myleott / jgibblabeledlda Goto Github PK

After some warning of "Unknown document label ( term ) for document doc_id."
An exception occurs:

java.lang.ArrayIndexOutOfBoundsException: 116
at jgibblda.Estimator.sampling(Estimator.java:114)
at jgibblda.Estimator.estimate(Estimator.java:66)
at jgibblda.LDA.main(LDA.java:52)

My data like:
[10] term1_1 term1_2 ...
term_2_1 term2_2 ...

according to README, i copy the example of input file ,however the result is
Unknown document label ( label_2,1 ) for document 1.
Unknown document label ( label_2,2 ) for document 1.
Unknown document label ( label_2,l_2 ) for document 1.
when i use only one line into the input,however,it can't figure out the "[",the model just think the charecter and the "[" as one word which means the model think the inputfile as unlabeled.But in the LDAcmdoption,-unlabeled is false.could you please give us a example of inputfile ,so i can know what's wrong with our inputfile?

I just ran Labelled LDA over a bunch of documents following the documentation, but after it got completed I am unable to locate where it has saved the model files. My initial presumption was the input directory but nothing is in there. There were no errors while training:

Dataset loaded:
        M:50902
        V:1931403
Sampling 100 iterations!
Iteration   100

Saving the final model!

Can you point me out where does it save the model?

Thanks for sharing the Labelled LDA solution. I tried with [news,1 tech,2], it shows warning for the unknown label. looking at the code in LDADataSet:addDoc,

try { label_set.add(Integer.parseInt(labelStr)); } catch (NumberFormatException nfe) { System.err.println("Unknown document label ( " + labelStr + " ) for document " + docs.size() + "."); }

I think the reason is that, 'Integer.parseInt(labelStr)' always fails for the label formate label_1,int.

Hi,
I am trying to learn LDA. Referrering to you code. But cant fine the input files
Can you pls put the sample input and other required files to run this example in some folder

Hi there,

I'm trying to use your code, but running into problems. After cloning the repo, I try to compile the code, but it issues errors. When I do:

$ javac -cp lib/args4j-2.0.6.jar:lib/trove-3.0.3.jar src/jgibblda/Estimator.java

I get

src/jgibblda/Estimator.java:37: error: cannot find symbol
    protected Model trnModel;
              ^
  symbol:   class Model
  location: class Estimator
src/jgibblda/Estimator.java:38: error: cannot find symbol
    LDACmdOption option;
    ^
  symbol:   class LDACmdOption
  location: class Estimator
src/jgibblda/Estimator.java:40: error: cannot find symbol
    public Estimator(LDACmdOption option) throws FileNotFoundException, IOException
                     ^
  symbol:   class LDACmdOption
  location: class Estimator
src/jgibblda/Estimator.java:44: error: cannot find symbol
        trnModel = new Model(option);
                       ^
  symbol:   class Model
  location: class Estimator
src/jgibblda/Estimator.java:46: error: illegal start of type
        if (option.est){
           ^
src/jgibblda/Estimator.java:49: error: illegal start of type
        else if (option.estc){
                ^
6 errors

I rarely use java, but my instinct tells me this is because of the code being written for a different versio of Java. I'm using 1.8.0_131. What is the intended version, or, how do I solve this. Thanks in advance!

can current algorithm produce a label for unlabelled documents?