Victory for Sun* Bear at the national online contest on natural language processing

This sixth workshop of VLSP in 2019 introduced 4 shared tasks aiming at tackling some of the most challenging issues in natural Vietnamese processing, Hate Speech Detection on Social Networks, Vietnamese Dependency Parsing; Automatic Speech Recognition; Text To Speech.

The workshop attracts many research groups in the country to come and share experience, and it also promotes collaboration among universities, research institutions and tech businesses. 

In VLSP 2019, Sun* participated in Hate Speech Detection on Social Networks and Text To Speech.

For Hate Speech Detection on Social Networks, the organizer sent the teams 20,000 flagged Facebook posts or comments (regular posts/posts with vulgarities, unsuitable to the Vietnam's fine customs/posts with negative content, attacking particular targets) for system training. 

After this phase, the organizer sent 5,000 unflagged posts or comments, and the systems that the teams had developed would filter and submit the result to the organizer.

The final victory belongs to Sun* Bear of AI Team, with an exceptional score (0.61971), surpassing the 2 other contestants. The runner-ups are ABCD from University of Information Technology, Vietnam National University - Ho Chi Minh City and Try Hard from Vietnam AI System, with the scores of 0.58883 and 0.58445 respectively. 

Furthermore, the AI Team also participated in Text To Speech. Each team has 45 minutes of 1000 audios in Northern Vietnamese accent, and 23 hours of 15,000 audios in Southern Vietnamese accent.

Then, each team receives 120 sentences, and the the teams must return all of these given sentences played in 2 accents based on the previous training data. The real recordings and the synthesized speeches are mixed together, then sent randomly to 24 people. They will listen to the audio files, determining which one is human voice and which one is not, then give ratings based on the file quality.   

The result of this contest will be announced in the official Workshop on Vietnamese Language and Speech Processing on 13/10/2019 at University of Science, 19 Le Thanh Tong, Hoan Kiem, Hanoi.