1 year ago
August 5, 2020
Intento Publishes the State of Machine Translation 2020 Report
Over the years, Intento reports have become a go-to benchmark in the machine translation ecosystem for enterprises, consulting agencies, and LSPs. The research aims to analyze the performance of commercially available machine translation engines on comprehensive datasets.
This year, Intento has focused its efforts on researching how stock MT engines perform in different domains. TAUS, the industry leader in language datasets, provided domain-specific data in 14 language pairs and 16 industry sectors, including Computer Software, Legal Services, Financial, and Healthcare. Intento has evaluated 15 pre-trained MT engines to assess how machine translation performance varies depending on the business sector or content type, what language pairs and industries show the highest MT quality across the board, and how many MT engines are needed for best translation results.
- Each of the 15 evaluated machine translation engines is best at something. Seven of them are enough to get the best quality for all 16 industries and 14 language pairs. For any given industry, one to four engines are enough to achieve the best quality.
- The highest MT quality has been seen in Computer Software, Legal Services, and Telecommunications, with Software Strings and Documentation, Support Content, Policies, Processes, and Procedures being the most accessible for machine translation.
- As many would expect, content in Professional and Business Services, as well as Instructions for Use and Sales and Marketing Content in other industries, are the hardest for MT to get right.
- An incredible spike in language coverage is observed: + 2 000 language pairs since June 2020, many low-resource languages have been added.
- MT Landscape continues to evolve: 11 more vendors offer pre-trained MT engines since June 2019.
The full report can be found at this link.
A CALL FOR LSP VOLUNTEERS
Following this broad study, Intento will proceed with more in-depth research of MT quality in selected domains and language pairs. We are looking for LSP volunteers to perform human LQA for this research, become Intento MT Evaluation Partners, and get limited free access to the Intento platform.
Intento helps global companies procure and utilize the best-fit cognitive AI services. The Intento AI Hub connects AI models trained on multiple platforms (such as Amazon, Google AutoML, or Microsoft Cognitive Services) with many enterprise software systems. Launched in 2017, Intento recently received a U.S. patent for its core platform and works with global technology, retail, and travel companies, augmenting their Localization, Content Management, Customer Support, and Marketing Operations with AI. For more information, visit inten.to.