Closed olegs closed 4 years ago
Small investigation showed that this is due to the fact that not all the papers are assigned with subtopic, i.e. some have -1.
Before dropping journals for papers with missing subtopics:
[2020-01-18 13:03:57,045: WARNING/ForkPoolWorker-1] journal_stats
[2020-01-18 13:03:57,045: WARNING/ForkPoolWorker-1] 262
[2020-01-18 13:03:57,053: WARNING/ForkPoolWorker-1] After drop papers with undefined journal
[2020-01-18 13:03:57,054: WARNING/ForkPoolWorker-1] 255
[2020-01-18 13:03:57,189: WARNING/ForkPoolWorker-1] journal comp counts sum
129 ArXiv [-1, 0, 2, 4, 1, 3, 5] [41, 27, 20, 14, 11, 9, 8] 130
207 Nature Methods [-1, 3, 0] [3, 2, 1] 6
137 Brain and nerve = Shinkei kenkyu no shinpo [-1] [3] 3
92 2018 International Congress on Big Data, Deep ... [-1, 1] [2, 1] 3
26 2017 25th Signal Processing and Communications... [-1] [3] 3
62 2018 26th Signal Processing and Communications... [-1, 1] [2, 1] 3
After:
[2020-01-18 13:03:57,191: WARNING/ForkPoolWorker-1] After drop papers with undefined subtopic
[2020-01-18 13:03:57,192: WARNING/ForkPoolWorker-1] 143
[2020-01-18 13:03:57,326: WARNING/ForkPoolWorker-1] journal comp counts sum
68 ArXiv [0, 2, 4, 1, 3, 5] [27, 20, 14, 11, 9, 8] 89
117 Nature Methods [3, 0] [2, 1] 3
106 Journal of High Energy Physics [4] [2] 2
43 2018 IEEE/ACM International Conference on Comp... [3, 5] [1, 1] 2
49 2018 Second International Conference on Intell... [1, 2] [1, 1] 2
This agrees with the 6 papers for Nature we get when navigate to journals.
Open "deep learning" at Semantic Scholar. See 3 papers for Nature methods, but on click 6 papers are shown.