Leveraging Reusability: Cost-Effective Lexical Acquisition for Large-Scale Ontology Translation

Proceedings of the 21st International Conference on Computational Linguistics and 44th Annual Meeting of the ACL, pages 945–952, Sydney, July 2006. c⃝2006 Association for Computational Linguistics Leveraging Reusability: Cost-effective Lexical Acquisition for Large-scale Ontology Translation G. Craig Murray Bonnie J. Dorr Jimmy Lin Institute for Advanced Computer Studies University of Maryland {gcraigm,bdorr,jimmylin}@umd.edu Jan Hajič Pavel Pecina Institute for Formal and Applied Linguistics Charles University {hajic,pecina}@ufal.mff.cuni.cz Abstract Thesauri and ontologies provide impor- tant value in facilitating access to digital archives by representing underlying prin- ciples of organization. Translation of such resources into multiple languages is an important component for providing multilingual access. However, the speci- ficity of vocabulary terms in most on- tologies precludes fully-automated ma- chine translation using general-domain lexical resources. In this paper, we pre- sent an efficient process for leveraging human translations when constructing domain-specific lexical resources. We evaluate the effectiveness of this process by producing a probabilistic phrase dic- tionary and translating a thesaurus of 56,000 concepts used to catalogue a large archive of oral histories. Our experi- ments demonstrate a cost-effective tech- nique for accurate machine translation of large ontologies. 1 Introduction Multilingual access to digital collections is an important problem in today’s increasingly inter- connected world. Although technologies such as cross-language information retrieval and ma- chine translation help humans access information they could not otherwise find or understand, they are often inadequate for highly specific domains. Most digital collections of any significant size use a system of organization that facilitates easy access to collection contents. Generally, the or- ganizing principles are captured in the form of a controlled vocabulary of keyword phrases (de- scriptors) representing specific concepts. These descriptors are usually arranged in a hierarchic thesaurus or ontology, and are assigned to collec- tion items as a means of providing access (either via searching for keyword phases, browsing the hierarchy, or a combination both). MeSH (Medi- cal Subject Headings) serves as a good example of such an ontology; it is a hierarchically- arranged collection of controlled vocabulary terms manually assigned to medical abstracts in a number of databases. It provides multilingual access to the contents of these databases, but maintaining translations of such a complex struc- ture is challenging (Nelson, et al, 2004). For the most part, research in multilingual in- formation access focuses on the content of digital repositories themselves, often neglecting signifi- cant knowledge that is explicitly encoded in the associated ontologies. However, information systems cannot utilize such ontologies by simply applying off-the-shelf machine translation. Gen- eral-purpose translation resources provide insuf- ficient coverage of the vocabulary contained within these domain-specific ontologies. This paper tackles the question of how one might efficiently translate a large-scale ontology to facilitate multilingual information access. If we need humans to assist in the translation proc- ess, how can we maximize access while mini- mizing cost? Because human translation is asso- ciated with a certain cost, it is preferable not to incur costs of retranslation whenever compo- nents of translated text are reused. Moreover, when exhaustive human translation is …

Content truncated to 500 words