A Pragmatic Concept of Theme and Rheme for Machine Translation

Christa HauenschildCOLING 1982

TL;DR

This paper proposes a pragmatic approach to theme-rheme structure (TRS) for machine translation, arguing that preserving the underlying communicative order of sentence elements is key to accurate translation. They suggest a simplified procedure that focuses on maintaining this order, offering a potentially more efficient solution for machine translation systems.

What Problem Does This Paper Address?

Machine translation systems often struggle with accurately conveying the nuances of meaning that are embedded in the structure of a sentence, specifically its theme-rheme structure (TRS). The theme is typically the "given" information, while the rheme is the "new" information. Preserving this distinction is crucial for maintaining textual coherence and ensuring the translated text is easily understood. However, directly computing the TRS for every sentence in a translation process is computationally complex and prone to ambiguity, especially given the inherent ambiguity in written language regarding TRS. This paper aims to address this challenge by proposing a more pragmatic and simplified method for preserving TRS in machine translation.

Methods

The authors propose a simplified procedure for preserving the theme-rheme structure (TRS) in machine translation. Instead of exhaustively computing the TRS for every sentence, their approach focuses on preserving the underlying communicative order of main sentence elements. This underlying order is conceived as a strategy of communication, where elements carrying "new information" are placed at the end, following elements that connect to the preceding context. The procedure involves deriving this underlying order from the surface order of main constituents, with transformations applied for special cases like cleft-sentences. The synthesis in the target language then aims to yield the corresponding surface order with a minimal set of transformations. This procedure is controlled by semantic factors closely related to TRS, such as anaphoric relations, the definiteness of noun-phrases, and the scope of quantifiers and operators, which must be kept constant independently.

Datasets

Not discussed in the paper.

Key Insights and Ideas

  • TRS as a Communication Strategy: The paper frames the theme-rheme structure (TRS) not just as a linguistic feature, but as a fundamental strategy of communication that influences how information is presented and processed.
  • Pragmatic Solution for MT: By viewing TRS pragmatically, the authors suggest a more efficient approach to machine translation, moving away from complex, exhaustive analysis towards preserving the communicative flow.
  • Underlying Order is Key: The central hypothesis is that preserving the underlying communicative order of main sentence elements is often sufficient to maintain the TRS, simplifying the translation process.
  • "Linearity Principle" in Texts: The authors observe that texts, particularly those for international organizations, tend to follow a "linearity principle" where sentences start with contextual information and end with new information, making the underlying order more predictable.
  • Semantic Factors as Control: Key semantic factors like anaphoric relations and definiteness are identified as crucial for controlling the preservation of TRS, ensuring that the pragmatic intent is maintained.

Key Results

The paper does not present quantitative results or specific performance metrics. Instead, it outlines a theoretical framework and a proposed procedure for machine translation. The key "result" is the conceptual shift towards a pragmatic approach to TRS and the proposal of a simplified method that relies on preserving the underlying communicative order of sentence elements. The authors suggest that this approach is "very often possible" to implement for certain languages (German, French, Russian) and text types.

Limitations and Future Work

The authors acknowledge that the validity of their hypothesis is limited to certain languages (German, French, and Russian) and specific types of texts, particularly those aiming at optimal information transmission (e.g., reports for international organizations). They also note that their simplified procedure is not intended to be a complete analysis of TRS but rather a pragmatic solution for machine translation. Future work would likely involve implementing and evaluating this proposed procedure in actual machine translation systems and exploring its applicability to a wider range of languages and text genres. The paper also hints at further arguments that could be added to the list of reasons for preserving TRS, such as the relation between TRS and the scope of operators, suggesting areas for deeper investigation.

Summary updated

BibTeX
@inproceedings{hauenschild1982pragmatic,
  author = {Christa Hauenschild},
  title = {{A Pragmatic Concept of Theme and Rheme for Machine Translation}},
  booktitle = {International Conference on Computational Linguistics},
  year = {1982},
}
Figure