The fourth “Oriental Language Recognition Challenge” (OLR), jointly hosted by Speechocean and Tsinghua University and co-organized by Xiamen University, Northwestern Polytechnical University and Duke Kunshan University, announced the contest results in the APSIPA ASC 2019 in Lanzhou on November 20, 2019. The award ceremony was held at the dinner party that day.
Wang Dong, a professor from Tsinghua University, Dr. Tang Zhiyuan and Song Liming, marketing director from Speechocean, jointly presented the champion teams of the three competition tasks with awards.
(Tsinghua University and Speechocean presented the champions with awards)
Specific competition task content and the championship:
Task 1: AP19-OLR-dev doesn't include short-duration speech data specifically, but the short-duration test set in previous challenges can be used for development.
Task 2: for the cross-channel task, AP19-OLR-dev includes a subset that contains 500 utterances per target language.
Champion: Samsung Research Institute China - Beijing
Task 3: In the zero-resource subset of AP19-OLR-dev, 3 new languages are provided with 5 utterances for enrollment and 500 for test. Note that the 3 languages in the final test are different from those in the development.
Champion: Xiamen University
The oriental language family is considered to be very diverse and complicated. After a long history of national exchanges and integration, especially with the influence of English on the local languages of oriental countries in the past century, there are various acoustic and linguistic phenomena in the oriental language family, which has attracted the attention of phonetic and language researchers. The research into these fields often requires a large number of language databases, including text data and speech data. However, at present, the oriental language database resources suitable for scientific research are very scarce. To promote the research on oriental Language in academic circles, Tsinghua University, together with Speechocean, an internationally renowned provider of the data resource and data service, launched the special topic and language recognition competition for oriental language research at the APSIPA ASC 2016.
In the past four years, multiple well-known institutions have jointly participated in hosting OLR — Xiamen University, Northwestern Polytechnical University, and Duke Kunshan University. The number of participating teams has increased from 8 in the first session to 45 this year, including many internationally renowned academic institutions and enterprises. The training data of the competition has grown to over 140 hours with 16 languages — mostly uncommon minority language in the AI research field. Besides, cross-channel LID and zero-resource LID are also added to the competition tasks this year, making the competition more challenging and interesting. All these changes indicate that with the expansion of the connotation and extension of OLR, the competition itself is also closely combined with applications step by step.
In order to promote communication between the teams, we plan to host a technical seminar about the results of this competition in the near future. At that time, the winning teams of each task will be invited to share their experience in algorithm optimization. Everyone is welcome to follow the Linkedin account of Speechocean so as to get the latest information on time.
Speechocean, one of the initiators and organizers of this competition, is a globally leading provider of AI basic data resource and data service. With the data reserves including 130+ kinds of language resources in 70+ countries and regions, Speechocean has always devoted itself to providing specialized engineering data products and services for enterprises and scientific research institutions in the whole industry chain of AI. The business involves various domains such as speech recognition, speech synthesis, computer vision, lexicon, and natural language processing and provides relevant services for the design, collection, transcription, annotation, etc. of data. Moreover, the products and services of Speechocean have been applied to various AI domains, such as smart home, smart speaker, autopilot, and vehicle navigation.