[Paper Review] Using Multiple Sources of Information for Constraint-Based Morphological Disambiguation

Gökhan Tür|ArXiv.org|Jul 30, 1996

Natural Language Processing Techniques3 citations

TL;DR

This paper presents a constraint-based morphological disambiguation system for agglutinative languages like Turkish, integrating hand-crafted linguistic rules, unsupervised learning from training corpora, and statistical information from target texts. By combining these sources, the system achieves 96–97% recall, 93–94% precision, and less than 1.03 ambiguous parses per token, with under 1% of tokens remaining unknown through a secondary morphological processor.

ABSTRACT

This thesis presents a constraint-based morphological disambiguation approach that is applicable to languages with complex morphology--specifically agglutinative languages with productive inflectional and derivational morphological phenomena. For morphologically complex languages like Turkish, automatic morphological disambiguation involves selecting for each token morphological parse(s), with the right set of inflectional and derivational markers. Our system combines corpus independent hand-crafted constraint rules, constraint rules that are learned via unsupervised learning from a training corpus, and additional statistical information obtained from the corpus to be morphologically disambiguated. The hand-crafted rules are linguistically motivated and tuned to improve precision without sacrificing recall. In certain respects, our approach has been motivated by Brill's recent work, but with the observation that his transformational approach is not directly applicable to languages like Turkish. Our approach also uses a novel approach to unknown word processing by employing a secondary morphological processor which recovers any relevant inflectional and derivational information from a lexical item whose root is unknown. With this approach, well below 1% of the tokens remains as unknown in the texts we have experimented with. Our results indicate that by combining these hand-crafted, statistical and learned information sources, we can attain a recall of 96 to 97% with a corresponding precision of 93 to 94%, and ambiguity of 1.02 to 1.03 parses per token.

Motivation & Objective

• To address the challenge of morphological disambiguation in agglutinative languages with rich inflectional and derivational morphology, such as Turkish.
• To develop a system that combines multiple information sources—hand-crafted rules, learned constraints, and corpus statistics—for improved disambiguation performance.
• To reduce the number of unknown words in morphological processing by employing a secondary processor for root-inferred parsing.
• To achieve high recall and precision while minimizing ambiguity in morphological analysis.
• To extend the applicability of constraint-based methods beyond Brill’s transformational approach, which is not directly suitable for Turkish.

Proposed method

• The system uses linguistically motivated, hand-crafted constraint rules to improve precision without reducing recall.
• It applies unsupervised learning to extract additional constraint rules from a training corpus, enhancing coverage and adaptability.
• Statistical information from the target corpus is integrated to guide disambiguation decisions.
• A secondary morphological processor analyzes unknown words by inferring root forms and recovering derivational and inflectional markers.
• The system combines all three sources—hand-crafted rules, learned constraints, and statistical data—within a constraint-based inference framework.
• The disambiguation process resolves ambiguity by ranking possible morphological parses based on the combined evidence from all sources.

Experimental results

Research questions

• RQ1 Can a constraint-based system effectively handle morphological disambiguation in agglutinative languages like Turkish, where inflectional and derivational morphology is highly productive?
• RQ2 How does combining hand-crafted rules, learned constraints, and statistical information improve disambiguation performance compared to single-source approaches?
• RQ3 To what extent can a secondary morphological processor reduce the number of unknown tokens in morphological analysis?
• RQ4 Does the proposed method achieve high recall and precision while maintaining low ambiguity in morphological parses?
• RQ5 Is the constraint-based approach viable for Turkish when transformational methods like Brill’s are not directly applicable?

Key findings

• The system achieves a recall of 96–97% and a precision of 93–94% on morphological disambiguation tasks for Turkish.
• The ambiguity rate is reduced to 1.02–1.03 parses per token, indicating high disambiguation confidence.
• Fewer than 1% of tokens remain unknown after processing, thanks to the secondary morphological processor.
• The integration of hand-crafted rules, learned constraints, and statistical data leads to a significant improvement in disambiguation accuracy.
• The approach successfully overcomes limitations of transformational methods like Brill’s, which are not directly applicable to Turkish’s morphological structure.
• The system demonstrates robust performance on real-world Turkish text with high morphological complexity.