|
|
|
|
|
by SchemaLoad
393 days ago
|
|
It doesn't though. It was going off usage of very common words like "its", "he", "and" rather than topic specific ones. Just that alone seems to work shockingly well. If you combined it with a few more data points like timestamps and topics of interest it would get even more accurate. |
|
I'm not questioning what would theoretically be possible to do, but the one that I saw failed the test.