An Intelligent Multi-Dictionary Environment

Gabor ProszekyCOLING 1998

TL;DR

This paper introduces MoBiDic, an intelligent, extendible multi-dictionary environment designed to revolutionize translator workflows. It addresses the limitations of traditional single-dictionary lookups by enabling simultaneous access to numerous dictionaries, supporting various media types, and leveraging morphological analysis for accurate lemma retrieval, ultimately aiming to make translation more efficient and comprehensive.

What Problem Does This Paper Address?

The paper tackles the inefficiencies and limitations faced by translators relying on traditional, single-source dictionaries. While computational tools have advanced, automatic translation remains imperfect, necessitating robust support for human translators. Existing electronic dictionaries often merely digitize paper versions, failing to fully leverage modern linguistic technology. A key challenge is the sheer volume and diversity of lexical resources required for accurate translation, coupled with the difficulty of finding multi-word expressions or idioms when the user doesn't know the exact headword. Furthermore, the need to manage and integrate personal glossaries and user-defined dictionaries within a unified environment is a significant hurdle.

Methods

MoBiDic employs a client-server architecture with distinct linguistic and dictionary servers. Its core innovation lies in its multi-dictionary access, allowing users to query an unlimited number of dictionaries simultaneously with a single search. A crucial component is its morphological analysis and lemmatization capability, which ensures that inflected word forms are automatically converted to their base lemmas before dictionary lookup. This prevents missed entries due to variations in word form.

The system also supports multimedia dictionaries (text, speech, pictures) and can dynamically find words and expressions within parallel corpora, providing real-world examples. MoBiDic is designed with an open architecture and a well-defined API, facilitating integration with other software modules and user-defined resources. For handling legacy dictionary data, a semi-automatic conversion system to SGML is used to enable structured access.

Datasets

Not explicitly discussed in the paper, beyond mentioning that the lexicographic basis is supplied by various publishing houses, with licenses to "almost 50 dictionaries already published in paper format." The system also supports user-created dictionaries and glossaries.

Key Insights and Ideas

  • Simultaneous Multi-Dictionary Querying: The ability to query an unlimited number of dictionaries at once is a paradigm shift, offering a comprehensive view of lexical information in a single step.
  • Lemma-Based Lookup: By automatically finding the lemma of inflected words, MoBiDic overcomes a major limitation of simple string matching, significantly improving lookup accuracy.
  • Idiom and Multi-Word Expression Handling: The system's ability to find multi-word expressions even when the user doesn't know the headword is a significant advantage over traditional dictionaries.
  • Integration of User Dictionaries: Allowing translators to incorporate their own glossaries and user-defined dictionaries into the system fosters personalized and collaborative translation environments.
  • Open Architecture and API: This design choice is key to MoBiDic's extensibility, allowing it to integrate with other tools and adapt to future linguistic technologies.

Key Results

The paper highlights the functional capabilities of MoBiDic rather than presenting specific quantitative performance metrics. Key results include:

  • Simultaneous access to an unlimited number of dictionaries: Users can survey all available dictionaries (translations, explanations, synonyms, etc.) with a single query.
  • Support for various dictionary types: MoBiDic handles textual, speaking, and picture dictionaries.
  • Dynamic example provision from corpora: Translators can access examples from their own or others' translations.
  • Dissemination of user glossaries: Workgroups can share custom glossaries with defined access rights.
  • Commercial availability and testing: MoBiDic is presented as a research project that is also a commercially available tool, tested by thousands of end-users.

Limitations and Future Work

  • Semi-automatic SGML Conversion: While functional, the conversion process for legacy data is described as semi-automatic and involves manual correction of inconsistent entries.
  • Integration with Third-Party Software: While an API is provided, integration with other commercial products requires collaboration with their developers.
  • Future Work: The paper mentions that combining MoBiDic with a thesaurus (like MorphoLogic's Helyette) is planned for the future. The ongoing development of morphological components for more languages is also noted. The authors implicitly suggest that further advancements in linguistic technology will continue to be integrated into the MoBiDic environment.

Summary updated

BibTeX
@inproceedings{proszeky1998intelligent,
  author = {Gabor Proszeky},
  title = {{An Intelligent Multi-Dictionary Environment}},
  booktitle = {International Conference on Computational Linguistics},
  year = {1998},
}
Figure