ActivaTM is the most advanced multi-channel, multi-format and multi-language translation memory system. It is based on Elastic Search technologies. Because of its open-source origins, it has been designed in order to supersede the expensive model of server-based CAT tool providers.

ActivaTM is indeed a server translation memory, but with no string attached. As reported by Jost Zetsche in his Tool Box Journal (, ActivaTM brings the first kind of real innovation to the CAT tool – Translation Memory environment in decades. In short, ActivaTM separates the need of every Computer Assisted Tool having its own Translation Memory

You can install ActivaTM on your own server or in the cloud as a SaaS, populate it with all the Translation Memories you want and leverage full and subsegment matches from whatever initial translation tool. Yes, you can mix TMs from any program, from free TMs and retrieve translation from all of them simultaneously.

Through intelligent, deep learning tag handling systems developed as part of our R&D program with the EU’s Expert Project, interoperability issues and different ways of handling tags are resolved. You can run ActivaTM with multiple translation platforms simultaneously. Translators can work in a variety of translation programs and all access the same TM remotely and instantaneously.

Better still, ActiveTM will compare all incoming material so that it will create triangular TMs. If you have translated a text from English into Chinese and from English to Japanese, presto! You also have a Chinese-Japanese translation memory. ActivaTM was first known as “ElasticTM”. It provides unknown levels of flexibility and ownership over linguistic assets at a fraction of the cost.

We have developed ActivaTM as part of our EXPERT Project, with a working grant from the European Union. ActivaTM works as a server-based fast and scalable Translation Memory Manager (TMM) with extremely advanced search and fuzzy match capabilities beyond traditional TM systems.

Technical specification

ActivaTM is developed mainly in Python. It is based on ElasticSearch (for storage and full-text TM queries) and Apache Spark (for parallel computations). It has a full-featured REST API so you can open it up immediately to all your freelance or in-house translators. Major features include:

  • Fast TM queries
  • Reverse language pairs (an English-Arabic translation generates Arabic into English entries, too)
  • Import/export of TMX files
  • Extensive TM filtering capabilities (for export and queries) by languages, domains, organizations, industry, creation date etc.
  • Built-in new language pair generation using pivot language (for example, French -> Spanish and French -> English will generate Spanish into and out of English)
  • Flexible, customizable fuzzy match (using various regex patterns and linguistic features).
  • XML tags handling (including deep learning model for accurate transferring of XML tags)
  • Supports output mode for Moses for easy integration in Machine Translation flow
  • Compatible with SDL Trados Studio via dedicated plugin (including manual update of translation units)
  • Pretranslation of segments when only one or two items differ (words, dates, etc.)
  • Built-in TM content analysis & maintenance: clean “dirty” segments, create new segments by splitting existing ones based on linguistic rules and POS tagging
  • Advanced user scope management by limiting access to certain language pairs, domains and queries amount
ActivaTM was first introduced as ElasticTM at TAUS Riga 2016. It was presented officially in TAUS webinar on September 7th, 2016.