pkt / whatswrong

Automatically exported from code.google.com/p/whatswrong
GNU General Public License v3.0
0 stars 0 forks source link

Client specified line for spans #14

Open GoogleCodeExporter opened 9 years ago

GoogleCodeExporter commented 9 years ago
It would be great if the client could specify which row a span belongs to.  
Presently, whatswrong dynamically chooses where to put spans.

My first workaround was to add a "divider span" that spanned all tokens so 
spans added before and after the divider span would be forced to different 
lines.  This did not work as whatswrong put the divider span at the bottom 
instead of dividing rows like I thought it would.

My second workaround was to add filler spans on each row so that there would be 
a span for each token.  I thought that if I filled a row with spans, the next 
span I added would go into the second row.  However, whatswrong would sometimes 
put the later spans into the first row and the earlier spans into the second 
row.

My final work around was to have a separate image for each span row I want to 
present.  I'm displaying SRL frames so each row has a relation and arguments 
for that relation.

It's also not possible to have a span that exactly matches an existing span.  
They get merged into one.  However, in SRL it's entirely possible that two 
frames share some arguments.

This is probably a non-trivial feature request, but I thought I'd bring it up 
because it seemed like it should be possible (and maybe I'm missing something).

Attached are two images.  I would like to combine them into one with two rows 
of spans.

Original issue reported on code.google.com by sch...@gmail.com on 20 Feb 2013 at 11:12

Attachments:

GoogleCodeExporter commented 9 years ago
This may be a little more tricky. I wonder, at least for the issue of merged 
spans, whether you could give arguments different labels for different frames. 
That is, instead of labelling Michael with "A0" twice, label it it with 
"A0(hit)" and "A0(ran)" (or using token indices). I am not 100%, but thought 
they may be mechanism that avoid merging in this case (but I haven't looked at 
the code for a long time, so may be complete nonsense).

Original comment by sebastian.riedel@gmail.com on 20 Feb 2013 at 11:31

GoogleCodeExporter commented 9 years ago
Yes, this does work.  Although I decider earlier that for my purposes
having multiple images (one per frame) is the best solution for the
time being.

Original comment by sch...@gmail.com on 20 Feb 2013 at 11:38