Successfully reported this slideshow.
We use your LinkedIn profile and activity data to personalize ads and to show you more relevant ads. You can change your ad preferences anytime.
2017
The 55th Annual Meeting of the

Association for Computational Linguistics ║
Vancouver, Canada ║ July 30 – August 4 ║
...
ACL is family friendly
Professionally managed on-site childcare is available for kids
aged 6 months to 12 years old. Drop-...
Recruitment lunch
Today from noon until 2pm there is a recruitment
lunch for job seekers. It is limited to 400 participant...
Conference App
This year have a smartphone app for the
conference. First download the GuideBook app,
then download the “AC...
Social Media
The conference social media channels are:
Facebook: @acl2017
Twitter: @acl2017
Instagram: @aclnlp
Please incl...
Live Captioning
Live captions are available for the conference
(with translations available if you want). Thanks
Microsoft...
Thank you to our sponsors!ACL 2017
║We are grateful to our ACL 2017 Sponsors for their support!
Platinum
Gold
Gold
Silver
Bronze
Supporter
Organizing Team
Program Co-Chairs Regina Barzilay and Min-Yen Kan 

Local Organizing Chairs Priscilla Rasmussen and Anoop ...
Aurélie Névéol
Karin Verspoor Roger Levy Anders Søgaard Ron Artstein
Raquel Fernandez Oliver Lemon Yangfeng Ji Sujian Li B...
Statistics
1318 Submitted Papers

302 Accepted Papers

Overall 22% Acceptance Rate (Main Conference)

367 Presentations
Statistics
1318 Submitted Papers

302 Accepted Papers

Overall 22% Acceptance Rate (Main Conference)

367 Presentations

…...
Improving Paper Selection Process
ACL Survey:
I was really disappointed by the reviewing quality of ACL
sometimes. Many re...
• 10% of paper were rejected due to formatting violations
• In most cases, violations were honest mistakes
• Disproportion...
• Challenge: large number of reviewers, many unfamiliar to ACs
• Solution tried: Toronto Paper Matching System (TPMS)
• Mi...
• Hierarchical Area Organization, with fewer papers per AC
• Short review period vs long discussion period
• Run scripts t...
Impact of Rebuttal/Discussion
• Approach: compare ranking obtained from the original reviews and
accepted papers
• Finding: 78% of papers ranked based o...
• Approach: train a classifier to predict review sentiment (two
classes, with 3s excluded)
• Accuracy: F-measure is 79%
Ra...
• Approach: train a classifier to predict review sentiment (two
classes, with 3s excluded)
• Accuracy: F-measure is 79%
Ra...
• Approach: train a classifier to predict review sentiment (two
classes, with 3s excluded)
• Accuracy: F-measure is 81%
Ra...
ACL 1997
ACL 2017 Opening Session
ACL 2017 Opening Session
ACL 2017 Opening Session
Upcoming SlideShare
Loading in …5
×

ACL 2017 Opening Session

2,843 views

Published on

Delivered by Chris Callison Burch (University of Pennsylvania), Regina Barzilay (Massachusetts Institute of Technology) and Min-Yen Kan (National University of Singapore). 31 Aug 2017. Vancouver, Canada

Published in: Education
  • Be the first to comment

  • Be the first to like this

ACL 2017 Opening Session

  1. 1. 2017 The 55th Annual Meeting of the
 Association for Computational Linguistics ║ Vancouver, Canada ║ July 30 – August 4 ║ Welcome General Chair: Chris Callison-Burch Program Chairs: Regina Barzilay and Min-Yen Ka ACL Vice President: Marti Hearst ACL President: Joakim Nivre
  2. 2. ACL is family friendly Professionally managed on-site childcare is available for kids aged 6 months to 12 years old. Drop-ins are welcome. Stop by the Oak Rooms or call Lisa on 1–450–466–6897. The Arbutus Room is also available for nursing mothers. Your registration includes full dinners at the poster and demo sessions, and breakfasts and coffee breaks during the conference. You are invited to bring your loved ones by purchasing additional meals at the registration desk. The social event will be held at the Vancouver Aquarium tomorrow evening after the poster session. Your ticket is included. You may purchase additional tickets for family members.
  3. 3. Recruitment lunch Today from noon until 2pm there is a recruitment lunch for job seekers. It is limited to 400 participants, and it was paid for by 20 companies. Attending this event requires an admission ticket (which is located in your registration packet). If you did not get an admission ticket, I encourage you to visit our many sponsors who are exhibiting at ACL 2017 throughout the main conference where you can speak to them about employment possibilities.
  4. 4. Conference App This year have a smartphone app for the conference. First download the GuideBook app, then download the “ACL 2017” guide from within that app. https://guidebook.com/g/acl2017nlp/ Wifi Password: ACL2017
  5. 5. Social Media The conference social media channels are: Facebook: @acl2017 Twitter: @acl2017 Instagram: @aclnlp Please include hashtag #acl2017nlp in any social media posts.
  6. 6. Live Captioning Live captions are available for the conference (with translations available if you want). Thanks Microsoft! To access the captions, go:                http://www.aka.ms/2ACL    If you’re a speaker, please use the lapel mic. If you’re an audience member please be sure to use the handheld mic when asking your question.
  7. 7. Thank you to our sponsors!ACL 2017 ║We are grateful to our ACL 2017 Sponsors for their support! Platinum Gold
  8. 8. Gold Silver Bronze Supporter
  9. 9. Organizing Team Program Co-Chairs Regina Barzilay and Min-Yen Kan Local Organizing Chairs Priscilla Rasmussen and Anoop Sarkar Workshop Chairs Wei Xu and Jonathan Berant Tutorial Chairs Maja Popović and Jordan Boyd-Graber Publication Chairs Wei Lu, Sameer Singh, and Margaret Mitchell Demonstration Chairs Heng Ji and Mohit Bansal Student Research Workshop Organizers Spandana Gella, Allyson Ettinger, and Matthieu Labeau Faculty Advisors SRW Cecilia Ovesdotter Alm, Mark Dredze and Marine Carpuat Publicity Chair Charley Chan Conference Handbook Chair Christian Federmann Student Volunteer Coordinator Maryam Siahbani Webmaster & Appmaster Nitin Madnani
  10. 10. Aurélie Névéol Karin Verspoor Roger Levy Anders Søgaard Ron Artstein Raquel Fernandez Oliver Lemon Yangfeng Ji Sujian Li Bo nnie Webber Eugene Agichtein Chia-Hui Chang Jing Jiang Sarvnaz Karimi Zornitsa Kozareva Kang Liu Tie-Yan Liu Mausam Alessandro Moschitti Smaranda Muresan Grzegorz Chrupała Amir Globerson Tommi Jaakkola Sujith Ravi Wil liam Yang Wang Yang Liu Minh-T hang Luong Hai tao Mi Graham N eubig Deyi Xio ng Michael Piot rowski Karën F ort Omri Abend Mona Diab Jaso n Eisner Hinric h Schütze Soph ie Rosset Wajdi Zaghouani Manaal Faruqui Hannaneh Hajishirzi Anna Korhonen Presla v Nakov Mehroosh Sadrzadeh Aline Villavicencio Alexandra Balahur Lun-Wei Ku Saif M Mohammad Zhiyuan Liu Shimei Pan Svitlana Volkov a Chiori Hori Chia-ying Lee Wenjie Li Alexander M Rush Verena Rie ser Emily Pitler Barbara Plank Yue Zhang Hai Zhao Mohit Bansal Na te Kushman Dan Gildea Rebecca Passonneau Tim Baldwin Martha Palme r Mark Johnson Zhang Min Kristina Toutanova Lu Wei Julia Hockenma
  11. 11. Statistics 1318 Submitted Papers 302 Accepted Papers Overall 22% Acceptance Rate (Main Conference) 367 Presentations
  12. 12. Statistics 1318 Submitted Papers 302 Accepted Papers Overall 22% Acceptance Rate (Main Conference) 367 Presentations … 36 Blog Posts
  13. 13. Improving Paper Selection Process ACL Survey: I was really disappointed by the reviewing quality of ACL sometimes. Many reviewers are actually not qualified to review papers and many did not spend enough time on the papers. A substantial number of respondents express concern about the current conference reviewing process with increasing reviewer load and decreasing review quality. 16 I believe the lack of good reviewers is a much more severe problem that may ruin the standard of ACL in the near future. We reject a lot of good papers, and let in a lot of bad ones
  14. 14. • 10% of paper were rejected due to formatting violations • In most cases, violations were honest mistakes • Disproportionally impacts newcomers from developing communities Rejection Without Review length, margins, font, anonymization, … Solution: Use on-line format checker as part of submission
  15. 15. • Challenge: large number of reviewers, many unfamiliar to ACs • Solution tried: Toronto Paper Matching System (TPMS) • Mixed outcomes due to limited TPMS enrollment (50%) Improving Review Quality: Matching Papers and Reviews Proposed Solution: Create an up-to-date reviewer database, keeping institutional memories across conferences Dear Regina, I just want to bring to your attention our situation that might bring to light lack of appropriate reviewers for papers….
  16. 16. • Hierarchical Area Organization, with fewer papers per AC • Short review period vs long discussion period • Run scripts to flag problematic reviews • Enable direct communication between authors and ACs Improving Review Quality: Towards Journal-like Reviewing Proposed Solution: Create more effective tools for ACs to identify problematic reviews and calibrate individual reviewers.
  17. 17. Impact of Rebuttal/Discussion
  18. 18. • Approach: compare ranking obtained from the original reviews and accepted papers • Finding: 78% of papers ranked based on the original reviews are accepted • Average score of upgraded papers — 3.14, downgraded — 3.4 Impact of Discussion Period Conclusions: 1. Original reviews drive the acceptance process. 2. Discussion does determine the fate of 22% of submissions Place to reduce unnecessary load, focusing on challenging papers
  19. 19. • Approach: train a classifier to predict review sentiment (two classes, with 3s excluded) • Accuracy: F-measure is 79% Rationales behind Decisions This is nice ongoing work that attempts to explore understanding of XXX. The focus on the role of XXX is well motivated, and the treatment of the various ways in which YYY is comprehensive and methodical ….
  20. 20. • Approach: train a classifier to predict review sentiment (two classes, with 3s excluded) • Accuracy: F-measure is 79% Rationales behind Decisions unclear, hard, lacks novelty, enough, baseline, writing, compare, missing, novelty, description, well written, good, solid, well, nice, liked, interesting, useful, paper well, novel, evaluation
  21. 21. • Approach: train a classifier to predict review sentiment (two classes, with 3s excluded) • Accuracy: F-measure is 81% Rationales behind Decisions unclear, hard, lacks novelty, enough, baseline, writing, compare, missing, description, well written, good, solid, well, nice, liked, interesting, useful, paper well, novel, evaluation
  22. 22. ACL 1997

×