a) Start GATE, load Alignment and ANNIE plugins.
b) Create new ANNIE English Tokeniser and Segment Processing PR.
c) Create new document from the attached gate_segment_test.xml file.
d) Create new corpus with this document.
e) Create corpus pipeline with the Segment Processing PR.
analyser: the ANNIE English Tokeniser instance
inputASName: Original markups
f) Run the pipeline with the corpus instance.
g) Look at default AS of the document. It should look like on the attached screen shot gate_segment_test.png. Four token annotations are missing (for words 'two', 'three', 'five' and 'six').
This bug is new with Gate 8. It is OK with Gate 7.1
The content of the attached gate_segment_test.xml file is:
<Doc><Section>one two three</Section>
<Section>four five six</Section>
<Section>seven eight nine</Section></Doc>
This was actually two bugs; one in that we were getting the next annotation ID from the wrong (possibly non-existant) document andtwo that we weren't ensuring the next annotation ID was always big enough. Both of these have now been fixed.