Worum geht es in dem Projekt?
The aim is to build a community-based open source translation system – based on the existing open source projects openTM2, translate5 and Moses. The system will enable the simple integration of additional systems, including proprietary systems such as SDL GroupShare or Google Translate or other MT-providers like Slate Desktop.
What makes this project really different
is the community approach:
The major strategic goal is to build a community that enable
small and mid-sized companies as well as the translation
departments of large companies to have their own translation
software. A software
- they can truly extend and customize the way they need
- they can integrate the way they need
- they have no license fees for
- that enables them to compete with the big players who use their own software to push smaller companies out of the market
- and that is performing well regarding the new and advanced technical trends on the market.
As for this crowdfunding the specific technical objective is,
openTM2 users gain ...
- ... a modern & user-friendly graphical user interface
- ... the possibility to work online with multiple users and offline as a single user in a new modern interface with openTM2
- ... an easy and workflow-based way of proofreading their files
translate5 users gain ...
- ... translation memory and machine translation support
- ... getting matches from multiple sources
- ... parallel concordance search in multiple sources
Moses users gain ...
- ... a flexible GUI made for big data
- ... integration with file-based Translation Memories
- ... quality assurance designed by European MT research
Every Language Service Provider gains
- ... an easy solution to translate those tiny translation jobs through the web, that have a large management overhead, when dealt with the traditional way
- ... a translation system they can integrate and customize the way they need it without license fees
- ... the certainty of using a translation system, that can never and will never be owned by another Language Service Provider (due to its Open Source nature)
- ... a translation system for that they can develop features by themselves (or ask developers for it) and share the costs with others in the community.
- ... a lead development with a long-time background in translation-centric IT, that is yet completely independent of Language Service Providers or big companies.
The idea on the spot
Most small and medium-sized translation companies as well as big language buyers are looking for translation software that they can customize, integrate and upgrade themselves without high license fees.
Even more importantly: In the increasingly complex software landscape of the translation industry, small and medium-sized companies will only manage to stay independent and survive if they cooperate with each other on software.
The aim of this project is to solve this need through enhancing and integrating the existing open source solutions of the language industry based on a community approach of open source.
Yet another translation system?
Translation companies and freelancers already suffer from the increasing number of different translation tools they need to learn to use and support. Our system is not intended to be yet another one. Instead it will integrate other existing systems via API, support their bilingual file formats, use translation interchange packages, specific usage of translation memory, etc. This will enable users to reduce the number of systems they need to learn and support.
This integration is best performed by an ecosystem of community-driven open source systems that has no interest in binding users by keeping them from switching the system. By a system where the users themselves can enhance interfaces, integrations and plug-ins.
The targeted architecture: What will be developed with the money raised by this crowdfunding?
The goal of this crowdfunding project is to integrate translate5, openTM2 and Moses-based MT systems to achieve an integrated open source translation system that is installable on a desktop as well as on a server. translate5 will serve as a Graphical User Interface and server back-end and will connect to the other systems via REST-API.
The result of this is that openTM2 can run on the same server as translate5, but does not have to. OpenTM2 will be integrated in translate5 Windows installer (translate5 is platform-independent, openTM2 Windows only). Moses systems must be heavily trained and configured to make use of them. Moses will therefore not be included in the translate5 installer but will be connectable via REST-API.
The architecture will be designed in a way that will make it easy to write plug-ins to integrate more TM/MT resources, for example Google Translate or SDL GroupShare.
The attached mock-up screens show how the integration will look in the Graphical User Interface:
The TM/MT overview
- Is only available to users with project manager rights
- Shows the available resources
- File-based openTM2 Translation Memories (TM) can be added, downloaded or deleted here
- Adding a file-based openTM2 TM can be done by uploading a TMX file or a native openTM2 TM. The same applies to downloading a TM.
- Moses-based resources cannot be added through the GUI by the project manager. They are configured in the system configuration by the admin.
- As is usual for tabular overviews in translate5, the columns can be filtered and sorted.
The TM/MT task association
- Is located as a tab in the task properties window. This window can be opened in translate5 by clicking on the cog wheel-icon of a task in the task overview.
- By checking the check-box of a TM/MT-entry of the list, the entry is associated as match-resource for this task.
- As usual for tabular overviews in translate5 the columns can be filtered and sorted.
The TM/MT match look-up in ergonomic view
- Looks up all associated resources of the opened task
- Shows max. 5 matches for each resource
- Matches are ordered by match rate – resources are not ordered
- As soon as a resource responds, its matches are displayed. Matches of slower resources are inserted at the right position of the order, as soon as the resource responds.
- In tooltips the user sees the match attributes and the name of the originating resources.
- For sorting purposes, MT matches are treated in the same way as 70% fuzzy matches. The percentage can be adjusted in the translate5 configuration.
- Matches can be copied to the edited segment by keyboard shortcut
The TM/MT match look-up in all-column view
- Features identical to the TM/MT match look-up in ergonomic view of translate5
The Concordance Search
- Looks up all associated resources of the opened task
- Shows 5 matches for each resource, ordered by resource, first responding resources first
- Below the 5 matches of each resource is a “More” button. By clicking on this button all results of this resource are shown in a separate new tab.
- Matches are shown as soon as the resource responds. A resource that has not yet responded is shown as “loading”.
- All actions that can be triggered through the GUI can be triggered through the REST-API as well.
The architecture described will not cover all usual features of translation environments, but rather the features described in this architecture paragraph and the following paragraphs plus the features already present in translate5.
The systems to be integrated: openTM2
openTM2 is the open source version of the TM2 – the Translation Manager of IBM. IBM has been developing TM2 since the 90s and uses it to process the company's internal translations. Some years ago IBM decided to release TM2 as open source. For the open source version, the old TM2 has been modularized and improved. Many features have been converted to plug-ins.
In autumn 2014, IBM implemented a company-wide switch to openTM2 instead of TM2. Since October 2015, IBM has released the latest version of openTM2 to the public. It is in the interest of IBM to integrate other parties in the development of openTM2.
openTM2 is a proven and extremely fast translation memory engine. Due to its use in one of the biggest IT companies in the world, it has a lot of features tuned for the efficiency, automation and integration of machine translation. However, the Graphical User Interface of openTM2 has not changed since the 90s and many people will not like its GUI.
This makes the integration of openTM2 and translate5 a perfect match: openTM2 benefits from translate5's modern browser-based GUI and its server-based architecture. translate5 on the other hand benefits from openTM2's fast and robust file-based translation memory. Also, in future it will perhaps benefit from the vast range of features relating to machine translation, process automation and efficiency, as well as other features.
The result will be a stable and fast open source translation memory, installable as a stand-alone desktop tool as well as a server-based multi-user environment.
To achieve this integration, an openTM2 REST-API needs to be developed. These development efforts are part of what will be financed though this crowdfunding.
The systems to be integrated: Moses
Moses is the leading open source machine translation (MT) system. Most of the known proprietary MT systems build on Moses. Since Moses has to be trained and configured thoroughly (which is a lot of work and needs a lot of data), it does not make sense to simply include a basic Moses installation with translate5 installer, as this crowdfunding aims to do with openTM2.
Instead Moses will be integrated with translate5 via REST-API. This enables the user to pull machine translation matches for segments via web or LAN from any Moses-based MT system that uses the standard Moses REST-API.
The systems to be integrated: translate5
translate5 is a powerful browser-based system for handling, editing and quality checking bilingual texts. It is used in the translation industry as well as in MT research.
Its main user-centric features are:
- Modern and flexible Graphical User Interface
- Flawless handling, scrolling and high performance of thousands of files and hundreds of thousands of segments when editing a task
- Integrated terminology recognition via openTMSTermTagger
- Task management interface
- User management
- Basic workflow
- Can be installed and updated on the desktop or the server with a few clicks
For the integration of other tools translate5 offers:
- An event-based plug-in system for integrating functionality (and thus third-party tools) into GUI and backend
- A comprehensive REST-API
- A sophisticated queuing and load-balancing system for internal and external processes (currently e.g. used to handle multiple openTMStermTagger instances in combination with dependent translate5 internal tasks).
Was sind die Ziele und wer ist die Zielgruppe?
The project goal
A variety of open source tools exist in the translation industry. Together they cover large areas of what is needed in modern translation business. What is missing is a solution that plugs different tools together and combines them in a system that can be used in daily translation business. A system that can be customized and expanded to fulfil the needs of users. This is what this project strives to do.
Crowdfunding serves as a starting point for this goal by enhancing and integrating 2 open source systems (openTM2 and translate5) and loosely integrating Moses-based machine translation.
This starting point will establish a modern, browser-based open source translation memory system that is robust, capable of processing large amounts of data and can easily be expanded to integrate more systems. For more details please see above.
Who is the project for
Small and medium-sized translation companies and the translation departments of large language buyers are all under pressure in a translation industry that is becoming ever more technically complex.
They struggle by spending more and more money on tools and processes – but eventually they risk becoming completely dependent on the big players of the translation industry – or go out of business.
To maintain their independence and their mastery of their own processes, they need to cooperate on software. This project aims to initiate this cooperation for the core technology of the translation industry: The translation tools.
Freelance translators also benefit: They get a stand-alone open source translation memory tool with a modern GUI. The same GUI that translation companies can offer them for free when working with one instance of translate5. This reduces the tool costs for freelancers, gives them the possibility to engage in enhancing their tool (as it is open source) if they want and reduces the need to use a variety of tools.
Warum sollte jemand dieses Projekt unterstützen?
Are you a small or medium-sized translation company or do you work for the translation department of a large language buyer, and want to stay or become technologically independent with regard to your core technology? In a translation industry that is becoming all the time technologically more complex?
Do you strive for technological excellence in order to manage the future and wish to be and remain in control of your processes?
Do you want to be able to expand, customize and integrate your translation tool the way you need it?
This crowdfunding project aims to initiate and help manage the cooperation of small and medium-sized translation companies, as well as the language departments of large language buyers, in terms of software. A cooperation like this will be the only way for these players to stay independent in the future.
So if you agree or simply need a cost-efficient, customisable browser-based translation tool: You should engage in this project.
Would you like a free translation tool that is fast, easy to use, has a modern GUI and is capable of dealing with a variety of bilingual formats? A translation tool that you can expand yourself, if you want to? You should engage in our project – be either a part of the community or support the development.
Was passiert mit dem Geld bei erfolgreicher Finanzierung?
MittagQI is able to issue invoices for the funded money (for details see below).
The outlined functionality will be developed for translate5 and openTM2 and will be provided to the public as source code.
The first phase of the crowdfunding is successfully finished and the development started. In the first phase € 21.000,-- from 6 different companies have been raised (please see the companies listed as partners below). The new developments will be delivered to all supporting companies, as soon as they are finished and to the public as soon as the second phase will be funded.
The first phase successfully financed with € 21.000,--
the following development steps:
- € 18,520,-- translate5 part of the development. In detail: GUI integration, plugging in openTM2 and Moses connectors, managing of parallel calls to different resources, configuration, integration of openTM2 in installer (will be developed by MittagQI)
- € 2,000,-- coordination of the needs and interests of a high number of different companies and users (by MittagQI)
- € 480,-- from the first phase are kept for the second phase - so it's first drop is funded :-)
The second phase of the crowdfunding is now open for
funding. Funding goal: € 15.000,--
It's money will be used for the needed steps, that are still not financed:
- € 13,000,-- REST-API for openTM2 and support for translate5 internal translation format through openTM2 (will be developed by Gerhard Queck (Q-Soft), long-term developer and one of the lead developers of openTM2)
- € 2000,-- project management of openTM2 integration and testing of the openTM2 REST-API (by MittagQI)
- € 500,-- fee for StartNext
How does an "all-or-nothing campaign" work?
StartNext will only debit your account if the funding goal of the campaign is reached (for payment methods through direct debit and credit card). For the payment methods Sofort banking and Advance payment, StartNext will transfer the money back to you if the campaign is not successful.
Invoices for your participation
MittagQI is able to issue invoices for your donation. If you need an invoice, are located within the European Union and do not have a VAT number, or you are located in Germany, we will have to add German VAT to the money donated by you. This will need to be transferred to us independent of StartNext as soon as you receive your invoice.
Tax and selection of a reward
If you select a reward and you are located within the European Union and do not have a VAT number, or you are located in Germany, we will have to add German VAT to the money donated by you. This will need to be transferred to us independent of StartNext as soon as you receive your invoice.
Wer steht hinter dem Projekt?
Development & Organisation
MittagQI, represented in this project by Marc Mittag, initiated the project and will lead it, and will be developing the translate5 aspect of the functionality.
openTM2 functionality will be developed by Q-Soft, Gerhard Queck (openTM2/TM2 development lead since decades).
Supporters & Community
Michael Schneider, beo
Remy Blättler, Supertext
Geert Vanderhaeghe, Lexitech
Andrea Modersohn, oneword
Marjana Plukavec, IOLAR
Gábor Bessenyei, Globalese
Bertrand Gillert, Locasoft