2017 Jelinek Summer Workshop on Speech and Language Technology (JSALT) at CMU
Watch Closing Day Presentations HERE.
Watch Plenary Lectures HERE.
Watch Opening Day Presentations HERE.
The 2017 Annual Jelinek Memorial Workshop on Speech and Language Technology will be held at Carnegie Mellon University Language Technologies Institute in Pittsburgh, PA.
It is a continuation of the Johns Hopkins University CLSP summer workshop series from 1995-2016. It consists of a two-week summer school, followed by a six-week workshop. Notable researchers and students come together to collaborate on selected research topics. The Workshop is named after the late Fred Jelinek, its former director and head of the Center for Speech and Language Processing.
2017 Workshop Program (July 3-August 11, 2017)
Topics and Research Teams:
- Neural Machine Translation with Minimal Parallel Resources (Leader: George Foster - National Research Council Canada)
- Enhancement and Analysis of Conversational Speech (Leader: Mark Liberman - Linguistic Data Consortium)
- The Speaking Rosetta Stone - Discovering Grounded Linguistic Units for Languages without Orthography (Leader: Emmanuel Dupoux - École des Hautes Etudes en Sciences Sociales, Odette Scharenborg - Radboud University Nijmegen)
Closing Presentations - August 10-11, 2017
Gates and Hillman Centers, Rashid Auditorium (4401)
Thursday, August 10th, 2017
Workshop Review and Summary Presentations, Part I
08:30 AM – Continental Breakfast (Lobby outside Rashid Auditorium)
09:00 AM – Opening Remarks, Florian Metze, Alan Black and Sanjeev Khudanpur
09:15 AM – Team Presentation: The Speaking Rosetta Stone - Discovering Grounded Linguistic Units for Languages without Orthography
10:45 AM – Break
11:00 AM – Team Presentation Continued
11:45 AM – Discussion, Questions and Comments
12:15 PM – Lunch Break (on your own)
01:30 PM – Team Presentation: Enhancement and Analysis of Conversational Speech
03:00 PM – Break
03:15 PM – Team Presentation Continued
04:00 PM – Discussion, Questions and Comments
04:30 PM – Adjourn for the day
Friday, August 11th, 2017
Workshop Review and Summary Presentations, Part II
08:30 AM – Continental Breakfast (Lobby outside Rashid Auditorium)
09:00 AM – Team Presentation: Neural Machine Translation with Minimal Parallel Resources
10:30 AM – Break
10:45 AM – Team Presentation Continued
11:30 AM – Discussion, Questions and Comments
12:00 PM – Concluding Remarks, Florian Metze, Alan Black and Sanjeev Khudanpur
12:15 PM – Lunch Break (on your own)
01:45 PM – Teams return to workshop laboratory to wrap-up
6:00 – 9:00 PM – WS17 Farewell Dinner - Schenley Park Visitor Center (https://www.pittsburghparks.org/schenley-park-cafe-visitor-center). Dress code: business casual.
EVERYONE Please RSVP (Presentation ONLY or ALSO dinner on 11th) by August 7 to Jae Cho
Summer School (June 19-30, 2017)
The summer school is meant to be an introduction to the state-of-the-art research in the speech and language technology area for graduate and undergraduate students. It also contains an introduction to this year's workshop research topics.
- View the last year's agenda and topics here.
- Click here for a description of the undergraduate program and the opportunity to apply. (Now closed)
- Apply here for NAACL-supported undergraduate spot for the summer school (Apply by Monday, May 15 2017.)
Opening Day Presentations - July 3, 2017
For a map of the CMU campus, please see the Participant Information Section at the bottom of this page.
- 08:30 AM – Morning Sessions, Gates and Hillman Centers, Rashid Auditorium (4401)
- 12:15 PM – Adjourn to 407 S Craig St, Workshop Laboratory Setup, 407 S Craig St., Room 104
- 12:30 PM – Lunch on your own
- 02:00 PM – Team Meetings (in respective laboratories)
- 05:30 PM – Opening Day Picnic, Gates and Hillman Centers, 6th floor lounge and patio
Recordings of the Opening Day Presentations can be viewed here.
Plenary Lectures - Wednesdays, July 5-August 2
RASHID AUDITORIUM (GATES AND HILLMAN CENTERS 4401)
- 10:30AM-12PM Plenary Lectures by Invited Speakers
Speakers:
July 5 Brian MacWhinney - "Language Learning Theory"
July 12 Luke Zettlemoyer (TO BE HELD IN Wean Hall 7500)
July 19 Sharon Goldwater
July 26 Jason Pelecanos (IBM) - "A History of Speaker Recognition: Of Hammers and Nails"
In this talk we provide a selective historical overview of the progress made in text-independent automatic speaker recognition. Specifically, we discuss a recurring trend where new developments build upon existing speaker recognition innovations. Many of these innovations are also adopted from other fields.
This principle is also known as the “Law of the Instrument”. In its more well-known form, it may be said that “given only a hammer, it is tempting to treat everything as a nail.” When a new research approach is developed (a hammer) this tool is shared across research areas to address many longstanding problems (the nails). One may need only examine the recent developments in deep learning to realize that these developments are permeating many research areas, including speaker recognition.
More than a decade ago, NIST speaker recognition evaluations reported Equal Error Rates (EER) of approximately 10%. With these accumulated advances, systems today are giving an error rate of 0.6% (NIST 2010 Speaker Recognition Evaluation data). While there are many challenges in speaker recognition remaining, what can be expected in the next decade and what will be the role of the “Law of the Instrument”?
August 2 Emily Mower Provost - "Human-Centered Computing: Using Speech to Understand Behavior"
Engineering approaches to human behavior analysis are complicated by the lack of a one-to-one mapping between the behavioral cues that an individual generates and how an external observer interprets those cues. This many-to-many mapping injects noise into both the data and ground truth. As a result, many of the models and assumptions used in traditional machine learning and signal processing must be used with caveats or adapted to meet the needs of this domain. I will discuss our work on algorithmic approaches to characterize and predict how humans perceive signals that modulate spoken communication, focusing on emotion and mood. I will highlight our efforts in tracking mood for individuals with bipolar disorder. These technologies have the potential to forward diagnosis and treatment by providing constrained, repeatable, and easily modifiable assessment protocols, objective measures, and interaction scenarios.
- 12:00-1:00PM Catered Lunch (Please RSVP, if not team members)
- 1:00-2:30PM Weekly Progress Meetings
Participant Information
- Accommodations
Hampton Inn Pittsburgh-University Center (1.13 Miles)
Shadyside Inn All Suites (0.70 Miles)
Residence Inn Pittsburgh University/Medical Center (1.10 Miles)
Wyndham Pittsburgh University Center (0.74 Miles)
Hilton Garden Inn Pittsburgh University Place (0.90 Miles)
- Dining (YELP)
- Transportation
- Local Attractions
Heinz Field (Football)
PNC Park (Baseball)
PPG Paints Arena (Hockey)
Benedum Center
Heinz Hall
Duquesne Incline
Carnegie Museums (Art, Natural History, Science Center, The Andy Warhol)
Frank Lloyd Wright's Fallingwater
JustDuckyTours
We thank our sponsors
Coming soon
Organizing Committee
Sanjeev Khudanpur
Florian Metze
Alan Black
Local Contacts
Jae Cho
412-268-5480