Two of the speech team from Speechmatics visited Hyderabad in India last week for the 2018 Interspeech Conference. Held in the state-of-the-art Hyderabad International Convention Centre, the conference brought over 5,000 delegates from around the world.
Key objectives were to learn about the latest developments in Automatic Speech Recognition (ASR), to discuss current trends, and to discover the future of speech.
Interspeech is the largest dedicated speech conference, addressing all aspects of speech science and technology, from fundamental theories through to advanced applications including computational modelling and technology development inspired by recent advances in artificial intelligence (AI) and machine learning (ML). The Speechmatics team attended presentations on numerous research papers, discussed the latest speech developments with other researchers, and showed a technical demo of Speechmatics’ ASR. Major learnings from the conference included discovering that there is an interest in end-to-end solutions but it is still unclear as to whether it surpasses other approaches, and ascertaining future challenges with speech technology including the problems with far-field and noisy ASR, as well as under-resourced languages and domains. The key highlights from the speech team comprised:
Seeing Speechmatics listed as one of the most accurate speech companies in the world in an Adobe Research paper
Learning about the future trends of speech technology
Witnessing speech technology being applied to more and more real applications and domains
Observing the ASR community getting bigger and better
Nicolás Serrano Martínez Santo, Speech Recognition Engineer at Speechmatics said,
“I had an incredible time at Interspeech 2018 in Hyderabad last week. The ASR community assembled at the conference to present the latest research and progress achieved in the field. It’s exciting to see that the improved quality of current systems has enabled speech technologies to be applied in our daily life, such as the Amazon, Apple or Google assistants. However, real applications come with new problems and challenges that need to be solved.
“One of my personal highlights was seeing Speechmatics referred to in a number of research papers and listed as one of the most accurate after some evaluations, such as one done by Adobe Research. We need to continue to innovate and build on existing and new R&D to maintain our position in this growing and increasingly challenging field.
“I had an insightful time at Interspeech and am really looking forward to discovering where we will be in the next 12 months. See you at Graz for Interspeech 2019!”
Interested in speech recognition? Come and try our latest speech recognition demo at IBC in Amsterdam this weekend, Hall 8, stand E17. Alternatively, you can try our real-time demo here.