Examples with TreePrint edu.stanford.nlp.trees.TreePrint used on opensource projects

Example 6 with TreePrint

use of edu.stanford.nlp.trees.TreePrint in project CoreNLP by stanfordnlp.

the class StanfordCoreNLP method construct.

//
// AnnotatorPool construction support
//
private void construct(Properties props, boolean enforceRequirements, AnnotatorImplementations annotatorImplementations, AnnotatorPool pool) {
    Timing tim = new Timing();
    this.numWords = 0;
    this.constituentTreePrinter = new TreePrint("penn");
    this.dependencyTreePrinter = new TreePrint("typedDependenciesCollapsed");
    if (props == null) {
        // if undefined, find the properties file in the classpath
        props = loadPropertiesFromClasspath();
    } else if (props.getProperty("annotators") == null) {
        // this happens when some command line options are specified (e.g just "-filelist") but no properties file is.
        // we use the options that are given and let them override the default properties from the class path properties.
        Properties fromClassPath = loadPropertiesFromClasspath();
        fromClassPath.putAll(props);
        props = fromClassPath;
    }
    this.properties = props;
    if (pool == null) {
        // if undefined, load the default annotator pool
        pool = getDefaultAnnotatorPool(props, annotatorImplementations);
    }
    // Set threading
    if (this.properties.containsKey("threads")) {
        ArgumentParser.threads = PropertiesUtils.getInt(this.properties, "threads");
        this.availableProcessors = new Semaphore(ArgumentParser.threads);
    } else {
        this.availableProcessors = new Semaphore(1);
    }
    // now construct the annotators from the given properties in the given order
    List<String> annoNames = Arrays.asList(getRequiredProperty(props, "annotators").split("[, \t]+"));
    Set<String> alreadyAddedAnnoNames = Generics.newHashSet();
    Set<Class<? extends CoreAnnotation>> requirementsSatisfied = Generics.newHashSet();
    for (String name : annoNames) {
        name = name.trim();
        if (name.isEmpty()) {
            continue;
        }
        logger.info("Adding annotator " + name);
        Annotator an = pool.get(name);
        this.addAnnotator(an);
        if (enforceRequirements) {
            Set<Class<? extends CoreAnnotation>> allRequirements = an.requires();
            for (Class<? extends CoreAnnotation> requirement : allRequirements) {
                if (!requirementsSatisfied.contains(requirement)) {
                    String fmt = "annotator \"%s\" requires annotation \"%s\". The usual requirements for this annotator are: %s";
                    throw new IllegalArgumentException(String.format(fmt, name, requirement.getSimpleName(), StringUtils.join(Annotator.DEFAULT_REQUIREMENTS.getOrDefault(name, Collections.singleton("unknown")), ",")));
                }
            }
            requirementsSatisfied.addAll(an.requirementsSatisfied());
        }
        alreadyAddedAnnoNames.add(name);
    }
    // Sanity check
    if (!alreadyAddedAnnoNames.contains(STANFORD_SSPLIT)) {
        System.setProperty(NEWLINE_SPLITTER_PROPERTY, "false");
    }
    this.pipelineSetupTime = tim.report();
}

Example 7 with TreePrint

use of edu.stanford.nlp.trees.TreePrint in project CoreNLP by stanfordnlp.

the class EvaluateTreebank method testOnTreebank.

/** Test the parser on a treebank. Parses will be written to stdout, and
   *  various other information will be written to stderr and stdout,
   *  particularly if <code>op.testOptions.verbose</code> is true.
   *
   *  @param testTreebank The treebank to parse
   *  @return The labeled precision/recall F<sub>1</sub> (EVALB measure)
   *          of the parser on the treebank.
   */
public double testOnTreebank(Treebank testTreebank) {
    log.info("Testing on treebank");
    Timing treebankTotalTimer = new Timing();
    TreePrint treePrint = op.testOptions.treePrint(op.tlpParams);
    TreebankLangParserParams tlpParams = op.tlpParams;
    TreebankLanguagePack tlp = op.langpack();
    PrintWriter pwOut, pwErr;
    if (op.testOptions.quietEvaluation) {
        NullOutputStream quiet = new NullOutputStream();
        pwOut = tlpParams.pw(quiet);
        pwErr = tlpParams.pw(quiet);
    } else {
        pwOut = tlpParams.pw();
        pwErr = tlpParams.pw(System.err);
    }
    if (op.testOptions.verbose) {
        pwErr.print("Testing ");
        pwErr.println(testTreebank.textualSummary(tlp));
    }
    if (op.testOptions.evalb) {
        EvalbFormatWriter.initEVALBfiles(tlpParams);
    }
    PrintWriter pwFileOut = null;
    if (op.testOptions.writeOutputFiles) {
        String fname = op.testOptions.outputFilesPrefix + "." + op.testOptions.outputFilesExtension;
        try {
            pwFileOut = op.tlpParams.pw(new FileOutputStream(fname));
        } catch (IOException ioe) {
            ioe.printStackTrace();
        }
    }
    PrintWriter pwStats = null;
    if (op.testOptions.outputkBestEquivocation != null) {
        try {
            pwStats = op.tlpParams.pw(new FileOutputStream(op.testOptions.outputkBestEquivocation));
        } catch (IOException ioe) {
            ioe.printStackTrace();
        }
    }
    if (op.testOptions.testingThreads != 1) {
        MulticoreWrapper<List<? extends HasWord>, ParserQuery> wrapper = new MulticoreWrapper<>(op.testOptions.testingThreads, new ParsingThreadsafeProcessor(pqFactory, pwErr));
        LinkedList<Tree> goldTrees = new LinkedList<>();
        for (Tree goldTree : testTreebank) {
            List<? extends HasWord> sentence = getInputSentence(goldTree);
            goldTrees.add(goldTree);
            pwErr.println("Parsing [len. " + sentence.size() + "]: " + SentenceUtils.listToString(sentence));
            wrapper.put(sentence);
            while (wrapper.peek()) {
                ParserQuery pq = wrapper.poll();
                goldTree = goldTrees.poll();
                processResults(pq, goldTree, pwErr, pwOut, pwFileOut, pwStats, treePrint);
            }
        }
        // for tree iterator
        wrapper.join();
        while (wrapper.peek()) {
            ParserQuery pq = wrapper.poll();
            Tree goldTree = goldTrees.poll();
            processResults(pq, goldTree, pwErr, pwOut, pwFileOut, pwStats, treePrint);
        }
    } else {
        ParserQuery pq = pqFactory.parserQuery();
        for (Tree goldTree : testTreebank) {
            final List<CoreLabel> sentence = getInputSentence(goldTree);
            pwErr.println("Parsing [len. " + sentence.size() + "]: " + SentenceUtils.listToString(sentence));
            pq.parseAndReport(sentence, pwErr);
            processResults(pq, goldTree, pwErr, pwOut, pwFileOut, pwStats, treePrint);
        }
    // for tree iterator
    }
    //Done parsing...print the results of the evaluations
    treebankTotalTimer.done("Testing on treebank");
    if (op.testOptions.quietEvaluation) {
        pwErr = tlpParams.pw(System.err);
    }
    if (saidMemMessage) {
        ParserUtils.printOutOfMemory(pwErr);
    }
    if (op.testOptions.evalb) {
        EvalbFormatWriter.closeEVALBfiles();
    }
    if (numSkippedEvals != 0) {
        pwErr.printf("Unable to evaluate %d parser hypotheses due to yield mismatch\n", numSkippedEvals);
    }
    // only created here so we know what parser types are supported...
    ParserQuery pq = pqFactory.parserQuery();
    if (summary) {
        if (pcfgLB != null)
            pcfgLB.display(false, pwErr);
        if (pcfgChildSpecific != null)
            pcfgChildSpecific.display(false, pwErr);
        if (pcfgLA != null)
            pcfgLA.display(false, pwErr);
        if (pcfgCB != null)
            pcfgCB.display(false, pwErr);
        if (pcfgDA != null)
            pcfgDA.display(false, pwErr);
        if (pcfgTA != null)
            pcfgTA.display(false, pwErr);
        if (pcfgLL != null && pq.getPCFGParser() != null)
            pcfgLL.display(false, pwErr);
        if (depDA != null)
            depDA.display(false, pwErr);
        if (depTA != null)
            depTA.display(false, pwErr);
        if (depLL != null && pq.getDependencyParser() != null)
            depLL.display(false, pwErr);
        if (factLB != null)
            factLB.display(false, pwErr);
        if (factChildSpecific != null)
            factChildSpecific.display(false, pwErr);
        if (factLA != null)
            factLA.display(false, pwErr);
        if (factCB != null)
            factCB.display(false, pwErr);
        if (factDA != null)
            factDA.display(false, pwErr);
        if (factTA != null)
            factTA.display(false, pwErr);
        if (factLL != null && pq.getFactoredParser() != null)
            factLL.display(false, pwErr);
        if (pcfgCatE != null)
            pcfgCatE.display(false, pwErr);
        for (Eval eval : evals) {
            eval.display(false, pwErr);
        }
        for (BestOfTopKEval eval : topKEvals) {
            eval.display(false, pwErr);
        }
    }
    // these ones only have a display mode, so display if turned on!!
    if (pcfgRUO != null)
        pcfgRUO.display(true, pwErr);
    if (pcfgCUO != null)
        pcfgCUO.display(true, pwErr);
    if (tsv) {
        NumberFormat nf = new DecimalFormat("0.00");
        pwErr.println("factF1\tfactDA\tfactEx\tpcfgF1\tdepDA\tfactTA\tnum");
        if (factLB != null)
            pwErr.print(nf.format(factLB.getEvalbF1Percent()));
        pwErr.print("\t");
        if (pq.getDependencyParser() != null && factDA != null)
            pwErr.print(nf.format(factDA.getEvalbF1Percent()));
        pwErr.print("\t");
        if (factLB != null)
            pwErr.print(nf.format(factLB.getExactPercent()));
        pwErr.print("\t");
        if (pcfgLB != null)
            pwErr.print(nf.format(pcfgLB.getEvalbF1Percent()));
        pwErr.print("\t");
        if (pq.getDependencyParser() != null && depDA != null)
            pwErr.print(nf.format(depDA.getEvalbF1Percent()));
        pwErr.print("\t");
        if (pq.getPCFGParser() != null && factTA != null)
            pwErr.print(nf.format(factTA.getEvalbF1Percent()));
        pwErr.print("\t");
        if (factLB != null)
            pwErr.print(factLB.getNum());
        pwErr.println();
    }
    double f1 = 0.0;
    if (factLB != null) {
        f1 = factLB.getEvalbF1();
    }
    //Close files (if necessary)
    if (pwFileOut != null)
        pwFileOut.close();
    if (pwStats != null)
        pwStats.close();
    if (parserQueryEvals != null) {
        for (ParserQueryEval parserQueryEval : parserQueryEvals) {
            parserQueryEval.display(false, pwErr);
        }
    }
    return f1;
}

Example 8 with TreePrint

use of edu.stanford.nlp.trees.TreePrint in project CoreNLP by stanfordnlp.

the class EvaluateTreebank method processResults.

public void processResults(ParserQuery pq, Tree goldTree, PrintWriter pwErr, PrintWriter pwOut, PrintWriter pwFileOut, PrintWriter pwStats, TreePrint treePrint) {
    if (pq.saidMemMessage()) {
        saidMemMessage = true;
    }
    Tree tree;
    List<? extends HasWord> sentence = pq.originalSentence();
    try {
        tree = pq.getBestParse();
    } catch (NoSuchParseException e) {
        tree = null;
    }
    List<ScoredObject<Tree>> kbestPCFGTrees = null;
    if (tree != null && kbestPCFG > 0) {
        kbestPCFGTrees = pq.getKBestPCFGParses(kbestPCFG);
    }
    //combo parse goes to pwOut (System.out)
    if (op.testOptions.verbose) {
        pwOut.println("ComboParser best");
        Tree ot = tree;
        if (ot != null && !op.tlpParams.treebankLanguagePack().isStartSymbol(ot.value())) {
            ot = ot.treeFactory().newTreeNode(op.tlpParams.treebankLanguagePack().startSymbol(), Collections.singletonList(ot));
        }
        treePrint.printTree(ot, pwOut);
    } else {
        treePrint.printTree(tree, pwOut);
    }
    // print various statistics
    if (tree != null) {
        if (op.testOptions.printAllBestParses) {
            List<ScoredObject<Tree>> parses = pq.getBestPCFGParses();
            int sz = parses.size();
            if (sz > 1) {
                pwOut.println("There were " + sz + " best PCFG parses with score " + parses.get(0).score() + '.');
                Tree transGoldTree = collinizer.transformTree(goldTree);
                int iii = 0;
                for (ScoredObject<Tree> sot : parses) {
                    iii++;
                    Tree tb = sot.object();
                    Tree tbd = debinarizer.transformTree(tb);
                    tbd = subcategoryStripper.transformTree(tbd);
                    pq.restoreOriginalWords(tbd);
                    pwOut.println("PCFG Parse #" + iii + " with score " + tbd.score());
                    tbd.pennPrint(pwOut);
                    Tree tbtr = collinizer.transformTree(tbd);
                    // pwOut.println("Tree size = " + tbtr.size() + "; depth = " + tbtr.depth());
                    kGoodLB.evaluate(tbtr, transGoldTree, pwErr);
                }
            }
        } else // Huang and Chiang (2006) Algorithm 3 output from the PCFG parser
        if (op.testOptions.printPCFGkBest > 0 && op.testOptions.outputkBestEquivocation == null) {
            List<ScoredObject<Tree>> trees = kbestPCFGTrees.subList(0, op.testOptions.printPCFGkBest);
            Tree transGoldTree = collinizer.transformTree(goldTree);
            int i = 0;
            for (ScoredObject<Tree> tp : trees) {
                i++;
                pwOut.println("PCFG Parse #" + i + " with score " + tp.score());
                Tree tbd = tp.object();
                tbd.pennPrint(pwOut);
                Tree tbtr = collinizer.transformTree(tbd);
                kGoodLB.evaluate(tbtr, transGoldTree, pwErr);
            }
        } else // Chart parser (factored) n-best list
        if (op.testOptions.printFactoredKGood > 0 && pq.hasFactoredParse()) {
            // DZ: debug n best trees
            List<ScoredObject<Tree>> trees = pq.getKGoodFactoredParses(op.testOptions.printFactoredKGood);
            Tree transGoldTree = collinizer.transformTree(goldTree);
            int ii = 0;
            for (ScoredObject<Tree> tp : trees) {
                ii++;
                pwOut.println("Factored Parse #" + ii + " with score " + tp.score());
                Tree tbd = tp.object();
                tbd.pennPrint(pwOut);
                Tree tbtr = collinizer.transformTree(tbd);
                kGoodLB.evaluate(tbtr, transGoldTree, pwOut);
            }
        } else //1-best output
        if (pwFileOut != null) {
            pwFileOut.println(tree.toString());
        }
        //Print the derivational entropy
        if (op.testOptions.outputkBestEquivocation != null && op.testOptions.printPCFGkBest > 0) {
            List<ScoredObject<Tree>> trees = kbestPCFGTrees.subList(0, op.testOptions.printPCFGkBest);
            double[] logScores = new double[trees.size()];
            int treeId = 0;
            for (ScoredObject<Tree> kBestTree : trees) logScores[treeId++] = kBestTree.score();
            //Re-normalize
            double entropy = 0.0;
            double denom = ArrayMath.logSum(logScores);
            for (double logScore : logScores) {
                double logPr = logScore - denom;
                entropy += Math.exp(logPr) * (logPr / Math.log(2));
            }
            //Convert to bits
            entropy *= -1;
            pwStats.printf("%f\t%d\t%d\n", entropy, trees.size(), sentence.size());
        }
    }
    // Perform various evaluations specified by the user
    if (tree != null) {
        //Strip subcategories and remove punctuation for evaluation
        tree = subcategoryStripper.transformTree(tree);
        Tree treeFact = collinizer.transformTree(tree);
        //Setup the gold tree
        if (op.testOptions.verbose) {
            pwOut.println("Correct parse");
            treePrint.printTree(goldTree, pwOut);
        }
        Tree transGoldTree = collinizer.transformTree(goldTree);
        if (transGoldTree != null)
            transGoldTree = subcategoryStripper.transformTree(transGoldTree);
        //Can't do evaluation in these two cases
        if (transGoldTree == null) {
            pwErr.println("Couldn't transform gold tree for evaluation, skipping eval. Gold tree was:");
            goldTree.pennPrint(pwErr);
            numSkippedEvals++;
            return;
        } else if (treeFact == null) {
            pwErr.println("Couldn't transform hypothesis tree for evaluation, skipping eval. Tree was:");
            tree.pennPrint(pwErr);
            numSkippedEvals++;
            return;
        } else if (treeFact.yield().size() != transGoldTree.yield().size()) {
            List<Label> fYield = treeFact.yield();
            List<Label> gYield = transGoldTree.yield();
            pwErr.println("WARNING: Evaluation could not be performed due to gold/parsed yield mismatch.");
            pwErr.printf("  sizes: gold: %d (transf) %d (orig); parsed: %d (transf) %d (orig).%n", gYield.size(), goldTree.yield().size(), fYield.size(), tree.yield().size());
            pwErr.println("  gold: " + SentenceUtils.listToString(gYield, true));
            pwErr.println("  pars: " + SentenceUtils.listToString(fYield, true));
            numSkippedEvals++;
            return;
        }
        if (topKEvals.size() > 0) {
            List<Tree> transGuesses = new ArrayList<>();
            int kbest = Math.min(op.testOptions.evalPCFGkBest, kbestPCFGTrees.size());
            for (ScoredObject<Tree> guess : kbestPCFGTrees.subList(0, kbest)) {
                transGuesses.add(collinizer.transformTree(guess.object()));
            }
            for (BestOfTopKEval eval : topKEvals) {
                eval.evaluate(transGuesses, transGoldTree, pwErr);
            }
        }
        //PCFG eval
        Tree treePCFG = pq.getBestPCFGParse();
        if (treePCFG != null) {
            Tree treePCFGeval = collinizer.transformTree(treePCFG);
            if (pcfgLB != null) {
                pcfgLB.evaluate(treePCFGeval, transGoldTree, pwErr);
            }
            if (pcfgChildSpecific != null) {
                pcfgChildSpecific.evaluate(treePCFGeval, transGoldTree, pwErr);
            }
            if (pcfgLA != null) {
                pcfgLA.evaluate(treePCFGeval, transGoldTree, pwErr);
            }
            if (pcfgCB != null) {
                pcfgCB.evaluate(treePCFGeval, transGoldTree, pwErr);
            }
            if (pcfgDA != null) {
                // Re-index the leaves after Collinization, stripping traces, etc.
                treePCFGeval.indexLeaves(true);
                transGoldTree.indexLeaves(true);
                pcfgDA.evaluate(treePCFGeval, transGoldTree, pwErr);
            }
            if (pcfgTA != null) {
                pcfgTA.evaluate(treePCFGeval, transGoldTree, pwErr);
            }
            if (pcfgLL != null && pq.getPCFGParser() != null) {
                pcfgLL.recordScore(pq.getPCFGParser(), pwErr);
            }
            if (pcfgRUO != null) {
                pcfgRUO.evaluate(treePCFGeval, transGoldTree, pwErr);
            }
            if (pcfgCUO != null) {
                pcfgCUO.evaluate(treePCFGeval, transGoldTree, pwErr);
            }
            if (pcfgCatE != null) {
                pcfgCatE.evaluate(treePCFGeval, transGoldTree, pwErr);
            }
        }
        //Dependency eval
        // todo: is treeDep really useful here, or should we really use depDAEval tree (debinarized) throughout? We use it for parse, and it sure seems like we could use it for tag eval, but maybe not factDA?
        Tree treeDep = pq.getBestDependencyParse(false);
        if (treeDep != null) {
            Tree goldTreeB = binarizerOnly.transformTree(goldTree);
            Tree goldTreeEval = goldTree.deepCopy();
            goldTreeEval.indexLeaves(true);
            goldTreeEval.percolateHeads(op.langpack().headFinder());
            Tree depDAEval = pq.getBestDependencyParse(true);
            depDAEval.indexLeaves(true);
            depDAEval.percolateHeadIndices();
            if (depDA != null) {
                depDA.evaluate(depDAEval, goldTreeEval, pwErr);
            }
            if (depTA != null) {
                Tree undoneTree = debinarizer.transformTree(treeDep);
                undoneTree = subcategoryStripper.transformTree(undoneTree);
                pq.restoreOriginalWords(undoneTree);
                // pwErr.println("subcategoryStripped tree: " + undoneTree.toStructureDebugString());
                depTA.evaluate(undoneTree, goldTree, pwErr);
            }
            if (depLL != null && pq.getDependencyParser() != null) {
                depLL.recordScore(pq.getDependencyParser(), pwErr);
            }
            Tree factTreeB;
            if (pq.hasFactoredParse()) {
                factTreeB = pq.getBestFactoredParse();
            } else {
                factTreeB = treeDep;
            }
            if (factDA != null) {
                factDA.evaluate(factTreeB, goldTreeB, pwErr);
            }
        }
        //Factored parser (1best) eval
        if (factLB != null) {
            factLB.evaluate(treeFact, transGoldTree, pwErr);
        }
        if (factChildSpecific != null) {
            factChildSpecific.evaluate(treeFact, transGoldTree, pwErr);
        }
        if (factLA != null) {
            factLA.evaluate(treeFact, transGoldTree, pwErr);
        }
        if (factTA != null) {
            factTA.evaluate(tree, boundaryRemover.transformTree(goldTree), pwErr);
        }
        if (factLL != null && pq.getFactoredParser() != null) {
            factLL.recordScore(pq.getFactoredParser(), pwErr);
        }
        if (factCB != null) {
            factCB.evaluate(treeFact, transGoldTree, pwErr);
        }
        for (Eval eval : evals) {
            eval.evaluate(treeFact, transGoldTree, pwErr);
        }
        if (parserQueryEvals != null) {
            for (ParserQueryEval eval : parserQueryEvals) {
                eval.evaluate(pq, transGoldTree, pwErr);
            }
        }
        if (op.testOptions.evalb) {
            // empty out scores just in case
            nanScores(tree);
            EvalbFormatWriter.writeEVALBline(treeFact, transGoldTree);
        }
    }
    pwErr.println();
}