Core Concepts
Ambiguity types in natural language pose challenges for NLP systems, requiring a nuanced approach for resolution.
Abstract
1. Introduction
- Ambiguity enhances communication efficiency.
- NLP systems may struggle with human-level ambiguity understanding.
- Different ambiguity types require distinct resolution approaches.
2. Background
- Refinement and extension of AMBIENT's ambiguity categories.
- Linguistic definitions influence the taxonomy but are not exhaustive.
- Taxonomy aids in meaningful splits for detailed assessments.
3. Types of Ambiguity
3.1 Lexical Ambiguity
- Words with multiple meanings create ambiguity.
3.2 Syntactic Ambiguity
- Multiple grammatical structures lead to ambiguity.
3.3 Scopal Ambiguity
- Relative ordering of quantifiers causes ambiguity.
3.4 Elliptical Ambiguity
- Identity of elided words or phrases is ambiguous.
3.5 Collective/Distributive Ambiguity
- Plural expressions can be collective or distributive.
3.6 Implicative Ambiguity
- Sentences carry implicit meanings leading to ambiguity.
3.7 Presuppositional Ambiguity
- Presuppositions in sentences may be ambiguous.
3.8 Idiomatic Ambiguity
- Sequences interpreted as idioms but also literally possible.
3.9 Coreferential Ambiguity
- Pronoun reference ambiguity occurs in sentences.
3.10 Generic/Non-Generic Ambiguity
- Sentences can have both generic and non-generic readings.
3.11 Type/Token Ambiguity
-Terms ambiguous between type and token readings.
4. Conclusion
Taxonomy contributes to understanding NLP's handling of ambiguity, aiding in model assessment and dataset creation.
Stats
"Lexical ambiguity occurs when words have multiple possible meanings."
"Syntactic ambiguity happens when multiple grammatical structures are possible for a sequence of words."
"When a sentence contains multiple quantifiers or scopal expressions, their relative ordering may be ambiguous."