Error when calculating ranks for csentence

[l.main...@gmail.com]

Hi,

I'm calculating the ppr for 3000 synsets, but i get about a hundred lines like this:

Error when calculating ranks for csentence 00351000-n

Assuming that both the dictionary and the graph file are ok, maybe the problem is in the context file. Yet, it seems to be ok, since it is in this form:

00351000-n

00351000-n##fake_id#2#1

06888174-n

06888174-n##fake_id#2#1

...

Have you got any idea about the cause of these errors?

Thanks in advance.

Best,

Leonardo

[l.main...@gmail.com]

Then I have another question. In the file lkb_sources/Readme.txt you suggest to use

  ./compile_kb -o wn30+gloss.bin wnet30_rels.txt wnet30g_rels.txt

for creating a graph comprising both relations and glosses.

Unfortunately, when i run that command, ukb replies to me:

  multiple occurences

producing no outputs.

Maybe if i merge the 2 file it works, but i'd like to know your opinion.

Best,

Leonardo

[Aitor Soroa]

Hi Leonardo,

the documentation is wrong :-( you should create the graph like this:

cat wnet30_rels.txt wnet30g_rels.txt | ./compile_kb -o wn30+gloss.bin -

best,
aitor

> --
> You received this message because you are subscribed to the Google Groups "ukblist" group.
> To unsubscribe from this group and stop receiving emails from it, send an email to ukblist+u...@googlegroups.com.
> To post to this group, send email to ukb...@googlegroups.com.
> Visit this group at http://groups.google.com/group/ukblist?hl=en.
> For more options, visit https://groups.google.com/groups/opt_out.
>
>

--
ondo izan
aitor

[Aitor Soroa]

Hi again,

i've tried the context below and it works for me. The steps I've done:

# compile kb
% cat wnet30_rels.txt wnet30g_rels.txt | ./compile_kb -o wn.bin -
# copy your context to ctx.txt and run ukb_ppv
% ./ukb_ppv -K wn.bin ctx.txt

and it creates two files, 00351000-n.ppv and 06888174-n.ppv with the
personalized pagerank scores.

best,
aitor

[l.main...@gmail.com]

Yes, using the rels+glosses graph that context works well. I get those errors using the only glosses graph. And i remember that when i have used the only rels graph, i have got similar errors, but referring to different csentence.

So, maybe, when i use the rels+glosses graph, ukb calculates the ppr using only the glosses for a subset of synsets and using only the rels for another subset, and i don't know if it's ok.

What do you think about it?

[l.main...@gmail.com]

I have just noticed that maybe it happens when the graph does not contains the node for which i want to calculate ppr.

So maybe i was wrong about using the rels graph. 

Thank you for your replies.

Best,

Leonardo