Welcome Message by the Technical Chairs
To truly appreciate the amount of work that goes into managing and creating the technical program of a conference the size of Interspeech, one must experience it firsthand. We would like to give you some insights into its complexity, not to scare away those who will be technical program chairs in the future, but to actually help them plan for their work. Even though the process was arduous, we enjoyed every minute of it. It allowed us to establish an exciting team spirit among the chairs and the Area Coordinators and create a virtual contact between us and almost every participant to the conference. And more than that, it was a unique human and learning experience which we will always remember.
The first phase of creating a technical program is the enrollment of the reviewers and the Area Coordinators. This year, 864 colleagues (out of an initial pool of 1333 we initially contacted) kindly accepted offers to perform reviews. Of those, 115 volunteered to be in a special task force known as “crash reviewers”, whose job it is to perform emergency reviews on demand, if necessary, at the end of the review period and with a short notice. This year we chose to distribute the 19 scientific areas among 38 Area Coordinators, two for each area, with the specific goal of distributing the work among local and international scientists. A call for special sessions was also sent, and a committee including the general and technical program chairs and the chairs for special sessions indicated which ones were to form the program of the conference. Similarly, with the help of the tutorial chairs and of the keynote chairs we selected the best proposals for tutorials and keynotes.
The main bulk of the work begins towards the end of the submission phase when most of the authors start submitting their work. Most of the submissions do not require help from the technical program chairs, but some do. The help may take the form of accommodations for delay due to serious reasons—like our Japanese colleagues hit by one of the most catastrophic events of our times—or due to malfunctioning of the online system or human errors. Once all papers were in the system, we had to match each one of them with at least three reviewers. So, for 1439 paper we would need 4317 reviews, making sure that the load is evenly distributed across all reviewers, with typically not more than 8 reviews each. That, of course, cannot be done by hand, but requires a program which would match, at best, the topics selected by the authors for each paper with the expertise domain of each reviewer. The automatic program also checks for the possibility of conflicts of interest. However, the automatic approach is far from being perfect due either to technical limitations or limited knowledge about the submitted work. For instance, relying on the email address to identify the affiliation of the author, and relationships between reviewers and authors is not always accurate since many authors prefer to use generic email accounts such as Gmail or Yahoo. The Area Coordinators greatly helped review the final match by visually inspecting every single paper and suggesting potential reassignments. We then notified the reviewers of their assignments, requesting their help to check additional potential conflicts or issues. In fact, after the initial assignment, we had to resolve about 150 review reassignment requests due to conflicts which were not predicted or because of papers which were out of the range of expertise of the reviewer. After the initial reshuffling, several other reviewers’ reassignments took place due to conflicts of interest, incorrect topic assignments, changes of reviewers’ commitments during the course of the review period, etc. At the end of this iterative process, 678 manual reassignments where carefully completed.
Even with the reviewers’ best intentions, and even with a constant flow of reminders by the technical program chairs, there are always a number of papers which do not get a complete set of three reviews by the reviewer’s deadline. At the end of the official review period, we still had 594 reviews missing out of 4317 total. We decided to grant an additional week to the late reviewers, and the number of missing reviews went down to 117. At this point in time we switched into emergency mode and escalated the code alert to ‘orange’. We assigned the missing reviews to the crash reviewers who, with the help of the Area Coordinators, helped bring the number of missing reviews to virtually zero in a few days.
After the reviews were done, we started the task of creating the sessions. We asked each pair of Area Coordinators to analyze the papers in their domain, check for mixed reviews, perform additional tiebreaking reviews if necessary, and decide about which papers were to be accepted and which were to be rejected. That may not be an easy decision in many cases. With 1439 papers - most of them being of excellent quality - and a target acceptance rate below 59%, we had to make a number of hard decisions. The decisions were reviewed during the Technical Program Committee meeting in Rome on May 19-20, where a selected number of Area Coordinators, together with the Interspeech 2011 General Chairs and Technical Program Chairs, the President of ISCA, and the Technical Program Chair of next year’s conference worked to create the final sessions and give shape to a draft program. The program was reviewed and refined during the next few days, and more than 5,000 acceptance/rejection notifications—not without technical mailing problems—were sent to all the authors of submitted papers. Some review mistakes—like for instance, the posting of wrong reviews—were spotted by the authors. These mistakes had to be fixed, which produced slight changes in the final program.
Finally, all of the newly created sessions had to be allocated to the time slots, matched in size and predicted attendance to the available conference rooms, and assigned to potential session chair candidates in each area of expertise. Even though particular care was taken to avoid the same authors having to present different papers simultaneously in different sessions, clashes of this type always occur. So, upon request of a few authors, we had to review the allocation in order to come to the final program.
As it is clear by now, managing the program of a large conference requires a huge amount of work, and we could not have accomplished it without the continuous support of ISCA’s President Isabel Trancoso and the ISCA’s board. A great thank to the Interspeech 2011 General Chairs: Piero Cosi, who sat day by day with us in the trenches helping respond to the hundreds of requests we received and solve the apparently infinite issues that arose every day, and Renato De Mori, who helped us with his always insightful, practical suggestions and wisdom. Also our unconditional thanks go to those Area Coordinators who promptly responded to all of our requests often on a short notice of a few hours. And of course, many thanks go to all of those reviewers who did their work conscientiously and with great professionalism. Last but not least, a great thank you to the authors of all papers submitted. Without them and without their dedication to the science and technology of speech we would not have this wonderful and unique conference.
We are all set for a great Interspeech 2011 in Florence, the cradle of the Renaissance and one of the most beautiful cities in the world.
See you all in Florence,
Roberto Pieraccini and Pino Di Fabbrizio
Interspeech 2011 Technical Program Chairs