Machine Translation Evaluation
Would you like to dramatically reduce your localization costs AND reduce your time to market?
Impossible right? Maybe not.
You have heard about Machine Translation (MT) but maybe are unsure about what it means to you. Typical concerns we hear from our clients include:
1. The up front investment is too great
2. What we are doing now works fine so why change?
3. Will I have to change all of my internal processes?
Machine Translation is a translation productivity tool. It is not a replacement to the translator. MT essentially automates many of the tasks currently undertaken by the translator, thereby greatly reducing the overall translation time, reducing costs and allowing you to get your products to the global market faster.
So where do you start? It’s quite easy really. Just talk to VistaTEC about our Simple Seven Step Solution.
The VistaTEC Simple Seven Step Solution to MT Evaluation
1. Discussion – VistaTEC understand that any initiative cannot be successful unless project parameters and objectives are clearly understood and agreed by all parties. Prior to engaging in any solution, VistaTEC will first discuss with you what YOU want to achieve. We sculpt all of our solutions to meet the individual client needs.
2. Target Languages – VistaTEC will work with you to understand and prioritise the languages you wish to localize.
3. Analysis – Once we are fully satisfied that we understand your current situation, your specific objectives, and all relevant project information including your target languages, the VistaTEC Professional Services team will get to work. Our industry leading teams will analyse all of the available data on the project and begin to build the MT picture for you.
4. Technology Selection - A key part of your Machine Translation Evaluation strategy will be selecting the appropriate MT tool or tools to best suit your requirements. As a vendor independent organization, VistaTEC will objectively select and recommend the tools that we feel best suit your business priorities.
5. Test – As part of the overall evaluation process we would advocate that thorough testing is completed to ensure that overall project objectives are achieved. Client reviewers would then be in a position to determine whether or not the post edit test was acceptable.
6. Benchmarking – Once the test phase has been completed VistaTEC will produce a series of performance statistics that will allow you to evaluate the proposed MT solution in your environment.
7. Project Plan – On completion of sufficient testing and benchmarking exercises you will now be in a position to confidently make your decision on an MT strategy and begin to plan a larger technology rollout.
Let’s take this forward...
To start evaluating what Machine Translation can do for your company contact the VistaTEC Productivity Solutions Group today at productivitysolutions@vistatec.ie for a FREE consultation with an experienced member of our team.
Other VistaTEC Machine Translation related services include:
Preparation of MT Output
Statistical
When preparing a Statistical system for production we concentrate on the following areas:
• Preparation of all available bilingual data for the engine, e.g. TMs, glossaries;
• Creation of base line and bespoke materials for the target languages;
Rule-Based
Through our experience of working with Client A our team has gained considerable experience in the following key areas of preparation for Rule-Based MT:
• Use of Acrocheck to extract key terminology from the source files;
• Encoding of terminology in Systran’s Dictionary Manager;
• Development and constant refinement of regular expressions to search and replace known linguistic errors made by the MT system.
Evaluation of MT Output
Once any MT output has been generated, be it from a rules based or a statistical based engine, it is essential to review the output in order to determine the quality and monitor how well the MT engine is working. There is an ever evolving number of ways to evaluate output and each can be tailored to fit a specific clients requirements.
Statistical
In our evaluation of Statistical MT output VistaTEC concentrate on the following key areas:
• Creation of quality metrics;
• Analysis of quality metrics.
Rule-Based
In our evaluation of Rule-Based MT output VistaTEC concentrate on the following key areas:
• Untranslated terms: Any terms that have not been translated by the MT system, are encoded in its Dictionary Manager.
• Mistranslations: Any terms that have been incorrectly translated by the MT system, are encoded in its Dictionary Manager.
• Misspellings: Misspellings are solved by encoding the correct terminology in its Dictionary Manager or with the use of regular expressions.
• Collection of recurrent linguistic patterns or structures regarding grammatical and syntax errors that can be fixed with the use of regular expressions.
Following a complete evaluation of the output and resolution of the issues outlined above a new output is generated to make sure that the fixes have produced the desired result.
Post-editing – Selection and Service
It is now well established that not every linguist or agency is suitable for the task of post editing. While the existing translators know the content and products extremely well, their affiliation to the traditional process can prejudice their opinions and make it difficult for them to maximize the potential efficiencies that MT and post-editing process can offer.
VistaTEC offers MT and post-editing services that are tailored to our clients needs. By engaging early in the MT set-up we can work with the client to identify the skills necessary for their situation and over the course of 2008 one of our objectives is to build a more robust team of post-editors who are chosen and trained specifically for this specialized service.
VistaTEC & MT Research
On 13 November 2007 Science Foundation Ireland announced the establishment of a €30 million research partnership to be funded under its Centre’s for Science, Engineering and Technology (CSET) Campus-Industry Partnerships Programme. Funding was awarded to a consortium of universities and industry partners for a collaborative, multidisciplinary research project that will carry out fundamental and applied research underpinning the design, development, implementation and evaluation of the next generation of high-tech automatic language translation. VistaTEC was invited to join the university-led consortium as an industry partner along with seven other domestic and international localization industry companies.
The project, which will run over five years, counts Prof. Andy Way of Dublin City University's Machine Translation group and Reinhard Schäler of the University of Limerick and LRC among its senior academic members and includes firms like IBM, Microsoft, and Alchemy Software. The project intends to tackle the problems of volume, content delivery and personalization, which are considered critical problems for the localization industry.
VistaTEC have made several commitments to CSET, however we are also participating in many other fields. We were one of the first of the Translation Automation Users Society (TAUS) and have strong links to DCU’s School of Applied Language and Intercultural Studies.
Our aim is to provide a marketable solution that blends an understanding of technology with the right linguistic approach, and we see our involvement in research, as well as execution, to be fundamental to this.
Let’s take this forward...
To start evaluating what Machine Translation can do for your company contact the VistaTEC Productivity Solutions Group today at productivitysolutions@vistatec.ie for a FREE consultation with an experienced member of our team.