Native Speech 2014 - a chance to make a name for young developers

Winners will receive valuable prizes and a dream job!

In the spring of 2014, the second open competition on the development of speech systems for young scientists - “Native Speech” - will be held. The winner will receive 120,000 rubles or participate in the international summer school on speech technologies, the silver medalist will receive an iPhone 5, and the finalist who will take the third place will receive an iPad 4.
')

Competition schedule:

January 15, 2014 - the beginning of the registration of participants
February 15, 2014 — opening access to the competition base, starting work
1-3 April 2014 - semi-final
April 26-28 - finals to be held in St. Petersburg

You can get acquainted with the conditions of participation and get all the necessary information on the contest page on the MDGs website spch.pro/x1fm2
A prerequisite for participation is to join the MDGs VKontakte vk.com/speechpro group.

In order to participate, it is enough to have programming skills, lively, creative thinking and a desire to solve interesting, non-standard tasks at the junction of several disciplines, such as dynamic programming, statistical modeling, etc.

The organizer of Native Speech is the Center for Speech Technologies company, the absolute leader of the Russian and a significant player in the international market of speech technologies and multimodal biometrics, the leading developer of speech systems in Russia. All three winners of the competition will be invited to join the MDG team.



Last year, contestants were asked to develop a prototype of their own speech recognition system in a month. This technology is closest to the human dream of the interaction between man and machine - everything that is said is literally transformed into text. The task of full recognition of continuous speech is not solved anywhere in the world, however, the recognition accuracy is already high enough to use the technology in practice. The potential scope of application of technology full-text recognition (speech-to-text) for commercial purposes is quite wide. These include automatic interpretation of phonograms, voice input of data into various electronic maps, voice control interfaces of equipment, preparation of hidden subtitles for TV and much more.

This year the task will be narrower. Participants will need to create a workable algorithm for converting the recognized sequence of phonemes into text that meets the norms of the Russian language. When creating the algorithm, it will be possible to use in-house developments and open software packages intended for research, such as: HTK, Kaldi, SRILM, mitLM, OpenFST. The use of commercial products and online services is not permitted.

For the victory in the contest “Native Speech 2013”, a grave struggle ensued, and the best result turned out to be quite impressive: the participant's system from the North Caucasus State University recognized 63% of what was said. This year, the organizers are waiting for the contestants of the same scientific passion and sports passion from the contestants!

Technical support, as in the past year, will be provided by one of the strongest Russian universities - ITMO.