Cooperative File Improvement and Use of a Computer-Based Chinese/English Dictionary

Jim MathiasCOLING 1973

TL;DR

This paper introduces the CETA File Improvement System, a novel, cooperative approach to building and refining a computer-based Chinese-English dictionary. By leveraging a distributed network of government and academic linguists, CETA has created a dynamic, machine-readable dictionary that can be updated continuously, overcoming the slow pace of traditional dictionary development.

What Problem Does This Paper Address?

Traditional dictionary creation and updating are incredibly slow, often taking decades for new editions. This makes it challenging to keep dictionaries current with evolving language, especially for specialized fields like technical translation. Furthermore, many existing dictionaries, even those with extensive entries, are not in a machine-readable format, making them difficult to update, consolidate, or integrate into computer-aided translation systems. The CETA group aimed to solve this by developing a computer-based Chinese-English dictionary that could be continuously improved and enriched through a collaborative effort.

Methods

The core of CETA's approach is the CETA File Improvement System. This system is built on a cooperative international process involving government agencies, universities, and individual linguists. The process begins with a computer-generated Chinese-English dictionary file, initially containing 90,000 terms. This file is then distributed to a network of linguists and language specialists who are tasked with reviewing existing entries and suggesting additions or corrections.

Key aspects of the method include:

  • Machine-Readable Format: The dictionary is stored in a computer file, allowing for easy manipulation, sorting, correction, and addition of data. This also enables the computer generation of Chinese characters, eliminating the need for traditional typesetting.
  • Distributed Review: Linguists are provided with instructions and a structured process for submitting improvements. They receive updated versions of the dictionary as a reward for their contributions.
  • Cyclical Refinement: The improvement process operates in cycles. Each cycle involves progressively more rigorous standards of review, starting with gross error and duplication checks and moving towards enrichment of lexical content, addition of grammatical information, and stylistic labels.
  • Automated Processing: The system uses computers to process suggested changes, generate prooflists for manual review, and update the master file. This automation significantly speeds up the update process.
  • On-line Computer Aid: The dictionary file is also intended to serve as the core for an on-line computer-aided translation system, demonstrating the practical application of the machine-readable format.

Datasets

The paper describes the creation and ongoing refinement of a 90,000-term Chinese-English dictionary file. This file is the primary dataset, and it is continuously being updated with suggested improvements. The paper does not mention any specific external datasets used for training or evaluation, as the focus is on the collaborative improvement of their own dictionary.

Key Insights and Ideas

  • The Power of Cooperative Improvement: The most significant insight is that a large-scale, complex task like dictionary building can be effectively tackled through voluntary, distributed cooperation. This model democratizes dictionary development and accelerates its pace.
  • Machine-Readability as a Foundation: Creating a machine-readable dictionary is crucial for enabling continuous updates and integration into modern technological tools like computer-aided translation systems.
  • Incentivizing Contribution: Providing participants with updated versions of the dictionary and the satisfaction of contributing to a valuable resource serves as an effective incentive for their voluntary efforts.
  • Dynamic Dictionaries: The concept of a "living" file that can be constantly changed and enriched offers a stark contrast to static, periodically reissued dictionaries.
  • Bridging the Gap: CETA's work bridges the gap between academic linguistic research and practical technological applications, demonstrating how computational linguistics can directly support translation and information access.

Key Results

  • Over 51,000 Suggested Improvements: As of the paper's writing, over 51,000 suggested improvements have been submitted and evaluated, with many awaiting update.
  • Continuous Updates: The File Improvement System allows for updates to the dictionary file within a matter of months, a dramatic improvement over the typical 10-year building and 20-year reissue cycles of traditional dictionaries.
  • Growing Dictionary Size: Due to more additions than deletions, the new file is expected to be slightly larger than the original 90,000 terms.
  • Capability for Character Display: The character vector file has been updated to support the drawing of approximately 10,500 Chinese characters, with ongoing updates planned.
  • Prototype Computer Aid: Prototype computer aid functions have been developed to illustrate how the dictionary file can be used interactively by translators.

Limitations and Future Work

  • Refined Data Base for Testing: CETA hopes to test the system further using a refined data base for evaluation of its potential for shared access by a wider government and academic community.
  • Further Enrichment: Future cycles of file review will emphasize further enrichment of lexical content, addition of grammatical information, incorporation of restrictive and stylistic labels, and identification of agglutinated phrases.
  • Inclusion of Romanization and Telecode: Future printings of the dictionary manuscript will include Pin Yin romanization with tone, telecode numbers, and customary English gloss and source information.
  • Machine Readability of Existing Dictionaries: The paper notes that many good dictionaries are not machine-readable, posing a challenge for consolidation and improvement. CETA's method addresses this by creating a machine-readable foundation from the outset.

Summary updated

BibTeX
@inproceedings{mathias1973cooperative,
  author = {Jim Mathias},
  title = {{Cooperative File Improvement and Use of a Computer-Based Chinese/English Dictionary}},
  booktitle = {International Conference on Computational Linguistics},
  year = {1973},
}
Figure