Timeline replay of new paper feature

[TathagataChakraborti]

Originated from discussions with William Yeoh and Alberto Camacho during ICAPS 2022 demo session.

Replay the timeline of papers in survey data and analyze to what extent the new paper generator feature matches with work actually done in the past.

[haz]

How would this conceptually work? We could measure the hamming distance between a suggested paper and the actual "next written one", but is that interesting in any way?

[TathagataChakraborti]

No good ideas yet 🙊 doing a top-k recall would be too much coz of the number of possibilities. Need to think about it a bit.

Also, as a measure of limitation of the taxonomic classes, we can also check how often it's among the no goods but was an actual paper which means that we need to extend classes (if we trust peer review).