Examples with WordNetManager edu.illinois.cs.cogcomp.edison.utilities.WordNetManager used on opensource projects

Example 6 with WordNetManager

use of edu.illinois.cs.cogcomp.edison.utilities.WordNetManager in project cogcomp-nlp by CogComp.

the class BIOCombinedReader method getTokensFromTAs.

private List<Constituent> getTokensFromTAs() throws InvalidPortException, InvalidEndpointException, IOException, JWNLException, DatastoreException {
    List<Constituent> ret = new ArrayList<>();
    WordNetManager wordNet = null;
    Gazetteers gazetteers = null;
    BrownClusters brownClusters = null;
    Datastore ds = new Datastore(new ResourceConfigurator().getDefaultConfig());
    File gazetteersResource = ds.getDirectory("org.cogcomp.gazetteers", "gazetteers", 1.3, false);
    gazetteers = GazetteersFactory.get(5, gazetteersResource.getPath() + File.separator + "gazetteers", true, Language.English);
    Vector<String> bcs = new Vector<>();
    bcs.add("brown-clusters/brown-english-wikitext.case-intact.txt-c1000-freq10-v3.txt");
    bcs.add("brown-clusters/brownBllipClusters");
    bcs.add("brown-clusters/brown-rcv1.clean.tokenized-CoNLL03.txt-c1000-freq1.txt");
    Vector<Integer> bcst = new Vector<>();
    bcst.add(5);
    bcst.add(5);
    bcst.add(5);
    Vector<Boolean> bcsl = new Vector<>();
    bcsl.add(false);
    bcsl.add(false);
    bcsl.add(false);
    brownClusters = BrownClusters.get(bcs, bcst, bcsl);
    WordNetManager.loadConfigAsClasspathResource(true);
    wordNet = WordNetManager.getInstance();
    for (TextAnnotation ta : currentTas) {
        View tokenView = ta.getView(ViewNames.TOKENS);
        String mentionViewName = "";
        if (ta.getId().startsWith("bn") || ta.getId().startsWith("nw")) {
            mentionViewName = ViewNames.MENTION_ACE;
        } else {
            mentionViewName = ViewNames.MENTION_ERE;
        }
        View mentionView = ta.getView(mentionViewName);
        View bioView = new SpanLabelView("BIO", BIOReader.class.getCanonicalName(), ta, 1.0f);
        String[] token2tags = new String[tokenView.getConstituents().size()];
        for (int i = 0; i < token2tags.length; i++) {
            token2tags[i] = "O";
        }
        for (Constituent c : mentionView.getConstituents()) {
            if (!_type.equals("ALL")) {
                String excludeType = _type;
                if (!c.getAttribute("EntityMentionType").equals(excludeType)) {
                    continue;
                }
            }
            Constituent cHead = ACEReader.getEntityHeadForConstituent(c, ta, "HEAD");
            if (!c.hasAttribute("EntityType")) {
                c.addAttribute("EntityType", c.getLabel());
            }
            if (cHead == null) {
                continue;
            }
            if (c.getAttribute("EntityType").equals("VEH") || c.getAttribute("EntityType").equals("WEA")) {
            // continue;
            }
            c.addAttribute("EntityType", "MENTION");
            /**
             * @Note that unlike BIOReader, the tagging schema is set to "BIOLU" here
             */
            if (cHead.getStartSpan() + 1 == cHead.getEndSpan()) {
                token2tags[cHead.getStartSpan()] = "U-" + c.getAttribute("EntityType") + "," + c.getAttribute("EntityMentionType");
            } else {
                token2tags[cHead.getStartSpan()] = "B-" + c.getAttribute("EntityType") + "," + c.getAttribute("EntityMentionType");
                for (int i = cHead.getStartSpan() + 1; i < cHead.getEndSpan() - 1; i++) {
                    token2tags[i] = "I-" + c.getAttribute("EntityType") + "," + c.getAttribute("EntityMentionType");
                }
                token2tags[cHead.getEndSpan() - 1] = "L-" + c.getAttribute("EntityType") + "," + c.getAttribute("EntityMentionType");
            }
        }
        for (int i = 0; i < token2tags.length; i++) {
            Constituent curToken = tokenView.getConstituentsCoveringToken(i).get(0);
            Constituent newToken = curToken.cloneForNewView("BIO");
            if (token2tags[i].equals("O")) {
                newToken.addAttribute("BIO", token2tags[i]);
            } else {
                String[] group = token2tags[i].split(",");
                String tag = group[0];
                String eml = group[1];
                newToken.addAttribute("BIO", tag);
                newToken.addAttribute("EntityMentionType", eml);
            }
            newToken.addAttribute("GAZ", ((FlatGazetteers) gazetteers).annotateConstituent(newToken, false));
            newToken.addAttribute("BC", brownClusters.getPrefixesCombined(newToken.toString()));
            if (!newToken.toString().contains("http")) {
                newToken.addAttribute("WORDNETTAG", BIOFeatureExtractor.getWordNetTags(wordNet, newToken));
                newToken.addAttribute("WORDNETHYM", BIOFeatureExtractor.getWordNetHyms(wordNet, newToken));
            } else {
                newToken.addAttribute("WORDNETTAG", ",");
                newToken.addAttribute("WORDNETHYM", ",");
            }
            if (_mode.contains("TRAIN")) {
                newToken.addAttribute("isTraining", "true");
            } else {
                newToken.addAttribute("isTraining", "false");
            }
            bioView.addConstituent(newToken);
        }
        ta.addView("BIO", bioView);
        for (Constituent c : bioView) {
            ret.add(c);
        }
    }
    return ret;
}