English-Japanese Translation Through Case-Structure Conversion

Fujio Nishida, Shinobu Takamatsu, Hiroaki KurokiCOLING 1980

TL;DR

This paper proposes a novel approach to English-Japanese machine translation by converting English sentence structures into a language-independent "case structure" representation. This intermediate representation, inspired by Hornby's verb patterns and augmented with semantic roles, allows for more robust translation by handling syntactic and semantic ambiguities before generating Japanese sentences.

What Problem Does This Paper Address?

Translating between English and Japanese presents significant challenges due to their vastly different grammatical structures. English relies heavily on word order and prepositions, while Japanese utilizes postpositions and a more flexible word order, often implying grammatical roles through context. Traditional machine translation systems struggle with these differences, leading to awkward or inaccurate translations. This paper aims to bridge this gap by introducing a "case structure" as an intermediate representation that captures the underlying semantic relationships of a sentence, making it easier to convert into the target Japanese grammatical framework.

Methods

The core of this approach lies in the creation and manipulation of "case structures." The process begins with parsing English sentences using A.S. Hornby's verb patterns to identify syntactic categories and dependency relations. This forms the basis of a case structure, which is then enriched with semantic information.

Key methodological components include:

  • Case Structure Representation: A hierarchical system of "case labels" (e.g., AGENT, OBJECT, LOCATION) and "categories" (e.g., PHYSICAL, MENTAL, LIVING OBJECT) is used to represent the semantic roles of sentence constituents.
  • Hornby's Verb Patterns: These patterns are crucial for initial parsing, helping to resolve multi-syntactic and semantic ambiguities of words and phrases.
  • Semantic Information Integration: Beyond syntax, semantic information about words is incorporated to determine their roles and find appropriate Japanese equivalents.
  • Case Transformation: The paper details specific transformations to convert English case structures into Japanese equivalents, addressing issues like non-living subjects and clause ordering for better readability.
  • Idiomatic Expression Handling: A dedicated mechanism is described for processing idiomatic expressions, which often pose significant translation challenges.
  • Japanese Grammar Generation: Finally, the transformed case structures are fed into a Japanese grammar to generate the translated sentences.

Datasets

Not discussed in the paper. The paper focuses on the methodology and provides examples rather than reporting on experiments with specific datasets.

Key Insights and Ideas

  • The Power of Intermediate Representations: The paper strongly advocates for using a language-independent "case structure" as a crucial intermediate step in machine translation, especially for languages with significant structural differences like English and Japanese.
  • Bridging Syntax and Semantics: By combining syntactic parsing (via verb patterns) with semantic role labeling, the system can achieve a deeper understanding of sentence meaning, leading to more accurate translations.
  • Systematic Handling of Linguistic Differences: The paper demonstrates that many of the challenges in English-Japanese translation, such as the "non-living subject problem" and clause ordering, can be systematically addressed through case structure transformations.
  • Importance of Lexical Information: The authors highlight the necessity of rich word dictionaries that include verb patterns, semantic categories, and Japanese equivalents to support the parsing and transformation processes.

Key Results

The paper presents a methodology and illustrates its application through several examples. While quantitative results are not provided, the qualitative examples demonstrate the system's ability to:

  • Translate sentences with complex semantic roles.
  • Handle idiomatic expressions effectively.
  • Address grammatical differences like non-living subjects and clause ordering to produce more natural-sounding Japanese.
  • Resolve ambiguities that might arise from multi-syntactic and semantic categories of words.

Limitations and Future Work

  • Incomplete Case Grammar: The authors acknowledge that case grammar is still in an incomplete state for general-purpose sentence comprehension.
  • Need for Detailed Dictionaries: The effectiveness of the system relies heavily on comprehensive word dictionaries containing detailed information about verb patterns, categories, and Japanese equivalents.
  • Future Work: The authors suggest that for more precise and multi-lingual translation, there is a need for more standardized definitions of case structures and hierarchical category systems, along with their integration into word dictionaries. This would be particularly beneficial for the translation of technical papers.

Summary updated

BibTeX
@inproceedings{nishida1980english,
  author = {Fujio Nishida and Shinobu Takamatsu and Hiroaki Kuroki},
  title = {{English-Japanese Translation Through Case-Structure Conversion}},
  booktitle = {International Conference on Computational Linguistics},
  year = {1980},
}
Figure