“Ambiguity is pervasive at all levels of analysis. It has been, is, and is likely to remain the key problem in natural language processing.” (Gadzar 1993:161) This statement by Gerald Gadzar expresses the necessity to cope with the challenge of ambiguity resolution. As the phenomenon of ambiguity is widespread in human language, an interesting question would be: How could a machine be able to handle ambiguity while even humans have difficulties in solving such problems?
This paper will first define the phenomenon of ambiguity and explain the different types of it. An interesting aspect will be the effect of garden path sentences.
Table of Contents
0 Introduction
1 The phenomenon of Ambiguity
1.1 Ambiguity: Definition
1.2 Different Types of Ambiguity
1.3 Garden Path Sentences
2 Syntactic Parsing
2.1 The Garden Path Model
2.2 Constraint-based Model of Parsing
2.3 Reanalysis in Sentence Processing
3 Ambiguity Resolution
3.1 The Way Humans Handle Ambiguity
3.2 The Way Machines Handle Ambiguity
4 Conclusion
5 References
Objectives and Core Topics
This paper explores the fundamental challenge of ambiguity in natural language, specifically examining how both human cognitive processes and computational parsing systems attempt to resolve ambiguous utterances. It investigates the mechanics of syntactic structure analysis and compares human linguistic interpretation with current machine-based resolution techniques.
- Theoretical definitions of linguistic ambiguity and its various types.
- The psychological impact and function of garden path sentences.
- Mechanisms of syntactic parsing, including late closure and minimal attachment principles.
- Comparison of human processing versus deterministic and constraint-based machine models.
- Algorithmic approaches to ambiguity resolution in computer systems.
Excerpt from the Book
1.3 Garden Path Sentences
Humans sometimes tend to misinterpret sentences which are grammatically correct but hard to understand. These sentences “lead the hearer down a garden path” because they lead one in the wrong direction and are therefore called garden path sentences (Newell: 1993:62). The reader or the hearer of such a sentence initially concentrates on one meaning of the sentence which in the end, when the sentence is finished, makes no sense. Thus, after recognizing that one’s intuition about the meaning was wrong, one has to reparse the sentence in a different way (Wingfield 1998:239). Consider the following example:
(1) The horse raced past the barn fell.
The initial understanding of the sentence would be that it concerns an active sentence with raced as its main verb. But by reaching the end of it the parser discovers that the word fell does not fit to this interpretation. Therefore one has to go back and reparse the sentence in order to find out that raced past the barn is a relative clause with a passive participle indicating that not the horse but someone else raced the horse and that not raced but fell is the main verb [INT2]. Garden path sentences are not only obstructive but they can also be a useful tool, for example for jokes:
(2) Time flies like an arrow. Fruit flies like a banana.
Furthermore, psycholinguists have often used the garden path effect in order to show how a “language processor is sent into an incorrect parse” [INT3].
Summary of Chapters
0 Introduction: This chapter introduces ambiguity as a key problem in natural language processing and outlines the paper's goal of comparing human and machine-based resolution strategies.
1 The phenomenon of Ambiguity: This chapter defines ambiguity, categorizes its various types, and illustrates the psychological effect of garden path sentences on human language comprehension.
2 Syntactic Parsing: This chapter examines the cognitive and computational procedures of syntactic structure analysis, focusing on models like the garden path model and constraint-based approaches.
3 Ambiguity Resolution: This chapter evaluates techniques used by both humans and machines, specifically analyzing the ROBIE parser and neutral algorithmic parsing systems.
4 Conclusion: This chapter summarizes the findings and highlights the necessity for further research into reconciling machine performance with human-like ambiguity handling.
5 References: This section lists all academic sources used throughout the paper.
Keywords
Ambiguity, Natural Language Processing, Syntactic Parsing, Garden Path Sentences, Ambiguity Resolution, Psycholinguistics, Late Closure, Minimal Attachment, Constraint-based Model, ROBIE Parser, Word-sense Disambiguation, Computational Linguistics, Reanalysis, Dependency Tree, Phonological Coding.
Frequently Asked Questions
What is the fundamental problem addressed in this paper?
The paper focuses on the challenge of ambiguity in natural language, exploring why it remains a persistent difficulty in both human communication and computer-based language processing.
What are the primary thematic areas covered?
The core themes include the definition and categorization of ambiguity, the mechanisms of syntactic parsing, the cognitive impact of garden path sentences, and practical computational techniques for resolving structural ambiguities.
What is the central research question?
The research asks how machines can be developed to handle linguistic ambiguity effectively, given the inherent difficulties even humans face when interpreting complex or ambiguous sentences.
Which scientific methods are employed?
The study utilizes a comparative literature review, analyzing psycholinguistic theories alongside computational parsing algorithms and algorithmic modeling.
What does the main body discuss?
The main body discusses the mechanics of syntactic parsing (such as the garden path model), the process of reanalysis in sentence processing, and the implementation of specific parsers like ROBIE.
Which keywords characterize this work?
Key terms include ambiguity, syntactic parsing, garden path sentences, word-sense disambiguation, and computational linguistics.
What is the significance of the "window of limited parallelism"?
It represents the part of a sentence analyzed in multiple, competing ways by a parser until the system selects the most plausible syntactic structure.
How does the ROBIE parser function?
ROBIE is a deterministic parser that uses a dictionary to assign syntactic characteristics to words and selects parts of speech based on an ordered list of morphological features, such as word endings.
What role does prosodic information play in sentence processing?
According to the text, prosodic information (often facilitated by phonological coding/the "inner voice") is crucial for human readers to resolve local syntactic ambiguities in written language.
- Quote paper
- Stefanie Dietzel (Author), 2006, Ambiguity Handling: Human vs. Machine, Munich, GRIN Verlag, https://www.grin.com/document/133403
