Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Montague semantics and modifier consistency measurement in neural language models

Danilo S. Carvalho, Edoardo Manino, Julia Rozanova, Lucas Cordeiro, André Freitas

TL;DR

This work proposes a novel methodology for measuring compositional behavior in contemporary language embedding models by introducing three novel tests of compositional behavior inspired by Montague semantics, and results indicate that current neural language models do not behave according to the expected linguistic theories.

Abstract

This work proposes a novel methodology for measuring compositional behavior in contemporary language embedding models. Specifically, we focus on adjectival modifier phenomena in adjective-noun phrases. In recent years, distributional language representation models have demonstrated great practical success. At the same time, the need for interpretability has elicited questions on their intrinsic properties and capabilities. Crucially, distributional models are often inconsistent when dealing with compositional phenomena in natural language, which has significant implications for their safety and fairness. Despite this, most current research on compositionality is directed towards improving their performance on similarity tasks only. This work takes a different approach, introducing three novel tests of compositional behavior inspired by Montague semantics. Our experimental results indicate that current neural language models do not behave according to the expected linguistic theories. This indicates that current language models may lack the capability to capture the semantic properties we evaluated on limited context, or that linguistic theories from Montagovian tradition may not match the expected capabilities of distributional models.

Montague semantics and modifier consistency measurement in neural language models

TL;DR

This work proposes a novel methodology for measuring compositional behavior in contemporary language embedding models by introducing three novel tests of compositional behavior inspired by Montague semantics, and results indicate that current neural language models do not behave according to the expected linguistic theories.

Abstract

This work proposes a novel methodology for measuring compositional behavior in contemporary language embedding models. Specifically, we focus on adjectival modifier phenomena in adjective-noun phrases. In recent years, distributional language representation models have demonstrated great practical success. At the same time, the need for interpretability has elicited questions on their intrinsic properties and capabilities. Crucially, distributional models are often inconsistent when dealing with compositional phenomena in natural language, which has significant implications for their safety and fairness. Despite this, most current research on compositionality is directed towards improving their performance on similarity tasks only. This work takes a different approach, introducing three novel tests of compositional behavior inspired by Montague semantics. Our experimental results indicate that current neural language models do not behave according to the expected linguistic theories. This indicates that current language models may lack the capability to capture the semantic properties we evaluated on limited context, or that linguistic theories from Montagovian tradition may not match the expected capabilities of distributional models.
Paper Structure (28 sections, 11 equations, 2 figures, 8 tables)

This paper contains 28 sections, 11 equations, 2 figures, 8 tables.

Table of Contents

  1. Introduction
  2. The modifier phenomena
  3. Modification semantics
  4. Adjective types and interpretations
  5. Distributional questions
  6. Methodology
  7. Set-based phrase denotations
  8. Embedding-denotation analogy
  9. Testing intersectivity (single phrase)
  10. Testing intersectivity (phrase pairs)
  11. Testing non-subsectivity
  12. Experimentation and discussion
  13. Experimental setup
  14. Data collection
  15. Phrase Generation
  16. ...and 13 more sections

Figures (2)

  • Figure 1: Methodology for testing model consistency regarding the modifier phenomena in adjective-noun phrases (ANs). $m$ represents a language model and $L$ the regular language $(adj~~)+noun$. $E_{m,L}$ is calculated by averaging the no. of combinations $(a, n, \phi, p = an)$ where the inequalities hold over the vocabulary size.
  • Figure 2: Intersective and non-intersective set relations in adjective-noun phrase denotations.