Text classification items are too large

[646e62]

The docs suggest that classifying large text chunks may lead to weak results. Although not all case decision paragraphs are several hundred entities long, many are. Classifying text by sentence, rather than by paragraph, may be the better choice.

[bbelderbos]

Interesting