Case Study: First Large-Scale Application of Auto-Adaptive MT

by Han Mai
3 Minute Read

Combining Machine Translation (MT) with auto-adaptive Machine Learning (ML) enables a new paradigm of machine assistance. Such systems learn from the experience, intelligence and insights of their human users, improving productivity by working in partnership, making suggestions and improving accuracy over time.

The net result is that human reviewers produce far higher volumes of content, with nearly the same level of quality, for a fraction of the time and cost. Machine assistance can save customers up to one half (or more) of the price of traditional high-quality human translation services. Or, if you’ve been used to machine translation alone and have been unhappy with the results, watch your translation quality rise dramatically with a marginal increase in price.

Case Study: Travel Portal Translation

A large travel and tour web site wanted to localize 1.77 million words of content from their catalog into 6 languages within a two week window. The hard deadline was to be ready to accommodate the summer vacation plans of millions of global users with more destinations and new activities. Successfully achieving this goal required rapid mobilization of a high-quality team of humans, fully-empowered by robust machine assistance technology.

e2f, based out of San Jose, California, with over 15 years of success in the translation and localization business, provided the “human capital” for the project. Their team was comprised of 100+ experienced translators, editors, and reviewers, plus seven project managers and a senior localization engineer.

null

Lilt, based out of Palo Alto, California, provided the translation engine for the project. Founded in 2015, its technology platform incorporates the latest research in Natural Language Processing (NLP), Human-Computer Interaction (HCI), and Machine Learning (ML).

The Lilt platform proved invaluable in augmenting e2f’s human staff, increasingly translation speeds far beyond the industry average of 335 words per hour. The automation process required transformation of source Excel documents into a format suitable for automated processing, which were then uploaded into the correct accounts in Lilt via scripts calling Lilt’s APIs. Once translations were made within the Lilt system, output was generated and transformed back into Excel documents in the target languages. 

null

Human translators could then accept or revise these segments. If Lilt’s suggestions were rejected, the approved human translations were fed back into the system, which learned from the human’s perspective and expertise. This positive feedback loop enabled faster and more accurate translation over time as the human translators contextually taught the system preferred translations of terms and phrases.

The client’s reaction was overwhelmingly positive. The number of errors was low compared to traditional machine translation solutions and the quality in line with standard human translations. Given a two-week window, the project was actually completed within 10 days.

Lilt’s productivity gains helped us submit the winning bid on the largest translation project in our company’s history, and deliver both ahead of schedule and within budget.
— Michel Lopez, CEO, e2f
This was the first time ever that teams of translators were training an MT system collectively, interactively and in real-time. The project proved that auto-adaptive machine translation technology is ready for large-scale production use.
— Spence Green, CEO, Lilt
null

Interactive and Adaptive Computer Aided Translation

5 Minute Read

Originally published on Kirti Vashee’s blog eMpTy Pages. Lilt is an interactive and adaptive computer-aided translation tool that integrates machine translation, translation memories, and termbases into one interface that learns from translators. Using Lilt is an entirely different experience from post-editing machine translations — an experience that our users love, and one that yields substantial productivity gains without compromising quality. The first step toward using this new kind of tool is to understand how interactive and adaptive machine assistance is different from conventional MT, and how these technologies relate to exciting new developments in neural MT and deep learning. Interactive MT doesn’t just translate each segment once and leave the translator to clean up the mess. Instead, each word that the translator types into the Lilt environment is integrated into a new automatic translation suggestion in real time. While text messaging apps autocomplete words, interactive MT autocompletes whole sentences. Interactive MT actually improves translation quality. In conventional MT post-editing, the computer knows what segment will be translated, but doesn’t know anything about the phrasing decisions that a translator will make. Interactive translations are more accurate because they can observe what the translator has typed so far and update their suggestions based on all available information.

Read More
screen_shot_2019-02-23_at_1.22.36_pm

Product update: improved translation review categories

1 Minute Read

We’re pleased to announce a new feature in our interface that improves the translation review process. Previously, reviewers could make general text comments about errors and introduce categories by using hashtags to indicate the type of error (such as #punctuation or #mistranslation). However, it was still a manual process to collect the different hashtags and identify precisely where the error occurred.

Read More