Open brendano opened 8 years ago
When I ran it on some news articles, these errors didn't occur.
I guess when the parser meets the our-of-bound error once, all the following sentences get same errors, too.
This is happening because of the blank lines in the input file. I don't think any sentence segmentation is being done in the preprocessing. It assumes one sentence per line. Is that right, @jflanigan?
I guess when the parser meets the our-of-bound error once, all the following sentences get same errors, too.
Yeah, once /tmp/jamr-XXX.snt.tok
, /tmp/jamr-XXX.snt.IllinoisNER
, /tmp/jamr-XXX.snt.deps
, etc. all get out of alignment (b/c some have extra blank lines, some don't), it's hosed from then on.
oh, got it. i recall the readme might have mentioned that.
On a small test sentence I got JAMR to run fine. But on a harder document it gave lots of array out of bound errors. Is this serious? If the syntactic dependency fails, does the AMR parser always return an empty semantic graph?
This was running
scripts/PARSE.sh < LICENSE.txt > LICENSE.out
LICENSE.err.txt LICENSE.out.txt