Hottest 2018aichallenger machine translation

  • Detail

On December 19, the defense and award ceremony of the finals of the "2018 AI Challenger global AI challenge" uploaded good news: Jinshan software 4. The oil pressure sensor of the electronic tensile testing machine needs to pay attention to the need to raise the upper jaw seat by about 10mm before clamping the sample. The two teams of AI Lab under the group (hereinafter referred to as "Jinshan") participated in this competition for the first time, After several rounds of competition with the R & D teams of enterprises and universities from 81 countries around the world on the electric box of the experimental machine base, he finally won the first and second place of this year's AI Challenger "English Chinese text machine translation" competition with a large score advantage

Li Kaifu, chairman and CEO of Innovation workshop, and Yao Dong, vice president of Jinshan office, and Dr. Li Changliang, head of Jinshan AI Lab

AI Challenger aims to "challenge the problems of the real world with AI", and has set up an open-source dataset and programming competition platform for global AI talents. It is committed to promoting the integration of AI in the field of scientific research and business, and creating a benign and sustainable new ecosystem of AI scientific research and application

the two teams sent by Jinshan won both awards, which not only proved the team strength and innovation potential of Jinshan team in the field of artificial intelligence technology, but also presented a congratulatory gift for the 30th anniversary of the founding of Jinshan with outstanding achievements

group photo of the 2018 champion team of the global AI challenge

the rise of new forces in the field of machine translation technology

machine translation is known as the "pearl on the crown" in the field of natural language processing. It can promote the elimination of cultural barriers, strengthen world exchanges, and make language no longer become a communication barrier that hinders people of different languages and cultures

in 1954, IBM launched the "Georgetown IBM" experiment, using a mainframe to translate 60 Russian Sentences into English. This experiment is regarded by academia as the first machine translation in history. In the following decades, many scientists have been trying various technical ideas, such as example based, rule-based, statistics based, etc. the research of machine translation has experienced ups and downs several times. Until the emergence of machine translation technology based on neural network and deep learning, the dream of "barrier free communication in the world" has become possible

It can make the light uniform and soft

2018 AI Challenger "English Chinese text machine translation" track, all teams are free to choose machine translation technology, but they need to train their own machine translation system according to the data provided by the evaluator. The evaluator uses the objective evaluation indicators of machine translation (Bleu, NIST, TER), of which Bleu is the most important evaluation indicator. In this year's competition, the total amount of data reached 13million pairs, and the bilingual data with context in China and Britain reached 3million pairs with the rising price of plastic chemical raw materials and economic development, which was significantly expanded compared with last year

a team from Jinshan AI Lab won the championship with an objective score of 3 Bleu points ahead of other competitors and a defense score of 30 points over other teams; Another team won the runner up with excellent results

taking part in such a world competition for the first time, you can enjoy the feast of artificial intelligence with "top giants" such as Stanford University, Tsinghua University, Chinese Academy of Sciences, bat, iFLYTEK and Yi, and achieve good results in the main track of machine translation, which shows that Jinshan, which has resolutely invested in AI research and development in recent years, has begun to reap rewards. In the field of machine translation technology, new forces have emerged

Jinshan AI Lab is young and full of vitality. In fact, the original intention of the team to participate in this competition is to test the team's ability with awards - in May 2017, general manager Lei personally led, Jinshan began to increase investment in research and development in the field of AI, and set up an AI team covering multiple research directions such as image, voice, language, etc., bringing together outstanding graduates from Chinese Academy of Sciences, Peking University, University of Melbourne, Harbin Institute of technology and other universities. In a short time, Jinshan AI team established a robust machine translation framework, built the infrastructure for training and tuning, and was equipped with a strong underlying computing service capability. Finally, it led the whole competition and won the championship in one fell swoop

explore core technologies to accelerate the implementation of AI

a hundred years ago, Mr. Yan Fu proposed the three principles of translation quality of "faithfulness, expressiveness and elegance". At present, the words translated by machine have just touched the level of "faithfulness" and "expressiveness", which is far from meeting the requirements of "elegance". However, from a practical point of view, machine translation technology is enough to create an efficient and reliable cross language intelligent auxiliary translation service. In the future, Kingsoft will strive to provide users of WPS, Ciba and other products with stable and near artificial translation, and users will be able to experience intelligent services such as instant text translation and simultaneous voice interpretation

in addition to machine translation, Jinshan AI Lab has also carried out in-depth exploration in image, text, voice and other directions. Jinshan invited Dr. Li Changliang of the Chinese Academy of Sciences to join in and preside over AI lab. It is under the leadership of Dr. Li that several young researchers can innovate unique algorithms such as hierarchical attention mechanism and Gaussian search, and accumulate advanced experience in engineering details and feature extraction. At present, Jinshan has applied for patents for relevant algorithms and technologies, and published relevant papers at international top-level conferences. At the same time, Jinshan pays more attention to the deep integration of technology and product services to complete the implementation of technology. At present, Jinshan is exploring the integration of a variety of artificial intelligence technologies to achieve human-computer cooperation and help the intellectualization of Jinshan office scene

although Zhou was an old country, his life was reformed. Today, the 30-year-old Jinshan has always adhered to technology, constantly made breakthroughs and transformations, and constantly created innovative products to serve hundreds of millions of users. At present, AI has become Jinshan's long-term strategy, and the company is going all out to explore core technologies and seek continuous breakthroughs

Jinshan future, AI on the road


Copyright © 2011 JIN SHI