bcorrigan
commented
5 years ago Could perhaps infer relationships or a hierarchy:
- Suppose tags "Science Fiction" and "posthuman" are likely to appear together
- "posthuman" appears with "Science Fiction" 30% of the time
- But "Science Fiction" only rarely has "posthuman"
From this we can infer that one is a subtype of the other. But this would only be statisticlaly meaningful for extremely large collections of books.
Another possible way is to have a static hierarchy of genres and subgenres that can be manually curated, and simply match tags against the hierarchy. That would probably be "good enough"
Already index all the tags as facets in tantivy.
Should take advantage of this and add a way to view tags in the frontend.
Although each book just has a list of tags, is it possible to infer relationships between the tags in some way? Doubtful