[Paper Review] A framework for lexical representation

José Miguel Goñi-Menoyo, José Carlos González|arXiv (Cornell University)|Jul 6, 1995

Natural Language Processing Techniques2 references3 citations

TL;DR

This paper presents a unification-based lexical framework for highly inflected languages, using lemma-based encoding to generate allomorph-indexed dictionaries automatically. It enables efficient morphological processing through a formalism and supporting software tools, with demonstrated application in a 15th International Conference on Artificial Intelligence and Language Engineering.

ABSTRACT

In this paper we present a unification-based lexical platform designed for highly inflected languages (like Roman ones). A formalism is proposed for encoding a lemma-based lexical source, well suited for linguistic generalizations. From this source, we automatically generate an allomorph indexed dictionary, adequate for efficient processing. A set of software tools have been implemented around this formalism: access libraries, morphological processors, etc.

Motivation & Objective

• To design a formalism for encoding lexical information in highly inflected languages using lemma-based representation.
• To enable linguistic generalizations through a structured, formalized lexical source.
• To automatically generate allomorph-indexed dictionaries from the lexical source for efficient processing.
• To implement a suite of software tools, including access libraries and morphological processors, around the formalism.
• To support efficient and scalable morphological analysis in languages like Spanish and other Romance languages.

Proposed method

• A unification-based formalism is defined to represent lexical entries using lemmas and morphological features.
• The system encodes lexical generalizations by grouping inflected forms under a common lemma with feature structures.
• Allomorph indexing is automatically generated from the lemma-based representation to support fast lookup.
• A compiler-like process transforms the high-level lexical source into optimized, allomorph-indexed dictionary structures.
• Software tools, including access libraries and morphological processors, are built on top of the formalism for runtime use.
• The framework is evaluated in the context of the AI95 conference, demonstrating practical applicability in language engineering.

Experimental results

Research questions

• RQ1 How can a unified formalism represent lexical entries in highly inflected languages while supporting linguistic generalizations?
• RQ2 What is the most effective way to encode morphological variation using lemma-based structures?
• RQ3 Can automatic generation of allomorph-indexed dictionaries improve processing efficiency in morphologically rich languages?
• RQ4 How can a software toolchain be built around a formal lexical representation to support real-world NLP applications?
• RQ5 What are the practical benefits of using unification-based feature structures in lexical modeling for inflected languages?

Key findings

• The framework successfully generates allomorph-indexed dictionaries from a lemma-based lexical source, enabling efficient morphological access.
• The unification-based formalism allows for compact and generalizable representation of inflected forms across highly inflected languages.
• The implemented software tools, including access libraries and morphological processors, demonstrate practical integration and usability.
• The approach was validated at the AI95 conference, where it was presented as a working solution for language engineering tasks.
• The system supports linguistic generalizations by encoding morphological variation through feature structures rather than listing all inflected forms.
• The formalism enables scalable and maintainable lexical representation, reducing redundancy in morphological data.