List of usage examples for weka.core.pmml PMMLFactory getPMMLModel
public static PMMLModel getPMMLModel(InputStream stream) throws Exception
From source file:es.ua.dlsi.experiments.id3.CheckCorrectCandidatePositionLeaveOneOutScoresMaximumEntropy.java
License:Open Source License
/** * @param args the command line arguments *///from w w w.jav a2s . com public static void main(String[] args) { CmdLineParser parser = new CmdLineParser(); CmdLineParser.Option odictionary = parser.addStringOption('d', "dictionary"); CmdLineParser.Option oremove1entry = parser.addBooleanOption("remove-1entrypars"); CmdLineParser.Option ooutput = parser.addStringOption('o', "output"); CmdLineParser.Option otreeoutput = parser.addStringOption("tree-output"); CmdLineParser.Option onotclosedcats = parser.addBooleanOption("remove-closedcats"); CmdLineParser.Option ovocabularypath = parser.addStringOption('v', "vocabulary"); CmdLineParser.Option oplf_tmp = parser.addStringOption('p', "plf-tmp-path"); CmdLineParser.Option olrm = parser.addStringOption('m', "linear-regression-model"); try { parser.parse(args); } catch (CmdLineParser.IllegalOptionValueException e) { System.err.println(e); System.exit(-1); } catch (CmdLineParser.UnknownOptionException e) { System.err.println(e); System.exit(-1); } String dictionary = (String) parser.getOptionValue(odictionary, null); String output = (String) parser.getOptionValue(ooutput, null); String treeoutput = (String) parser.getOptionValue(otreeoutput, null); String vocabularypath = (String) parser.getOptionValue(ovocabularypath, null); String plf_tmp = (String) parser.getOptionValue(oplf_tmp, null); String lrm = (String) parser.getOptionValue(olrm, null); boolean remove1entry = (Boolean) parser.getOptionValue(oremove1entry, false); boolean notclosedcats = (Boolean) parser.getOptionValue(onotclosedcats, false); //Preparing output stream PrintWriter pw; if (output != null) { try { pw = new PrintWriter(output); } catch (FileNotFoundException ex) { System.err.println("Error while traying to write output file '" + output + "'."); pw = new PrintWriter(System.out); } } else { System.err.println("Warning: output file not defined. Output redirected to standard output."); pw = new PrintWriter(System.out); } //Preparing output stream PrintWriter treepw = null; if (treeoutput != null) { try { treepw = new PrintWriter(treeoutput); } catch (FileNotFoundException ex) { System.err.println("Error while traying to write output file for the tree '" + treeoutput + "'."); treepw = new PrintWriter(System.out); } } //Reading the vocabulary Vocabulary vocabulary = null; try { vocabulary = new Vocabulary(vocabularypath); } catch (FileNotFoundException ex) { System.err.println("ERROR: File '" + vocabularypath + "' could not be found."); System.exit(-1); } catch (IOException ex) { System.err.println("Error while reading file '" + vocabularypath + "' could not be found."); System.exit(-1); } //Reading the dictionary and generating the set of lexical forms DictionaryReader dicReader = new DictionaryReader(dictionary); Dictionary dic = dicReader.readDic(); //Building the suffix tree Dix2suffixtree d2s; d2s = new Dix2suffixtree(dic); FeatureExtractor featextractor = new FeatureExtractor(dic, vocabulary, d2s, plf_tmp); LinearRegression lrmodel = null; try { PMMLModel pmmlModel = PMMLFactory.getPMMLModel(lrm); if (pmmlModel instanceof PMMLClassifier) { Classifier classifier = ((PMMLClassifier) pmmlModel); lrmodel = (LinearRegression) classifier; } } catch (Exception ex) { ex.printStackTrace(System.err); System.exit(-1); } //Loop that goes all over the entries of the dictionary for (Section s : dic.sections) { for (int i = 0; i < s.elements.size(); i++) { E e = s.elements.remove(i); //If the entry is a multiword is discarded if (e.isMultiWord()) { System.err.println("Multiword: " + e.toString()); } else { //Getting the stema nd paradign of the entry Candidate candidate = DicEntry.GetStemParadigm(e); if (candidate != null) { Pardef pardef = dic.pardefs.getParadigmDefinition(candidate.getParadigm()); if (pardef != null) { ParadigmProfiler pp = new ParadigmProfiler(new Paradigm(pardef, dic), dic); if (!remove1entry || pp.NumberOfWords() > 1) { String stem = candidate.getStem(); String bestsurfaceform; Pardef p = dic.pardefs.getParadigmDefinition(candidate.getParadigm()); Paradigm paradigm = new Paradigm(p, dic); //If indicated, entries generating forms from a closed category may be discarded if (!notclosedcats || !paradigm.isClosedCategory()) { //Choosing the most frequent surface form in the vocabulary bestsurfaceform = vocabulary.GetMostFrequentSurfaceForm(stem, paradigm); //If no one of the surface forms appear in the vocabulary: if (bestsurfaceform == null) { System.err.println("Warning: no occurrence for word with stem " + stem + " and paradigm " + paradigm.getName()); //Random form bestsurfaceform = stem + paradigm.getSuffixes().iterator().next().getSuffix(); } //If the lemma cannot be found, the system stops working if (candidate.GetLemma(dic) == null) { System.err.println("Error: lemma cannot be generated for stem " + stem + " and paradigm " + paradigm.getName()); System.exit(-1); } //Generating the list of candidates for the most common surface form //Set<Candidate> candidates=d2s.getSuffixTree(). // SegmentWord(bestsurfaceform); SortedSetOfCandidates candidates = d2s.CheckNewWord(bestsurfaceform, vocabulary, plf_tmp, null, notclosedcats); if (candidates.GetNumberOfDifferentCandidates() == 0) { String newsurfaceform; for (Suffix suf : paradigm.getSuffixes()) { newsurfaceform = stem + suf; if (!newsurfaceform.equals(bestsurfaceform)) { candidates = d2s.CheckNewWord(newsurfaceform, vocabulary, null, null, notclosedcats); if (candidates.GetNumberOfDifferentCandidates() > 0) { bestsurfaceform = newsurfaceform; break; } } } } if (candidates.GetNumberOfDifferentCandidates() == 0) { System.err.println("Warning: no candidates for candidate " + stem + "/" + paradigm.getName()); } else { Set<String> possiblesurfaceforms = new LinkedHashSet<String>(); //the key of this map is the set of surface forms and the value is the set of paradigms generating them Set<EquivalentCandidates> sf_candidate = new LinkedHashSet<EquivalentCandidates>(); for (RankedCandidate qc : candidates.getCandidates()) { possiblesurfaceforms.addAll(qc.getSurfaceForms(dic)); sf_candidate.add(qc); } for (EquivalentCandidates ec : sf_candidate) { RankedCandidate qc = (RankedCandidate) ec; FeatureSet featset = featextractor.GetFeatureSet(qc, notclosedcats); try { double probability = lrmodel .classifyInstance(featset.toWekaInstance()); qc.setScore(probability); } catch (Exception ex) { ex.printStackTrace(System.err); } } InstanceCollection records; // read in all our data records = new InstanceCollection(); records.buildInstances(possiblesurfaceforms, sf_candidate, dic); Tree tree = new Tree(records); tree.Print(treepw); treepw.flush(); try { int numberofquestions = tree.QuestionsToParadigm(candidate); //Printing the output pw.println(bestsurfaceform + ";" + stem + ";" + paradigm.getName() + ";" + numberofquestions); pw.flush(); s.elements.add(i, e); } catch (NotInTreeException ex) { System.out.println("Error: correct candidate for " + stem + ";" + paradigm.getName() + " is not in the ID3 tree."); } } } else { System.err.println("Closed category: " + e.toString()); } } else { System.err.println("Candidate " + candidate.toString() + " not processed: it is the only word in the paradigm"); } } else { System.err.println( "Paradigm " + candidate.getParadigm() + " does not appear in the dictionary"); } } else { System.err.println("Entry " + e.toString() + " does not contain any paradigm"); } } } } pw.close(); if (treepw != null) { treepw.close(); } }