What's new in the Cog-Sci world?
Can AI systems analyze social networks in literary works?
What do books like 1984 and Game of Thrones have in common? While they both possess intricate worlds and an interesting plot, a recent study was interested in whether their social networks were similar, regardless of when they were published. Additionally, they tested whether some language tools were able to hold up to today’s standards.
To analyze this, a machine-learning program was trained to identify text patterns and recognize relationships between characters based on how often they were mentioned. AAfterward this program was run against various science fiction literature, on novels from the 17th to 21st century.
The results showed that there were no significant differences between old and modern novels, in terms of being able to analyze their social networks. However, the data also highlighted gaps in the tools used; For instance, while nicknames like “Dany” for “Daenerys Targaryen” could be identified, apostrophe names like “d’Artagnan” could not. While data like this could be used to potentially develop a culturally-aware AI, it’s clear it’ll take some time before computers can accurately “read text.”