原书出处
Natural Language Processing in Action, Second Edition
In addition to the overlap across nationalities, the PyTorch tutorial dataset contained many duplicated names within nationalities. More than 94% of the Arabic names were duplicates, some of which are shown in listing 8.5. Other nationalities and languages, such as English, Korean, and Scottish, appear to have been deduplicated. Duplicates in your training set make your model fit more closely to common names than to
原书定位详情
来源:a19c7933-4eda-4176-aebd-30fe7e1ad292
正文位置:extracted/Natural Language Processing in Action, Second Edition/_epub-src/OEBPS/Text/chapter-8.html
EPUB SHA-256:0e9a35e49ca446e322e5524e53f39fb6bee5b2b23d9cda10ef052aca770b0b36