Interspeech 2018 Conference Review

This year's Interspeech conference took place in Hyderabad. This was the first time for India to host Interspeech and the organizers have done a remarkable job to ensure that it will be remembered as a truly special event. Immediately upon arrival to Rajiv Gandhi airport participants were greeted by volunteers, who promptly escorted delegates to numerous available taxi that took them to many hotels conveniently located in the Hitec area of Hyderabad, where Interspeech was hosted. Shuttle services, which ran twice a day, provided a convenient means of transport for the delegates to reach Interspeech and return back to their hotels. The main venue was a spacious area co-adjacent to one of the hotels, which featured one main hall that hosted sponsor booths and poster sessions, as well as multiple large and medium halls where oral and special sessions took place. In addition, there was an abundance of events to ensure that the delegates were pleasantly occupied in the evenings. Furthermore, the main conference banquette was a huge success, which featured a band of talented musicians that swept large crowds to the dance floor.

The main conferences started with a talk by ISCA medalist Bishnu Atal, who reminded us that cross-continental phone calls many of us make were once at the forefront of engineering science. Despite a significant progress made over the last hundred years, a large number of challenges nevertheless remain. As always, the keynotes of this Interspeech were broad and diverse. In the first keynote Jacqueline Vaissière discussed cross-linguistic prosodic tendencies. The third day of the main event started with Hervé Bourlard taking delegates on a journey of neural network evolution in speech recognition. Helen Meng presented the last keynote in which she looked at practical aspects of applying spoken language technology to language learning and wellbeing. The number of keynotes essentially doubled this year, thanks to the introduction of Perspective Talks that featured Nima Mesgarani, Dilek Hakkani-Tür, Sriram Ganapathy and Bhuvana Ramabhadran, who covered topics starting from speech processing in the brain to situated goal-oriented dialogue systems.

The main track stayed true to the spirit of previous Interspeech conferences by incorporating a very diverse range of subjects. One noticeable trend was an increase in the number of papers employing end-to-end approaches. Another noticeable trend was a rather sharp decline in the number of papers on language modelling. Compared to the previous 2017 Interspeech in Stockholm, which had featured three sessions, the current Interspeech featured only one full session devoted to language models. Overall, the oral sessions were well attended and the poster sessions, as always, attracted large buzzing crowds. In addition to the main track, a record number of special sessions and challenges took place during the conference including neural network interpretability, direct synthesis, low resource speech recognition, code switching, speech diarization, clinical applications, spoken CALL and computational paralinguistics.

The last day of the conference culminated in the awards ceremony. In addition to the always exciting announcement of the three best student paper awards out of twelve short-listed papers, ISCA honoured Christian Wellekens with the Special Service Medal whilst Elsevier awarded the best paper award for articles published in Speech Communication to Martin Cooke and his colleagues for their work on speech intelligibility. Finally, the joyful sounds from Austria, memorable to delegates from the past Interspeech events, reminded us that the next 2019 Interspeech will be taking place in Graz, Austria, where we hope to see many of you.