A Logic-Based Government-Binding Parser for Mandarin Chinese

Hsin-Hsi ChenCOLING 1990

TL;DR

This paper introduces a novel logic-based grammar formalism inspired by Government-Binding (GB) Theory to parse the highly flexible and context-sensitive Mandarin Chinese language. By integrating GB principles like movement transformations and binding theory into a declarative logic grammar, the authors aim to overcome the challenges of arbitrary word order and complex grammatical structures in Chinese.

What Problem Does This Paper Address?

Mandarin Chinese presents significant challenges for natural language processing due to its extreme flexibility and reliance on context. Unlike languages with more rigid word orders and explicit case marking, Chinese allows for a wide variety of sentence structures to convey the same meaning. This is particularly evident in topic-comment structures, where constituents can be moved to the beginning of a sentence, leading to complex binding relationships between elements and their traces. Accurately parsing these sentences requires handling arbitrary movement, detecting grammatical errors proactively, and resolving index assignments (similar to pronoun resolution) which are heavily context-dependent. Existing parsing approaches struggle to capture this fluidity and complexity efficiently.

Methods

The paper proposes a Government-Binding based Logic Grammar (GBLG) formalism. This approach leverages the principles of Government-Binding Theory, a generative grammar framework, and integrates them into a logic programming paradigm. Key aspects of their method include:

  • Formal Grammar Specification: The grammar is defined formally using a 6-tuple structure, including lexical terminals, non-terminals (phrasal, virtual, and movement), bounding non-terminals, logic connectives, and production rules.
  • Movement Transformations: The formalism explicitly handles movement using "movement non-terminals" (e.g., A <<< B for leftward movement, B >>> A for rightward movement). This allows for arbitrary numbers of movement steps and flexible placement of moved constituents.
  • Significance Condition: A crucial innovation is the "significance" condition for production rules. This condition ensures that movement transformations are properly licensed by the presence of a corresponding virtual non-terminal (trace) in the grammar body, effectively embedding the c-command condition and preventing illegal movements.
  • Binding Theory Integration: The paper incorporates Binding Theory principles (A, B, and C) to handle index assignment and pronoun resolution. This is implemented as a post-processing routine that traverses the parse tree to determine co-index relationships.
  • Declarative Semantics: The logic-based nature of the grammar allows for clear, declarative semantics, making it easier to reason about and potentially translate parsed structures into logical forms.

Datasets

Not discussed in the paper. The paper focuses on defining a formalism and demonstrating its application to Chinese sentence structures, rather than empirical evaluation on large datasets.

Key Insights and Ideas

  • GB Theory as a Natural Fit for Flexible Languages: The paper argues that GB Theory's core concepts, particularly the "move-α" rule and universal principles, are well-suited to model the flexibility of languages like Mandarin Chinese.
  • Significance Condition for Proactive Error Detection: The proposed "significance" condition for production rules is a powerful mechanism. It acts as a built-in grammar checker, ensuring that movement transformations are syntactically sound before parsing begins, thus preventing many potential errors.
  • Unified Treatment of Movement: The formalism allows for a uniform treatment of both leftward and rightward movement, as well as multiple movement steps within a single rule, which is essential for capturing complex Chinese sentence structures.
  • Binding Theory for Index Assignment: Integrating Binding Theory directly into the parsing process provides a principled way to handle the challenging task of index assignment, which is critical for semantic interpretation and pronoun resolution in context-dependent languages.

Key Results

The paper presents a formal grammar formalism and demonstrates its ability to represent various complex Chinese sentence structures, including:

  • Topic-comment structures with single and multiple topics.
  • Ba-constructions and Bei-constructions.
  • Relative clause constructions.
  • Appositive clause constructions.
  • Serial verb constructions.

It also touches upon a simple pronoun resolution mechanism based on Binding Theory. An experimental Chinese parser is mentioned as running on a Vax-11/785 with Quintus Prolog, utilizing a lexicon of about 200 words and 150 production rules. The paper highlights the formalism's capability for uniform treatment of movement, arbitrary movement non-terminals, automatic grammar error detection, and clear declarative semantics.

Limitations and Future Work

  • Limited Empirical Evaluation: The paper focuses on the formalism and its theoretical capabilities. While an experimental parser is mentioned, detailed quantitative results or performance evaluations on extensive corpora are not presented.
  • Pronoun Resolution Complexity: The paper acknowledges that while Binding Theory helps identify "impossible pairs" for co-indexing, determining "possible pairs" often requires more semantic information than is currently incorporated.
  • Semantic Interpretation: While the formalism aims for declarative semantics and touches on logical form composition, a full-fledged semantic interpretation system is beyond the scope of this paper.
  • Future Work: The authors suggest that their GB approach is also suitable for composing logical formulae from syntactic structures and can be extended to handle SVO-SOV inversion in logical forms. Further research into incorporating more semantic knowledge for robust pronoun resolution and developing a comprehensive semantic interpretation system is implied.

Summary updated

BibTeX
@inproceedings{chen1990logic,
  author = {Hsin-Hsi Chen},
  title = {{A Logic-Based Government-Binding Parser for Mandarin Chinese}},
  booktitle = {International Conference on Computational Linguistics},
  year = {1990},
}
Figure