Closed kernc closed 3 years ago
Typo in the configuration file. Fixed by https://github.com/insight-centre/naisc/commit/c3ed081a391f8e7c82eaed76c10bd95011788dea
Right. But now, running:
./naisc.sh /tmp/test.ttl /tmp/test.ttl -c configs/ontolex-default.json
I get:
[INITIALIZING] Loading Scorers
org.insightcentre.uld.naisc.main.ConfigurationException: Model file does not exist. (Perhaps you need to train this model?)
at org.insightcentre.uld.naisc.scorer.LibSVM.makeScorer(LibSVM.java:58)
at org.insightcentre.uld.naisc.main.Configuration.makeScorer(Configuration.java:206)
at org.insightcentre.uld.naisc.main.Main.execute(Main.java:287)
at org.insightcentre.uld.naisc.main.Main.execute2(Main.java:188)
at org.insightcentre.uld.naisc.main.Main.execute(Main.java:150)
at org.insightcentre.uld.naisc.main.Main.execute(Main.java:108)
at org.insightcentre.uld.naisc.main.Main.execute(Main.java:84)
at org.insightcentre.uld.naisc.main.Main.main(Main.java:584)
[FAILED] org.insightcentre.uld.naisc.main.ConfigurationException: Model file does not exist. (Perhaps you need to train this model?)
I don't think models/default.libsvm, referenced in ontolex-default.json, comes by get-models.sh
. I'm not supposed to train the model, am I? Can I switch it for a different model that is available/works?
Yeah, that model does need to be trained. There are two solutions here
train.sh
. (Better performance)"scorer.LibSVM"
with "scorer.RAdLR"
(worse performance)I guess for deployment we will provide a trained model, but you can work with the unsupervised model for now
Using the following OntoLex RDF/Turtle, stripped down from example:
Running Naisc with
-c configs/auto.json
, I get roughly the expected links:However, running with
-c configs/ontolex-default.json
, it crashes for me as below:Would you happen to have an idea what's the issue about and how I might mitigate it?