grobidOrg / grobid / 28018122208

39%

master: 39%

push

github

fix: emit balanced paragraph tags in fulltext training data #465

In testClosingTag(), the guard deciding whether to close a paragraph after a
reference marker used '||' between mutually exclusive equality checks
(citation/figure/table/equation marker), which is always true. As a result a
'</p>' was appended after a marker even when the following token was another
marker inside the same paragraph, producing redundant and unbalanced tags
(unclosed/extra <p>) in the generated *.training.fulltext.tei.xml. Use '&&'
so the paragraph is closed only when the next tag is neither the paragraph
continuation nor another embedded marker.

8652 of 24870 branches covered (34.79%)

Branch coverage included in aggregate %.

18423 of 45195 relevant lines covered (40.76%)

4.95 hits per line