Protocols￣ in NAACL (2018) Wet Lab Protocols: (Photo: Moley Robotic Chef)
http://web.cse.ohio-state.edu/~weixu/ Thank you Assistant ProfessorWei Xu Department of Computer Science and Engineering Ohio State University

More Related Content

Wei Xu - Innovative Applications of AI Panel

  • 1. How AI understands language Wei Xu Department of Computer Science and Engineering March-15-2018 @ Women in Analytics Conference
  • 2. The Turing Test Alan Turing (1912-1954) mathematician, world-class?Marathon?runner, forced to chemical castration for homosexuality conviction. . , , . .
  • 3. Natural language processing (NLP) 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 865 770 749 660 860 863 930 894 880 823 871 845 732 887 931 887 1,190 1,104 1,620 1,597 1,315 1,583 1,641 1,562 2,123 2,104 1,602 2,151 2,136 2,028 2,064 2,093 1,990 2,536 2,901 2,949 Number of Members Association for Computational Linguistics is an organization of world leading Natural Language Processing researchers since 1962. The goal is to enable computers to understand and generate human language. USA Worldwide
  • 4. The Holy Grail of AI / NLP Human language is ambiguous, creative, in?nite, and ever evolving.
  • 5. The Holy Grail of AI / NLP Human language is ambiguous, creative, in?nite, and ever evolving. . . . . Photo Credit: Kristy Wigglesworth / AP
  • 6. The Holy Grail of AI / NLP word ＾sel?e￣ in Google Trends Human language is ambiguous, creative, in?nite, and ever evolving. Images from http://www.ufunk.net/en/gadgets/sel?e-arm/
  • 7. The Holy Grail of AI / NLP . word , phrase . . sentence. . Solution: learning large-scale paraphrases Human language is ambiguous, creative, in?nite, and ever evolving. Wei Xu. ＾Data-driven Approaches for Paraphrasing Across Language Variations￣ PhD Thesis (2014)
  • 8. Natural language understanding match . . , . , Question Answering ., . . , . . , . . .
  • 9. Natural language generation Text Simpli?cation (adapting machine translation techniques) Wei Xu, Courtney Napoles, Ellie Pavlick, Chris Callison-Burch. ＾Optimizing Statistical Machine Translation for Simpli?cation￣ in TACL (2016)
  • 10. Designing Various [Machine learning] models ? Multi-instance Learning Models [Xu et al. 2014; Tabassum et al. 2016] Wei Xu, Alan Ritter, Chris Callison-Burch, Bill Dolan, Yangfeng Ji ＾Extracting Lexically Divergent Paraphrases from Twitter￣ in TACL (2014) (Viterbi approximation + online learning)
  • 11. Designing Various [Machine learning] models Wuwei Lan, Wei Xu. ＾The Important Role of Subword Embeddings in Sentence Pair Modeling￣ in NAACL (2018) ? Deep Pairwise Neural Networks [He et al. 2015; Lan et al. 2017; Lan and Xu 2018] Bi-LSTM Decompose sentence input into word context to reduce modeling dif?culty GloVe or subword CNN ? Highway Network
  • 12. Multiple vector similarity measurements used to capture word pair relationship Designing Various [Machine learning] models ? Deep Pairwise Neural Networks [He et al. 2015; Lan et al. 2017; Lan and Xu 2018] Wuwei Lan, Siyu Qiu, Hua He, Wei Xu ＾A Continuously Growing Dataset of Sentential Paraphrases￣ in EMNLP (2017)
  • 13. Designing Various [Machine learning] models Hua He and Jimmy Lin. ＾Pairwise Word Interaction Modeling with Deep Neural Networks for Semantic Similarity Measurement￣ in NAACL (2016) ? Deep Pairwise Neural Networks [He et al. 2015; Lan et al. 2017; Lan and Xu 2018] More attention added to top ranked word pairs.
  • 14. Designing Various [Machine learning] models ? Deep Pairwise Neural Networks [He et al. 2015; Lan et al. 2017; Lan and Xu 2018] Sentence pair relationship identi?ed by pattern recognition through ConvNet. Hua He and Jimmy Lin. ＾Pairwise Word Interaction Modeling with Deep Neural Networks for Semantic Similarity Measurement￣ in NAACL (2016)
  • 15. [data science] for linguistic styles she says he says wonderfully delightfully beautifully ?ne well good nicely superbly Daniel Preotiuc, Wei Xu, Lyle Ungar. ＾Discovering User Attribute Stylistic Differences via Paraphrasing￣ in AAAI (2016) (also studied age & income)
  • 16. Machine reading of instructions Chaitanya Kulkarni, Wei Xu, Alan Ritter, Raghu Machiraju. ＾An Annotated Corpus for Machine Reading of Instructions in Wet Lab Protocols￣ in NAACL (2018) Wet Lab Protocols: (Photo: Moley Robotic Chef)
  • 17. http://web.cse.ohio-state.edu/~weixu/ Thank you Assistant ProfessorWei Xu Department of Computer Science and Engineering Ohio State University