Constantin Orasan is Reader in computational linguistics at University of Wolverhampton, and the deputy head of the Research Group in Computational Linguistics. His background is computer science and received his PhD from University of Wolverhampton with a thesis in the field of computational linguistics. He is currently the coordinator of the EXPERT project. In the past he managed other European projects like FIRST, QALL-ME and MESSAGE. His expertise is in information extraction, automatic summarisation, translation memories, question answering and corpus linguistics.
will present the poster…
The EXPERT project: Advancing the state of the art in hybrid translation technologies
Machine translation is playing an increasingly important role in our multilingual society, but in many cases the technology is not mature enough to be able to produce high-quality translations completely automatically. Current research is addressing this problem by developing better translation methods and by improving the way human translators can use computers in the translation process. Despite its importance, the field is lacking enough world-class researchers to ensure its fast progress. This paper gives a brief overview of the EXPloiting Empirical appRoaches to Translation (EXPERT) project, an FP7 Marie Curie Initial Training Network1 which is focusing on these issues.
The purpose of the EXPERT project is two-fold. As a training network, the project is preparing 15 Marie Curie fellows to become future leaders in the field. This is achieved by employing 12 Early Stage Researchers (ESRs) and 3 Experienced Researchers (ERs) at one of the 9 partners in the consortium, by organising dedicated training events and enabling intersectoral and translational secondments. The researchers employed in the project work together with established researchers from the consortium to promote the research, development and use of hybrid language translation technologies.
The work of the researchers is organised into 15 individual projects which aim to improve the state of the art from five different perspectives:
• The user perspective: the real needs of translators are being considered when designing the workflow, considering which are the best features of translation engines to be employed, and proposing evaluation metrics. This is being achieved by an extensive user requirement analysis and the collection of feedback from users. Collaboration with industrial partners ensures that the views of real users are considered.
• Data collection and preparation: the focus of the EXPERT project is on data-driven translation technologies. In light of this, the project is investigating innovative ways of collecting multilingual data from the web, and preparing it for specific applications.
• Incorporation of language technology in translation memories: most of the existing translation memories available today rely on little language processing when they match and retrieve segments. Research carried out in the EXPERT project shows that even incorporation of simple language processing such as paraphrasing can help translators.
• The human translator in the loop: in EXPERT we investigate how to exploit user feedback to improve the quality of translation. This is achieved by developing new quality estimation methods which can learn from the way the human translator interacts with the system.
• Hybrid approaches to translation: the project is proposing truly hybrid methods which combine the best features of all the techniques available, and test them in real environments.
1 More details about the project can be found at http://expert-itn.eu
The full paper will give more details about the project and its main outcomes.