[Paper Review] Finite-State Non-Concatenative Morphotactics

Kenneth R. Beesley, Lauri Karttunen|arXiv (Cornell University)|Jun 30, 2000

Natural Language Processing Techniques15 references12 citations

TL;DR

This paper introduces the compile-replace algorithm, a technique that re-applies a finite-state regular-expression compiler to its own output, enabling finite-state transducers to model non-concatenative morphotactics such as Malay full-stem reduplication and Arabic stem interdigitation. The method overcomes the fundamental limitation of traditional finite-state morphology by extending morphotactic description beyond concatenation, allowing full use of finite-state operations and significantly improving compilation efficiency and expressiveness in morphological analysis and generation.

ABSTRACT

Finite-state morphology in the general tradition of the Two-Level and Xerox implementations has proved very successful in the production of robust morphological analyzer-generators, including many large-scale commercial systems. However, it has long been recognized that these implementations have serious limitations in handling non-concatenative phenomena. We describe a new technique for constructing finite-state transducers that involves reapplying the regular-expression compiler to its own output. Implemented in an algorithm called compile-replace, this technique has proved useful for handling non-concatenative phenomena; and we demonstrate it on Malay full-stem reduplication and Arabic stem interdigitation.

Motivation & Objective

• To address the long-standing limitation of finite-state morphology in handling non-concatenative processes like reduplication and stem interdigitation.
• To overcome the dependency of traditional finite-state systems on concatenation as the sole morphotactic operation.
• To develop a general-purpose algorithm that extends the expressive power of finite-state transducers for morphological description.
• To demonstrate the effectiveness of the method on real-world languages such as Malay and Arabic.
• To reduce compilation time and improve system efficiency while maintaining full linguistic accuracy.

Proposed method

• The compile-replace algorithm re-applies the regular-expression compiler to its own output, transforming the compiled finite-state network into a modified but still finite-state structure.
• The method enables the use of arbitrary finite-state operations—such as intersection, reversal, and substitution—in morphotactic descriptions, moving beyond concatenation.
• A special marker symbol (e.g., 'XX') is inserted into strings during intermediate compilation steps to isolate and manipulate structural components.
• Replace transducers are composed with the network to substitute markers with complex operations like intersection and reversal.
• The technique is applied in two stages: first to identify candidate strings (e.g., palindromes or reduplicated forms), then to filter only those satisfying the required structural conditions.
• The approach is implemented in the Xerox finite-state framework using lexc, Replace Rules, and the compile-replace procedure, supporting both analysis and generation.

Experimental results

Research questions

• RQ1 Can finite-state morphology be extended to handle non-concatenative phenomena such as full-stem reduplication and stem interdigitation?
• RQ2 Is it possible to retain the efficiency and regularity of finite-state systems while supporting complex morphotactic operations beyond concatenation?
• RQ3 How can a regular-expression compiler be recursively applied to its own output to generate more expressive morphological networks?
• RQ4 What performance improvements can be achieved by applying the compile-replace technique to large-scale morphological systems?
• RQ5 Can the method be generalized to other non-concatenative processes beyond Malay and Arabic?

Key findings

• The compile-replace algorithm successfully models Malay full-stem reduplication by identifying and processing stem repetitions using marker insertion and finite-state operations.
• The method enables accurate modeling of Arabic stem interdigitation by allowing complex morpheme reordering and internal structure manipulation.
• Compilation time for the Arabic morphological system was reduced from hours to minutes after applying the compile-replace technique.
• The technique efficiently extracts all palindromes from a 25,000-word lexicon in just a few seconds, demonstrating its power for non-trivial string pattern recognition.
• The approach maintains the mathematical elegance and efficiency of finite-state systems while significantly expanding their linguistic expressiveness.
• The method is general-purpose and extensible, with demonstrated applicability to other non-concatenative morphological phenomena beyond the two case studies.