Unlocking the complexity of phrasal composition: An interplay between semantic features and linguistic relations

Area of Science:

• Cognitive Science
• Computational Linguistics
• Psycholinguistics

Background:

• Existing theories on conceptual composition in language are fragmented, bridging formal semantics and neurolinguistics.
• Formal semantic theories often overlook conceptual content, while neurolinguistics highlights its importance in word composition.
• The relationship between these theoretical approaches to language interpretation remains unclear.

Purpose of the Study:

• To investigate the computational operations underlying conceptual composition in phrases.
• To determine how semantic content and linguistic relations influence these operations.
• To reconcile formal semantic and neurolinguistic perspectives on word combination.

Main Methods:

• Collected human ratings for experiential semantic features of individual words and six types of phrases.
• Examined phrase types varying in modification, argument-saturation, and context sensitivity.
• Compared the efficacy of different computational rules in predicting phrase ratings from word ratings.

Main Results:

• Composition operations are not uniform but vary based on feature type and linguistic relations.
• In adjective-noun phrases, addition merges attention features, while color features depend on the adjective.
• Social features in verb-noun phrases are guided by the verb; in noun-noun phrases, multiplication combines noun features.

Conclusions:

• Conceptual composition is context-dependent, influenced by both the meaning of words and their grammatical structure.
• Mathematical operations for merging semantic information differ across feature types and linguistic constructions.
• Findings suggest a unified model where formal linguistic structures interact with semantic content during interpretation.