The aim of OTELO is to meet the industrial need for compatible, powerful, and user-friendly language products. The project will develop an integrated environment to improve the availability, cost-effectiveness, and productivity of translation services. All aspects of the translation cycle will be covered, including integration with work flow and document management systems, and network- based access to machine or human translation facilities. Two major commercial machine translation systems will be adapted to OTELO specifications to allow for the sharing and reuse of existing expensive linguistic resources.
The OTELO project has successfully completed a number of significant work packages, including:
- business plan,
- user requirements,
- market assessment,
- system design,
- the OTELO client demonstrator.
The project has developed a first prototype client, which is promoting use of online translation tools and stimulating further requirements and refinements.
Market Assessment and User Survey
The OTELO user survey was conducted with the aid of a questionnaire tool 'Quest' allowing a executable to be sent out or used online and automated evaluation of replies. The detailed survey with 160 questions received 107 replies. The results showed that over 95% of organisations use non-automated methods and 43% use machine translation, 46% using translation memories. Only 9% use online services, but 28% plan to do so soon. Nearly 60% of respondents wanted high quality translation services, whilst 51% were satisfied with some availability of output. The greatest demand (over 92%) was for access to terminology databases.
Recent analyses estimated annual translation volume in Western Europe alone at between 400-450 million pages. Estimates for the year 2000 go as high as 1,800 million pages.
The OTELO market assessment and user surveys produced a view of the situation in translation technology:
- a gap exists; there is an increasing demand for translation, but no corresponding usage of Natural Language Processing (NLP),
- there is an opportunity; 85% of the users surveyed wanted to apply more translation technology.
Most interesting were the barriers to adoption of language processing tools that were highlighted:
- the low quality of texts, especially in the case of raw machine translation output,
- fragmentation; incompatible applications cover only part of the user's needs,
- workflow; most NLP applications do not support the overall translation process.
To overcome these problems OTELO integrates a variety of existing technology in a accessible NLP 'pool', provides process and groupware support, and improves quality indirectly through provision of productivity tools and a synergy effect of combined technology.
The OTELO Client
This OTELO Client is the main interface for all user interaction with OTELO systems:
Machine Translation Systems: offering three competitive machine translation systems, Logos, Metal, and PaTrans, each one accessible by means of a common OTELO format.
Lexical Resource Management: providing a general interface and management capability for lexical resource handling; this includes support for terminology databases, as well as complete facilities for the lexicons of the various machine translation systems accessed via OTELO.
Productivity Tools: offering a range of tools that serve as adjuncts to the basic task of translating, e.g., pre- and postediting tools, pattern matchers.
Information Management: for distributed handling through groupware systems of documents to be translated; this includes the handling of related translation jobs or other processes, and all of the necessary control information, including job-specific MT terminology.
Central Database: lexical resources will be distributed across a number of sites (e.g., the different MT lexicons). The central database will maintain the consistency and integrity of these data by keeping track of the information stored at each location and providing cross-checking and control mechanisms.
Text-Handling: will include various capabilities, including the extraction of text portions and the reading and writing of text elements.
Job Management: will primarily entail monitoring and controlling the network and the jobs that flow through it.
Currently the client (freely available from the project web site) combines Logos MT and IBM Translation Manager with its own built-in productivity enhancements such as HyperLink editor.
The Way Ahead
Distribution of the OTELO client is bringing further requirements and feedback to the consortium for incorporation into subsequent releases. The OTELO client and supporting systems are continually being improved. In the short term the system will be improved with greater support for 'occasional users', integration of translation memories, filters for word processor formats and support for terminology exchange between different MT systems.
The consortium has also attracted two additional organisations, IBM who will bring their existing translation memory product to OTELO, and S&D who will extend the user base of the project, particularly with a focus on occasional use of OTELO.
Industrial production still depends to a large extent on a manual translation process. However, it is the clear experience of the user members of the consortium, who are major industrial users of translation services and NLP technology, that the linguistic potential of existing NLP tools would be sufficient for productive use, if this potential could only be exploited to its full extent. But NLP applications do not always meet the standards of user friendliness and ergonomics that users expect today. Worse, it is unacceptable, as present systems require, to maintain separate and incompatible terminology pools for almost each NLP product. Finally, NLP is focussed on individual tasks, but in a world of distributed authoring and translation, literally spanning across continents, process support is an essential requirement that no NLP tool really meets.
Progress and results
The OTELO project aims to address these issues by developing a complete, integrated environment for NLP technology. OTELO will provide integrated, transparent, network access to high-end NLP tools in the form of Machine Translation, Translation Memory, and Terminology Management systems, as well as to conventional human translation resources. Moreover, OTELO will define a common format for the creation, storage, and maintenance of lexical resources, allowing these to be shared among heterogenous, multi-vendor, products. Finally, OTELO aims to bring the benefits of groupware computing to the translation process, by providing a client integrating access to network based services with local work flow and information management services, and also incorporating its own productivity tools.
The OTELO consortium includes major industrial users of NLP technology as well as major European suppliers of NLP products, and of linguistic services in general. The partners have a common strategic aim in making existing NLP technology smarter to use and therfore more acceptable as a standard translators' tool, leading to an overall increase in productivity throughout the translation community. Consequently, the translation network established by OTELO is intended to be open, so that any vendor of services possessing the necessary network infrastructure can freely offer services to connected users. To this end, billing, accounting and security mechanisms necessary for commercial transactions are taken into account from the earliest requirement stages. It is intended, moreover, to distribute the OTELO client freely via the standard distribution channels of LOTUS Development, one of the major world developers of software products.
Funding SchemeCSC - Cost-sharing contracts
2300 Copenhagen S