Institut za hrvatski jezik

O projektu

Mrežna stranica projekta: https://metanet.icsi.berkeley.edu/metanet/

MetaNet: An Overview

The Metaphor Repository and Identification Pipeline

MetaNet, which is housed at the International Computer Science Institute in Berkeley, California, is a collaboration across multiple teams and campuses to systematize metaphor analysis in a computational way. Originally funded through the Intelligence Advanced Research Projects Activity (IARPA) agency, it was established to mine for metaphors in several languages, and design programs that can identify metaphors automatically while also link surface linguistic forms, such as "the pillars of democracy", to an underlying conceptual metaphor, such as IDEOLOGICAL SYSTEMS ARE PHYSICAL STRUCTURES. More detail from the initial press release of the project can be found here.

This website is mostly intended to house the activities, publications, and research news of the Berkeley branch, where the main focus was on the metaphor identification pipeline, corpus work, and metaphor network design. Additional information about the other branches can be found in the press release and in the About MetaNet section, and some of their publications are also listed in the Publications link above. As part of the larger project, we have developed the MetaNet Metaphor Repository and Identification System. The diagram above shows the components of the system and their interrelation.

The Understanding of Metaphor in MetaNet

In MetaNet, Conceptual Metaphor Theory is formally represented as a systematic network of searchable and interrelated metaphors, as well as a network of semantic frames that act as source and target domains of metaphors. In the MetaNet metaphor database, we have been creating a near-complete compendium of attested metaphors, including time metaphors, mind metaphors, event structure metaphors, emotion metaphors, and morality metaphors. It also includes novel metaphors pertaining to target domains of interest to the project, which have centered on social problems such as issues of poverty, taxation, bureaucracy, and governance. In addition to the searchable encyclopedia of systematically-linked metaphors and the semantic frames that constitute the source and target domains, the MetaNet project has been exploring metaphors in corpus data across four different languages, namely American English, Mexican Spanish, Iranian Persian, and Russian as spoken in Russia. The goal of the MetaNet project has been to build a system that will extract linguistic manifestations of metaphor and automatically interpret them. The diagram shows the system flow, from metaphor database creation and storage, to metaphor extraction from corpus texts, to analysis and querying of extracted data.

In order to evaluate the effectiveness of the metaphor repository and methodologies developed by the rest of the team, a team of cognitive linguists and neuroscientists have been creating stimuli experiments and analyzing brain scans.

For a more detailed explanation of the project, see MetaNet: A Multilingual Metaphor Repository.