TL;DR
This paper introduces LEDA (Lexicographical Data Base for German), a comprehensive system designed to revolutionize German lexicography. By integrating text corpora, existing dictionaries, and a structured results bank, LEDA aims to significantly improve the efficiency and accuracy of dictionary compilation, moving beyond traditional manual methods.
What Problem Does This Paper Address?
Creating comprehensive dictionaries, especially for a language as rich as German, is a monumental task. Traditionally, lexicographers relied on intuition, limited personal experience, and printed dictionaries, which often led to inconsistencies, gaps, and a lack of empirical grounding. The sheer volume of language use, including evolving meanings and new word formations, makes it nearly impossible for individual lexicographers to capture the full picture. Existing machine-readable dictionaries, while helpful, are often inconsistent in their structure and content. LEDA aims to bridge this gap by providing a systematic, computer-assisted approach that leverages vast amounts of textual data and existing lexicographical knowledge.
Methods
LEDA is built around three core components:
- Text Bank: This component provides access to large, machine-readable corpora of German text. The system allows lexicographers to search for specific words or phrases within these corpora, retrieve contextual examples, and analyze their frequency and distribution. A key tool here is the REFER query system, which enables interactive searches and provides detailed information about word occurrences.
- Dictionary Bank: This component aims to integrate existing machine-readable dictionaries. The goal is to create a unified resource where lexicographers can compare and contrast entries from various sources, identify discrepancies, and leverage established definitions and grammatical information.
- Result Bank: This is the central workspace for lexicographers. It provides a structured framework (a database) for entering and organizing new dictionary entries. This bank allows for the systematic description of lemmas, including their meanings, grammatical properties, and relationships with other words (e.g., synonyms, antonyms). It also facilitates the integration of textual examples from the Text Bank and information from the Dictionary Bank.
The system is designed for interactive use, allowing lexicographers to define their own search parameters and build dictionary entries collaboratively.
Datasets
The paper mentions the following datasets:
- Mannheim corpora of contemporary written language: Part of the Text Bank.
- Freiburg corpus of spoken language: Part of the Text Bank.
- East/West German newspaper corpus: Part of the Text Bank.
- Existing machine-readable dictionaries: Used to build the Dictionary Bank. The paper notes that the Bonn Institut für Kommunikationsforschung und Phonetik has been compiling a cumulative word data base for German using 11 existing dictionaries.
The paper states that by the end of 1985, texts of a total length of 20 million words will be available for the Text Bank.
Key Insights and Ideas
- The necessity of empirical data: The paper strongly advocates for dictionary compilation based on actual language use (corpora), highlighting that many existing dictionaries were created without this crucial foundation.
- The power of interactive systems: Moving from batch processing to interactive terminals significantly enhances the efficiency and flexibility for lexicographers.
- The challenge of semantic integration: While morphosyntactic descriptions can be standardized, achieving full automatic unification of semantic and pragmatic descriptions across dictionaries is difficult, emphasizing the need for human oversight.
- LEDA as a collaborative tool: The system is designed to support teams of lexicographers, ensuring consistency and facilitating the sharing of information.
- The potential for derived products: The structured nature of LEDA allows for the eventual creation of various specialized dictionaries or linguistic resources from the core data.
Key Results
The paper focuses on the design and planned implementation of LEDA rather than presenting quantitative results from a fully operational system. However, it highlights the following:
- Efficiency gains: The REFER system can complete searches in 3-5 seconds, regardless of the text bank size, allowing for rapid retrieval of textual examples.
- Comprehensive access: LEDA aims to provide access to textual samples, existing dictionary entries, and structured results, offering a holistic approach to lexicography.
- Planned operational status: A first version of LEDA incorporating all three components is planned to be ready for use in 1986.
- Pilot project: The "Manual of Hard Words" project will be the first to apply the Result Bank in its planning stage.
Limitations and Future Work
- Current stage of development: The paper describes a system under development, with a first version planned for 1986. The full capabilities and long-term impact are yet to be realized.
- Semantic unification challenges: As mentioned, achieving complete automatic unification of semantic and pragmatic descriptions across different dictionaries remains a significant hurdle.
- Data conversion: Transforming existing, heterogeneously coded corpora and dictionaries into the LEDA framework requires ongoing effort and program development.
- Future work: The paper implies continuous improvement of the REFER system, further development of lemmatized registers, and the integration of new dictionaries and corpora into the Dictionary Bank. The ultimate goal is to make LEDA accessible and applicable to a wide range of lexicographical projects.
Summary updated