2024-12-23 14:26:05
Author: Sogou Inc. / 2023-07-23 20:32 / Source: Sogou Inc.

Sogou's Co-developed Speech Translation System Ranked No.1 at IWSLT 2018

BEIJING,Nov.8,2018 -- Sogou Inc. (NYSE: SOGO) ("Sogou" or "the Company"),an innovator in search and a leader in China's internet industry,today announced that the Company's co-developed Speech Translation System ranked first in the Baseline Model of Spoken Language Translation (SLT) at the International Workshop on Spoken Language Translation (IWSLT) 2018 that took place in Bruges,Belgium,October 29-30,2018. Separately,two research papers co-authored by Sogou Vice President,Jingfang Xu,and AI researcher,Feifei Zhai,have been presented at the 2018 Conference on Empirical Methods in Natural Language Processing (EMNLP) held in Brussels,October 31 - November 4.

Places first at IWSLT Spoken Language Translation competition

Held yearly since 2004,IWSLT is one of the most influential speech translation scientific workshops. In the SLT Baseline Model,the Sogou-TIIC Speech Translation System ranked highest with 28.09 BLEU points,a common metric for measuring the accuracy of machine translation,1.6 BLEU points higher than the next company. Other participants included iFlytek,Alibaba,Applications Technology (AppTek) and the United States Air Force Research Laboratory (AFRL). Sogou's Speech Translation System was developed in collaboration with the Tiangong Institute for Intelligent Computing (TIIC),an AI research institute jointly-established by Sogou and Tsinghua University in 2016.

Yanfeng Wang,leader of Sogou's Voice Interaction Technology Center which co-developed the speech translation system used at IWSLT,said,"We are delighted to have won the Baseline Model competition,the toughest global contest in speech translation. These technological advances will soon be available as updates to Sogou's machine simultaneous translation system,Sogou's Smart Travel Translator,Smart Recording Translator and Sogou Input Method's translation function."

Research papers presented at EMNLP

EMNLP is a leading conference in Natural Language Processing (NLP) organized by SIGDAT,a special interest group at the Association for Computational Linguistics (ACL) that focuses on linguistic data and corpus-based approaches to NLP. Sogou is a sponsor of the 2018 EMNLP conference.

The papers titled Improving the Transformer Translation Model with Document-Level Context and Three Strategies to Improve One-to-Many Multilingual Translation,present ways to improve translation performance by feeding more contextual information to translation machines,or by developing one-to-many multilingual translation models,which are more efficient than the current one-to-one translation models commonly used in the machine translation industry.

Jingfang Xu,Sogou Vice President,"Our research papers will help to improve the accuracy of machine translation,putting us one step closer to human-machine parity in translation. The updated Transformer Translation Model means machines can now process information in the context of the entire document,rather than sentence-by-sentence or line-by-line,resulting in a better translation. These advances will also be applied to Sogou's current products,including Sogou's online translation tools and the Sogou Smart Travel Translator."

Safe Harbor Statement

This announcement contains forward-looking statements. It is currently expected that the Business Outlook will not be updated until release of Sogou's next quarterly earnings announcement. However,Sogou reserves the right to update its Business Outlook at any time for any reason. Statements that are not historical facts,including statements about Sogou's and Sogou management's beliefs and expectations,are forward-looking statements. These statements are based on current plans,estimates,and projections,which involve inherent risks and uncertainties. We caution you that a number of important factors could cause actual results to differ materially from those contained in any forward-looking statement. Potential risks and uncertainties include,but are not limited to,intense competition in the market for search and search-related services; our need to continually innovate and adapt in order to grow our business; our reliance on Tencent platforms for a significant portion of our user traffic; and uncertainty regarding the extent and reach of PRC governmental regulation of sponsored search. Further information regarding these and other risks is included in Amendment No. 2 to Sogou's Registration Statement on Form F‑1 (Registration No. 333-220928) filed with the Securities and Exchange Commission on November 6,2017,and other documents Sogou files with or submits to the Securities and Exchange Commission.

About Sogou

Sogou Inc. (NYSE: SOGO) is an innovator in search and a leader in China's internet industry. With a mission to make it easy to communicate and get information,Sogou has grown to become the second largest search engine by mobile queries and the fourth largest internet company by MAU in China. Sogou has a wide range of innovative products and services including the Sogou Input Method,which is the largest Chinese language input software for both mobile and PC. Sogou is also at the forefront of AI development and has made significant breakthroughs in voice and image technologies,machine translation,and Q&A,which have been successfully integrated into our products and services.

For media enquiries,please contact:

Rachael Layfield


Brunswick Group


Tel: +86 10 5960-8600


Email: sogou@brunswickgroup.com

Sogou's Co-developed Speech Translation System Ranked No.1 at IWSLT 2018

View original content:/news-releases/sogous-co-developed-speech-translation-system-ranked-no1-at-iwslt-2018-300746479.html

Tags: Computer Software Computer/Electronics

Previous:

Next:

Leave a comment

CUSMail

CusMail provide the Latest News , Business and Technology News Release service. Most of our news is paid for distribution to meet global marketing needs. We can provide you with global market support.

© CUSMAIL. All Rights Reserved. Operate by Paid Release