Accessing ModelFront

Accessing and integrating translation risk prediction

ModelFront provides its translation risk prediction technology to a diverse set of translation providers and users, from machine learning researchers to language service providers to enterprises to translation toolmakers.

ModelFront provides a translation risk prediction API and console. API integrations are implemented by clients, partners and third-parties.

The ModelFront console is the best way to handle offline use cases, like filtering and evaluation, and an easy way to get started on a proof of concept for production use cases like effort estimation, hybrid translation or final review.

Our listed partners work with us to make our technology accessible to their clients and users in a way that's transparent and optimized for specific translation use cases, as smart features within their ready-to-use platforms.


No-code with the ModelFront console

For offline use cases or as a proof of concept for integrations, risk prediction can be run with a few clicks via the ModelFront console, without implementing any integration.

For effort estimation, upload an untranslated XLIFF, choose the machine translation API, click Start and look at the overall score, histogram and preview.

For hybrid translation, one option is to run an evaluation, download a filtered TMX - slide the filter to keep only those machine translations below a certain risk - to boost the translation memory matches smartly, and then lock all exact matches in the packages sent to translators.

For final review, upload a parallel file, like a final XLIFF, and look at the riskiest segments in the preview.


Our partners

GlobalDoc

GlobalDoc is a full-service, quality-focused language service provider to clients operating at a global scale, like DRiV, IBM, Siemens, Toshiba and Xerox, with decades of experience building solutions for translation automation.

GlobalDoc's LangXpert translation management platform integrates ModelFront to implement hybrid translation, from pre-translation to post-editing and final review.

Contact GlobalDoc

translate5

translate5 is an innovative open-source cloud-based translation system available as a hosted or self-hosted solution, for translation, post-editing and review supported by companies and institutions like transline, Supertext, Pangeanic, World Translation and the German Research Centre for Artificial Intelligence.

translate5's "What you see is what you get" editor integrates ModelFront's risk prediction in the segment translation and document translation post-editing experience.

Contact translate5


Integrating ModelFront

Enterprises and language service providers have different options for integrating ModelFront, according to the workflow and technology stack, and their control of it.

The ModelFront team works with clients and partners to understand integration options and connect with qualified third-party developers.

Features

For hybrid translation, the most commonly implemented feature is locking auto-approved translations. In other scenarios, the risk score is just a visual hint to the translator, who can still edit.

Another option is filtering or sorting by risk, and potentially locking or flagging accordingly. Most TMS and CAT already support a filtering paradigm.

Implementations

For enterprises and language service providers with their own TMS or CAT, like platform builders themselves, can integrate the ModelFront API directly and implement features for localization managers, project managers and translators.

Those enterprises and language service providers without their own TMS or CAT but with their own developers can consider options like a developing a Trados plug-in, extending WorldServer or updating Smartcat or Lilt via their APIs.

With XLIFF and TMX, it's possible to lock segments across legacy file-based platforms. XLIFF includes the standard attributes translate=no and approved=yes and there are non-standard attributes like SDL locked=true. TMX can also be used to lock segments.


Open access

ModelFront also supports open research conferences like the Workshop on Machine Translation, labs like YerevaNN, projects like deepchar, and supports open datasets like the TAUS Corona Crisis Corpus.

Many of their projects, datasets and derivative works are open and downloadable in full.