Artificial intelligence may soon understand language metaphors
Metaphorical language has coloured our speech for millennia, proving its importance as well in fields as diverse as psychology, philosophy, education, politics and business. This uniquely human way of expressing oneself has presented a challenge for artificial intelligence (AI) systems that must understand metaphor in order to communicate better with people. The EU-funded GEN-META (Generating metaphors using a combination of AI reasoning and corpus-based modeling of formulaic expressions) project studied how AI could understand and produce metaphor. It worked on building a natural language understanding and generation system that focuses on dialogue and cutting-edge metaphor generation mechanisms. META aimed to teach the system when to use metaphor if at all, update it with particular metaphorical words or phrases, and promote natural forms of metaphorical expression. Combining two existing AI language systems in novel ways, the project team extended an initial proof-of-concept for the system under development. It aspired to achieve metaphor understanding that goes beyond the capabilities of existing systems and create metaphorically expressed linguistic output using new ways of understanding metaphors. Project work also revolved around detecting metaphors in text and collecting numerous texts that portray metaphor usage in a variety of contexts such as climate change, immigration and security. Overall the research has significantly advanced the relevance and usefulness of automated natural language processing in relation to a variety of everyday activities. This, for instance, could be used to facilitate the inclusion of marginalised citizens in Europe’s digital economy by streamlining human-computer interfacing. Another example is making progress in refining language teaching technology. It is only a matter of time before AI understands metaphor and makes communication with machines more humanlike than ever.
Keywords
Artificial intelligence, metaphors, GEN-META, digital economy, human-computer interfacing
