Case Studies

The Challenge

Speed is critical to Signal AI’s clients and foreign content in particular poses problems for global communication teams who need to stay ahead of the news cycle.

Back in 2017, when Signal AI was considering how to make the world’s non-English information available to their clients, it was common for media intelligence companies to charge a premium for access to foreign content, translating it only manually on request and with a significant delay. Moreover, foreign content was only accessible to users who were willing and able to search in the foreign language itself.

Signal AI wanted to overcome this bottleneck and give their clients instantaneous access to non-English information structured and analyzed by their AI platform. But it was not clear how to port Signal AI’s machine learning models to the 50 plus list of languages that Signal AI’s clients were interested in analyzing. This is where Signal AI decided to experiment with automatic machine translation.


The Solution

Signal AI realized that with the right machine translation service provider, they could translate and provide insights from the estimated 50% of the world’s information that is not in English.

Machine translated data could be fed into Signal AI’s platform as English, allowing Signal AI’s existing machine learning models to be applied to the data, irrespective of the language it was originally in. Moreover, the insights and indexed documents and entities would be searchable immediately using English search terms, and the source would be available in English or in the original language as preferred by the client, all in real time.

Signal AI quickly gathered a list of requirements for their prospective machine translation provider. The service would need to support a large number of languages (over 50 at least), support extremely high throughput with low latency and produce high quality English translations.

With machine translation from over 90 languages into English, automatic language detection and an API that easily scales to translate billions of symbols a day, Yandex Translate was an obvious partner for Signal AI.

First integrating Yandex Translate’s API in 2017, Signal AI has since seen translation quality significantly improve with successive generations of Yandex’s state-of-the-art Neural Machine Translation system.

In addition, building on their experience and infrastructure serving translations to millions of users a day, Yandex Translate has developed a specific high throughput solution for clients such as Signal who require massive amounts of data translated with extremely low latencies.


The Results

Signal AI’s platform currently ingests around 100 million documents per month. Some 36% of these are translated from non-English language sources in over 50 languages and nearly 200 countries.

By translating all foreign language sources in real-time and annotating them with their custom AI models, Signal AI is able to notify their users of a critical foreign language news story within seconds of it breaking wherever that story appears first and whatever language it is written in.

Translating data from hundreds of thousands of sources daily into English would be neither possible nor cost effective if done by human translators who can take hours to translate a single news article.

Dr Miguel Martinez, Co-Founder and Chief Data Scientist, Signal AI, «Yandex has been a valued partner as we have scaled our AI platform globally, helping to provide accurate and in the moment translation from over 50 languages, and ultimately driving better decisions by our clients, especially those who rely on our non English language news insight.

Make an Enquiry

Specialist Firm in Property and Immigration Law

RSL-LAW is authorised and regulated by the Solicitors Regulation Authorities –
SRA No: 591917

© 2019 RSL-LAW. All rights reserved.

Contact us

33 St James’s square, London, SW1Y 4JS

Telephone in London:
+44 (0) 207 060 5333


rsl-law webinars