Document Embedding with Paragraph Vectors

If by 'doc2vec' you mean the word2vec-like 'Paragraph Vectors' technique: even though that's a far simpler approach than the transformer embeddings, it usually works pretty well for coarse document similarity. Even the famous word2vec vector-addition operations kinda worked, as illustrated by some examples in the followup 'Paragraph Vector' paper in 2015: https://arxiv.org/abs/1507.07998 So if for you the resulting doc-to-doc similarities seemed nonsensical, there was likely some process error in model training or application.

Here's a more current paper from Quoc Le (http://research.google.com/pubs/QuocLe.html, https://scholar.google.com/citations?user=vfT6-XIAAAAJ) that expands on the "proof of concept" in the above paper: Document Embedding with Paragraph Vectors (2015) https://arxiv.org/pdf/1507.07998.pdf

Such a fluffy article. Here is what he's publishing: http://arxiv.org/find/cs/1/au:+Hinton_G/0/1/0/all/0/1 This seems to be a good introduction to the topic: http://arxiv.org/pdf/1310.4546v1.pdf This is about paragraph vectors: http://arxiv.org/abs/1507.07998 http://arxiv.org/abs/1405.4053

The gensim library now offers the 'Paragraph Vector' [1] approach to create vectors for longer ranges of text. It's available in a class named Doc2Vec – but I don't think that's what is being used here. The Paragraph Vector approach can give interesting results for document-similarity, including similarity after certain 'algebraic'-like additions/subtractions of other topics/word-concepts. [2] [1] http://arxiv.org/abs/1405.4053 [2] http://arxiv.org/abs/1507.07998