pvtopic_2020.tsv is an updated version of the pvtopic vectors described in [1] by N Smalheiser, et al.  The model and all inferred vectors were computed with the GenSIM package for Python [2].

Model specifics:
Articles were represented as described in [1] with concantenation of title abstract is at least 25 words.
Trained on a 50% corpus sample
Distributed Memory model with no word down-sampling, no vector concatenation, and using hierarchical softmax optimization
300-dimensional vectors for paragraphs and words
10 “iterations”
The exact GenSIM parameters used to fit the model are:
model = Doc2Vec(documents=corpus_iterator,dm=1,vector_size=300, window=10, min_count=10, epochs=10,
                              workers=16,alpha=0.025, min_alpha=0.0025, hs=1, dm_concat=0,
                              dm_mean=0, sample=0, negative=0)


File specifics:
pvtopic2020_vectors.tsv is a tab-delimited text file with two elements.
PMID <TAB> list of 300 vector values separated by spaces
There are 18239113 rows as of 2020-04-22.

pvtopic2020_wordvectors.tsv is a tab-delimited text file with two elements:
word <TAB> list of 300 vector values separated by spaces
There are 508166 rows.


References:
[1]  Unsupervised low-dimensional vector representations for words, phrases and text that are transparent, scalable, and produce similarity metrics that are not redundant with neural embeddings
N Smalheiser, A Cohen, G Bonifield - Journal of Biomedical Informatics

[2]  Software framework for topic modelling with large corpora
R Rehurek, P Sojka - In Proceedings of the LREC 2010 Workshop on New... , 2010