The glossary-controlled machine translation tool

What is IP.Translator?

IP.Translator is a translation program implemented as Microsoft-Word plugin, which is optimized for computer-aided translation of patent claims and specifications.

IP.Translator uses a glossary-controlled neural machine translation algorithm. The proprietary AI method is based on a specifically trained deep neural network and translation results can be conveniently adapted by the user, to obtain a high-quality human-controlled translation in most efficient way.

The currently available free trial version (usable until 31 December 2020) allows translations from English into German. The reverse translation direction as well as other languages are on the roadmap.

Are you unhappy with computer translations of which results do not fit to the specific field of a patent document or do not translate individual terms consistently?

keyboard_arrow_rightIP.Translator uses a document-specific dictionary (glossary), in which users enter preferred terminology, to control the choice of words to be used by the translation algorithm, and to create consistent translations.

Are you concerned about translating confidential text online with a machine translation service that transmits your text over the internet to a remote server?

keyboard_arrow_rightIP.Translator works as a local installation without internet connection. In contrast to the gigantic amounts of data required by other models, the installed neural machine translation model itself has only about 250 MBytes. The translation works on a current mid-range office PC without high-performance graphics card.

  • How to use IP.Translator?arrow_drop_down

    • In order to download the currently free of charge software, you need to create an IP.appify user account, providing an e-mail address to be verified, so that we can inform you about news such as updates.
    • After login to your user account, you can download an installation file that sets up a Microsoft-Word plugin, copies the IP.Translator language model and installs required library functions.
    • IP.Translator is not an independent application, but integrated as plugin in Word and can be found there, after opening a document to be translated, under the tab "IP.appify".
    • In the menu, the text to be translated can be automatically split into segments, which are to be translated individually, and when selected, the translation algorithm provides an adjustable translation proposal in a pop-up window.
    • During the translation process, you also create a glossary for the document to be translated, in order to achieve a consistent choice of words in the target language. The glossary is stored in the document and can be exported for reuse in other documents, imported from previous translations, and can also be removed again.
    • The video below provides further usage details, and a more detailed written guide can be found in these operating instructions.
  • What are the system requirements?arrow_drop_down

    • Microsoft Windows 10 with Office 2016 or later (no Mac OS)
    • RAM: 8 GB recommended (4 GB minimum)
    • CPU: 4 cores with 3.5 Ghz (boost) recommended (2 cores with 3.0 Ghz minimum)
    • GPU: not required

The technology behind IP.Translator

  • IP.Translator builds on most recent advances in language modeling and neural machine translation to deliver a highly performing AI solution for machine-aided translation. It employs a transformer-based deep neural network which has been extended to utilize explicit bilingual dictionary entries.
  • The underlying machine learning algorithm is specifically optimized for the translation of patent claims and specifications and is trained by up-to-date deep learning techniques using selected bilingual patent specifications and other patent literature as training data.
  • The dictionary entries made by the user influence the choice of words in the translation algorithm in order to achieve rare, document-specific preferred translations. This improves the quality of the machine translation and saves working time. Said advantage even increases when the dictionary is reused in other texts with similar subject matter.
  • The relatively small installation, which can be executed locally on an office computer, preserves confidentiality of the translated texts, and saves resources compared to cloud-based machine translations.
  • The technology for including explicit bilingual dictionary entries in the neural machine translation method and training has been invented by Dr. Thomas Eissfeller, the development lead of IP.appify GmbH. Patent applications covering this technology have been filed.