Computational Metaphorics

An Unsupervised Learning Method for Metaphor Detection (2000) .. by Gideon Mann

>>As a foray into a uncharted area of “computational metaphorics“, this project yielded a number of insights.

Computational Metaphorics: Prior Work

There is a growing literature on computational linguistic approaches to understanding metaphor. The earliest attempts to understand metaphors used Lakoff’s recent theory but have restricted themselves to toy-domains, simplifying the problem [Martin 92, Fass 91]. In these systems, an expert defines a series of literal relationships, and when the natural language system is presented with a relationship outside the normal bounds, it uses its knowledge base to fit this phrase into its semantic system. These systems are brittle and rely on extensive input from the programmer.

[Dolan 95] sketches an initial attempt to detect metaphors in free text. The technical report presents a method which uses a proprietary ontology developed from a machine readable dictionary to detect metaphors, finding example mappings from one to domain to another. No results are given, and it is unclear whether or not he has actually implemented the system.

Another attempt at metaphor recognition comes from recognition of lexical phrases used to indicate metaphorical relationships. [Ferrari 96] describes a way to use lexical tags (“like an” for example) to detect metaphors.

However, no previous work has demonstrated results on free-text metaphor detection.<<

= = =

Ocularcentrism, Inscription & the Figure of the Computer in Cognitive Theories of Language (2003) .. abstract by Chris Werry

>>I show how the computational metaphorics drawn on by Chomskyan and cognitive linguistics lead to a series of blindspots and limitations in their representation of language and mind.<<

Notes:

Computational metaphorics is the study of metaphor analytics, which involves the use of computational methods and algorithms to analyze and understand metaphors. This can include tasks such as identifying metaphors in text, classifying them based on their type or meaning, and extracting the underlying structure or meaning of the metaphor.

Computational metaphorics is a growing field, with many potential applications in a variety of different fields. For example, it could be used in natural language processing, to improve the ability of language processing algorithms to understand and interpret metaphors. It could also be used in fields such as psychology and cognitive science, to better understand the role of metaphor in human thought and language. Additionally, computational metaphorics could be used to develop tools for analyzing and understanding metaphors in large collections of text, such as books, articles, or social media posts, and to gain insights into the use of metaphor in different contexts and over time.

Computational or automated metaphor analysis is a subfield of metaphorology, which is the study of metaphor from a variety of different perspectives, including linguistic, cognitive, and cultural. Metaphorology is a broad and interdisciplinary field, and includes the study of metaphor from many different angles, including its role in language, thought, and culture.

Computational or automated metaphor analysis is a specific aspect of metaphorology that focuses on the use of computational methods and algorithms to analyze and understand metaphors. This can include tasks such as identifying metaphors in text, classifying them based on their type or meaning, and extracting the underlying structure or meaning of the metaphor. By using computational methods, researchers in this field can analyze and understand metaphors at a large scale and in a more systematic and rigorous way. This can provide valuable insights into the use and meaning of metaphors in different contexts, and help to advance our understanding of this complex and fascinating phenomenon.

References:

See also:

CogSketch | Computational Analogy | Computational Dreaming | Daydreaming Machines |  John A Barnden | Robopsychology