Isao Matsunami
Isao Matsunami
And please, please remember me #4425
In my opinion, Most of OpenRefine users use it, with real-world entities in mind, to process NOT data bytes BUT concrete concepts. So non-discernible bytes should not have any effects....
You both know OpenRefine much much better. I leave this entirely to you. I just wanted to report what I encountered when I merged plain text data and scraped data....
I will update the trained data file this week. It won't reach 98% and this is done just for a single font set. clstm has never vomited "floating point error"...
Sometimes it drops to 4-5%. but that is only several characters in randomized character strings. updated) 178000 iterations, 2.23% error rate for first 100 samples (which are not used for...
Use https://github.com/tmbdev/clstm I just set environment variables nhidden=800, lrate=1e-4
As of the current nightly-built OR, the import side of this problem is solved. The export side is still there. When the project name has non-ascii chars, for example "result選挙2020",...