A Pragmatic Approach to Machine Translation From Chinese to English

Ching-Yi DoughertyCOLING 1969

TL;DR

This paper proposes a pragmatic approach to Chinese-to-English machine translation by focusing on efficient organization of linguistic data within a computer system. It outlines the necessary linguistic components and introduces methods for reducing parsing ambiguities and handling syntactic differences between the two languages, aiming for accurate translation without sacrificing computational efficiency.

What Problem Does This Paper Address?

The paper tackles the challenge of building an effective Chinese-to-English machine translation system, emphasizing the constraints of computer storage space and processing time. The core problem is how to organize and process the complex linguistic information of both languages to achieve accurate translations efficiently. This involves overcoming ambiguities inherent in natural language, particularly in Chinese, and bridging the structural differences between Chinese and English sentence and phrase constructions. The authors highlight that while machine translation was once considered impossible, advancements in computing and linguistic research make it a feasible goal.

Methods

The proposed system relies on storing five key types of linguistic data: a Chinese-to-English dictionary, Chinese syntactic rules, syntactic conversion rules from Chinese to English structure, English morphological rules, and the text to be translated.

  • Dictionary Enhancement: The dictionary needs to distinguish multiple meanings of Chinese lexemes based on context and syntactic function. This is achieved through syntactic codes and a "longest match" approach for dictionary lookup.
  • Ambiguity Reduction: To minimize parsing ambiguities, the authors refine syntactic codes by incorporating semantic elements (e.g., human vs. inanimate agents) and plurality. They also introduce hierarchical levels for noun and verb phrases, creating a more structured representation.
  • Syntactic Conversion Rules: These rules are developed by comparing Chinese and English sentence structures. If structures are similar, simple lexical substitution and English morphological rules suffice. For differing structures, rules for permutation, addition, or deletion of elements are applied. Examples illustrate conversions for verb phrases, passive constructions, and the Chinese particle "de."
  • "Longest Match" Principle: This principle is applied in dictionary lookup and parsing to resolve ambiguities by selecting the longest possible matching unit, thereby eliminating many potential misinterpretations.

Datasets

Not discussed in the paper. The paper focuses on the methodology and linguistic rules rather than specific datasets for training or evaluation.

Key Insights and Ideas

  • Efficiency is Paramount: The paper strongly emphasizes that for machine translation to be practical, it must be computationally efficient, balancing accuracy with limited computer resources.
  • Context is King for Lexical Meaning: The meaning of a Chinese word is not fixed but highly dependent on its syntactic function and surrounding context, requiring sophisticated disambiguation mechanisms.
  • Syntactic Structures are Key Bridges: The core of translation lies in understanding and converting syntactic structures, not just individual words. The paper details how to map Chinese grammatical structures to their English equivalents.
  • "De" is a Multifaceted Particle: The Chinese particle "de" is shown to have numerous meanings and functions (possessive, partitive, connective, relative pronoun), requiring a set of specific conversion rules to translate accurately.
  • Systematic Rules for Ambiguity: The paper demonstrates that by systematically refining syntactic codes and employing strategies like "longest match," many of the inherent ambiguities in Chinese can be managed.

Key Results

The paper doesn't present quantitative results in terms of translation accuracy or performance metrics. Instead, it outlines a framework and methodology for achieving accurate machine translation. The "results" are the proposed linguistic data structures, syntactic rules, and conversion mechanisms designed to handle the complexities of Chinese-to-English translation. The authors demonstrate through examples how their methods can resolve ambiguities and correctly translate various Chinese phrases and sentences.

Limitations and Future Work

  • Limitations:
    • The paper acknowledges that the goal of completely unambiguous parsing is "hard to achieve."
    • While the authors claim a pragmatic approach, the paper doesn't provide empirical validation of the system's performance.
  • The application of English morphological rules to machine translation is presented as a problem "yet to be solved," with solutions needing to be incorporated into the dictionary or syntactic conversion rules.
  • The handling of Chinese telegraphic codes and punctuation is mentioned, but the complexity of these aspects might pose further challenges.
  • Future Work:
    • Further refinements to syntactic codes and the introduction of more semantic elements are suggested.
  • The paper implies that with more sophisticated linguistic data, efficient programming languages, and faster computers, machine translation will become a reality in the near future.
  • The authors suggest that research into comparative constructions, locative phrases, and other minor structures would be beneficial.

Summary updated

BibTeX
@inproceedings{dougherty1969pragmatic,
  author = {Ching-Yi Dougherty},
  title = {{A Pragmatic Approach to Machine Translation From Chinese to English}},
  booktitle = {International Conference on Computational Linguistics},
  year = {1969},
}
Figure