Topic-SIG Annual Meeting : 2012

The Topic Related SIGs held their yearly face to face meeting on Wednesday, September 12, 2012 during Interspeech in Portland.

Attendees (Existing SIGS):

AVISA: Chris Davis (University of Western Sydney, Australia)
SIGDIAL: Jason Williams (Microsoft)
SIGML: Joseph Keshet, Karen Livescu (both at Toyota Technological Institute at Chicago, USA)
SLATE: Martin Russell (University of Birmingham, UK)
SpLC: Joe Campbell (MIT Lincoln Laboratory, USA)
SPROSIG: Keikichi Hirose (University of Tokyo, Japan), Mark Hasegawa-Johnson (University of Illinois, USA)
SYNSIG: Martti Vainio (University of Helsinki, Finland)


SLIM: Guillaume Gravier (Institut de Recherche en Informatique et Systèmes Aléatoires (IRISA), France), Gareth Jones (Dublin City University, Ireland)
SIG-HIST: John Ohala (University of California, Berkeley, USA)
SIG-ROBUST: Ramon Fernandez Astudillo (Instituto de Engenharia de Sistemas e Computadores (INESC), Portugal), Emmanuel Vincent (Institut national de recherche en informatique et en automatique (INRIA), France)
SIG-CHILD: Kay Berkling (Karlsruhe University of Education, Germany)

We started the meeting by introducing participants who have/are starting new topic related SIGs (see above for attendees):

SLIM: Speech and Language Indexing for Multimedia (just chartered)
SIG-HIST: The History of Speech Communication Sciences (just chartered)
SIG-ROBUST: Robust Speech Processing (in formation)
SIG-CHILD: Child Computer Interaction (was being formed during Interspeech, now chartered)
SIG-SLPAT: Speech and Language Processing for Assistive Technologies (now chartered)

Review Process - SIG Perspective.
We then reviewed the need for SIGs to take a more active part in vetting and providing reviewers for Interspeech and related workshops. The SIGs provide centers of expertise in various areas and having SIGs suggest potential qualified reviewers is essential to providing conferences and workshops of quality to ISCA members. Michael specifically asked the SIGs to provide the ISCA board with candidates from their individual communities. We also had a more general discussion of the review process. Joe Campbell suggested a more formalized method of tracking reviewers. He also asked if a data base of reviewers could be provided to workshop organizers to make the workshop review process simpler. For Odyssey (the flagship workshop of SpLC) five people are actually assigned for each review. Assigments are all done via spreadsheet. Reviewer performance is recorded over time and reviewers assigned to minimize the number of low-performing reviewers per paper. Karen Livescu suggested a multi-layer review process over a longer period of time. This is what ICML uses. Reviewers commit way in advance and better reviews result. Joseph Keshet asked if there are formal ISCA guidelines (the answer is "no" - something for the board to discuss (MAP)). A suggestion was also made to make papers on line as soon as accepted; was noted.

Requests to SIGS
Michael then made a plea for SIGs to also nominate candidates for ISCA Fellow and Distinguished Lecturer. Note Fellow nominations are due February 10th, with an initial indication of interest in nominating someone by January 10th.

Another plea was also made to organize papers from workshops into journal (SPECOM, CSL) special issues. Joe suggested that each workshop should have a best paper as well than could be publicized.

SIG Workshop Experiences
We then had a general discussion of ways to make it as easy as possible to organize SIG workshops. Here are some "first person" experiences

SIGML: It was hard to fund the workshop making bringing in external speakers difficult. ISCA funding helped once but next year wont be possible. Second workshops currently not eligible for ISCA funding (except for travel grants). It would be helpful to be able to leverage the payment mechanism of the broader conference when the workshop is associated with the broader conference ACL and ICML do this. Note also that other conferences separate review process and software from rest of conference. Paper management can be separated from financial mechanism. It would be useful to have the software detect when the author has to be in two different sessions. Maybe also highlight the specific presenters?

Jason suggested having workshops at academic institution, but not at the same venue, to save money. There should be a mechanism for jointly advertising conference and workshops. ISCA's current mechanism is very different from other conferences that build workshops right into the conference organization.

Slate: Cost is an issue. Can we find some compromise here w/o creating more burdens for the organizers?

Kay: 2008, 9 we co-hosted WOCCI with ICMI. They took care of all the finance and registration part. We took care of all the reviews. Papers were submitted and reviewed via, which is a great tool for small conferences and works wonderful, puts together proceedings and multi-sessions to subdivide responsibilities of reviews. 2012 was easy because we charged participants 20 Euro for food only. Venu was free at OHSU. Food could be ordered 2 days in advance so that late registration was also no problem. We charged registration by using Google Checkout button on our website. You can see our website at

Emmanuel commented that he is involved in the organization of several challenges (such as CHiME).The deadline for special session paper submissions is the
same as that of the main conference. If enough papers are not submitted the special session may not be held. This causes difficulties for challenge participants so instead a separate workshop is held, often creating some financial risks. Alternatively, if a special session deadline/organization could be decided later, a separate workshop is not necessary. Asking for a few extra weeks for submission are not a big deal since the reviewers for such sessions usually understand the constraints and have always been happy to speed up the review process.

Activity Reports

Blizzard Challenge with a workshop taking place in conjunction with Interspeech (the following Friday).