Examples with EntityMention edu.stanford.nlp.ie.machinereading.structure.EntityMention used on opensource projects

Example 11 with EntityMention

use of edu.stanford.nlp.ie.machinereading.structure.EntityMention in project CoreNLP by stanfordnlp.

the class RelationExtractorAnnotator method annotate.

@Override
public void annotate(Annotation annotation) {
    // extract entities and relations
    Annotation output = mr.annotate(annotation);
    // transfer entities/relations back to the original annotation
    List<CoreMap> outputSentences = output.get(SentencesAnnotation.class);
    List<CoreMap> origSentences = annotation.get(SentencesAnnotation.class);
    for (int i = 0; i < outputSentences.size(); i++) {
        CoreMap outSent = outputSentences.get(i);
        CoreMap origSent = origSentences.get(i);
        // set entities
        List<EntityMention> entities = outSent.get(MachineReadingAnnotations.EntityMentionsAnnotation.class);
        origSent.set(MachineReadingAnnotations.EntityMentionsAnnotation.class, entities);
        if (verbose && entities != null) {
            log.info("Extracted the following entities:");
            for (EntityMention e : entities) {
                log.info("\t" + e);
            }
        }
        // set relations
        List<RelationMention> relations = outSent.get(MachineReadingAnnotations.RelationMentionsAnnotation.class);
        origSent.set(MachineReadingAnnotations.RelationMentionsAnnotation.class, relations);
        if (verbose && relations != null) {
            log.info("Extracted the following relations:");
            for (RelationMention r : relations) {
                if (!r.getType().equals(RelationMention.UNRELATED)) {
                    log.info(r);
                }
            }
        }
    }
}

Example 12 with EntityMention

use of edu.stanford.nlp.ie.machinereading.structure.EntityMention in project CoreNLP by stanfordnlp.

the class TextOutputter method print.

/**
   * The meat of the outputter
   */
private static void print(Annotation annotation, PrintWriter pw, Options options) throws IOException {
    double beam = options.beamPrintingOption;
    List<CoreMap> sentences = annotation.get(CoreAnnotations.SentencesAnnotation.class);
    // Display docid if available
    String docId = annotation.get(CoreAnnotations.DocIDAnnotation.class);
    if (docId != null) {
        List<CoreLabel> tokens = annotation.get(CoreAnnotations.TokensAnnotation.class);
        int nSentences = (sentences != null) ? sentences.size() : 0;
        int nTokens = (tokens != null) ? tokens.size() : 0;
        pw.printf("Document: ID=%s (%d sentences, %d tokens)%n", docId, nSentences, nTokens);
    }
    // Display doctitle if available
    String docTitle = annotation.get(CoreAnnotations.DocTitleAnnotation.class);
    if (docTitle != null) {
        pw.printf("Document Title: %s%n", docTitle);
    }
    // Display docdate if available
    String docDate = annotation.get(CoreAnnotations.DocDateAnnotation.class);
    if (docDate != null) {
        pw.printf("Document Date: %s%n", docDate);
    }
    // Display doctype if available
    String docType = annotation.get(CoreAnnotations.DocTypeAnnotation.class);
    if (docType != null) {
        pw.printf("Document Type: %s%n", docType);
    }
    // Display docsourcetype if available
    String docSourceType = annotation.get(CoreAnnotations.DocSourceTypeAnnotation.class);
    if (docSourceType != null) {
        pw.printf("Document Source Type: %s%n", docSourceType);
    }
    // display each sentence in this annotation
    if (sentences != null) {
        for (int i = 0, sz = sentences.size(); i < sz; i++) {
            CoreMap sentence = sentences.get(i);
            List<CoreLabel> tokens = sentence.get(CoreAnnotations.TokensAnnotation.class);
            String sentiment = sentence.get(SentimentCoreAnnotations.SentimentClass.class);
            if (sentiment == null) {
                sentiment = "";
            } else {
                sentiment = ", sentiment: " + sentiment;
            }
            pw.printf("Sentence #%d (%d tokens%s):%n", (i + 1), tokens.size(), sentiment);
            String text = sentence.get(CoreAnnotations.TextAnnotation.class);
            pw.println(text);
            // display the token-level annotations
            String[] tokenAnnotations = { "Text", "PartOfSpeech", "Lemma", "Answer", "NamedEntityTag", "CharacterOffsetBegin", "CharacterOffsetEnd", "NormalizedNamedEntityTag", "Timex", "TrueCase", "TrueCaseText", "SentimentClass", "WikipediaEntity" };
            for (CoreLabel token : tokens) {
                pw.print(token.toShorterString(tokenAnnotations));
                pw.println();
            }
            // display the parse tree for this sentence
            Tree tree = sentence.get(TreeCoreAnnotations.TreeAnnotation.class);
            if (tree != null) {
                options.constituentTreePrinter.printTree(tree, pw);
            }
            // language which doesn't have dependencies, for example.
            if (sentence.get(SemanticGraphCoreAnnotations.EnhancedPlusPlusDependenciesAnnotation.class) != null) {
                pw.print(sentence.get(SemanticGraphCoreAnnotations.EnhancedPlusPlusDependenciesAnnotation.class).toList());
                pw.println();
            }
            // display MachineReading entities and relations
            List<EntityMention> entities = sentence.get(MachineReadingAnnotations.EntityMentionsAnnotation.class);
            if (entities != null) {
                pw.println("Extracted the following MachineReading entity mentions:");
                for (EntityMention e : entities) {
                    pw.print('\t');
                    pw.println(e);
                }
            }
            List<RelationMention> relations = sentence.get(MachineReadingAnnotations.RelationMentionsAnnotation.class);
            if (relations != null) {
                pw.println("Extracted the following MachineReading relation mentions:");
                for (RelationMention r : relations) {
                    if (r.printableObject(beam)) {
                        pw.println(r);
                    }
                }
            }
            // display OpenIE triples
            Collection<RelationTriple> openieTriples = sentence.get(NaturalLogicAnnotations.RelationTriplesAnnotation.class);
            if (openieTriples != null && openieTriples.size() > 0) {
                pw.println("Extracted the following Open IE triples:");
                for (RelationTriple triple : openieTriples) {
                    pw.println(OpenIE.tripleToString(triple, docId, sentence));
                }
            }
            // display KBP triples
            Collection<RelationTriple> kbpTriples = sentence.get(CoreAnnotations.KBPTriplesAnnotation.class);
            if (kbpTriples != null && kbpTriples.size() > 0) {
                pw.println("Extracted the following KBP triples:");
                for (RelationTriple triple : kbpTriples) {
                    pw.println(triple.toString());
                }
            }
        }
    }
    // display the old-style doc-level coref annotations
    // this is not supported anymore!
    //String corefAnno = annotation.get(CorefPLAnnotation.class);
    //if(corefAnno != null) os.println(corefAnno);
    // display the new-style coreference graph
    Map<Integer, CorefChain> corefChains = annotation.get(CorefCoreAnnotations.CorefChainAnnotation.class);
    if (corefChains != null && sentences != null) {
        for (CorefChain chain : corefChains.values()) {
            CorefChain.CorefMention representative = chain.getRepresentativeMention();
            boolean outputHeading = false;
            for (CorefChain.CorefMention mention : chain.getMentionsInTextualOrder()) {
                if (mention == representative)
                    continue;
                if (!outputHeading) {
                    outputHeading = true;
                    pw.println("Coreference set:");
                }
                // all offsets start at 1!
                pw.printf("\t(%d,%d,[%d,%d]) -> (%d,%d,[%d,%d]), that is: \"%s\" -> \"%s\"%n", mention.sentNum, mention.headIndex, mention.startIndex, mention.endIndex, representative.sentNum, representative.headIndex, representative.startIndex, representative.endIndex, mention.mentionSpan, representative.mentionSpan);
            }
        }
    }
    // display quotes if available
    if (annotation.get(CoreAnnotations.QuotationsAnnotation.class) != null) {
        pw.println("Extracted quotes: ");
        List<CoreMap> allQuotes = QuoteAnnotator.gatherQuotes(annotation);
        for (CoreMap quote : allQuotes) {
            pw.printf("[QuotationIndexAnnotation=%d, CharacterOffsetBegin=%d, Text=%s]%n", quote.get(CoreAnnotations.QuotationIndexAnnotation.class), quote.get(CoreAnnotations.CharacterOffsetBeginAnnotation.class), quote.get(CoreAnnotations.TextAnnotation.class));
        }
    }
    pw.flush();
}

Example 13 with EntityMention

use of edu.stanford.nlp.ie.machinereading.structure.EntityMention in project CoreNLP by stanfordnlp.

the class AceReader method convertAceEntityMention.

/**
   * Convert an {@link AceEntityMention} to an {@link EntityMention}.
   *
   * @param entityMention {@link AceEntityMention} to convert
   * @param docId ID of the document containing this entity mention
   * @param sentence
   * @param tokenOffset An offset in the calculations of position of the extent to sentence boundary
   *                    (the ace.reader stores absolute token offset from the beginning of the document, but
   *                    we need token offsets from the beginning of the sentence => adjust by tokenOffset)
   * @return entity as an {@link EntityMention}
   */
private EntityMention convertAceEntityMention(AceEntityMention entityMention, String docId, CoreMap sentence, int tokenOffset) {
    //log.info("TYPE is " + entityMention.getParent().getType());
    //log.info("SUBTYPE is " + entityMention.getParent().getSubtype());
    //log.info("LDCTYPE is " + entityMention.getLdctype());
    AceCharSeq ext = entityMention.getExtent();
    AceCharSeq head = entityMention.getHead();
    int extStart = ext.getTokenStart() - tokenOffset;
    int extEnd = ext.getTokenEnd() - tokenOffset + 1;
    if (extStart < 0) {
        logger.severe("READER ERROR: Invalid extent start " + extStart + " for entity mention " + entityMention.getId() + " in document " + docId + " in sentence " + sentence);
        logger.severe("This may happen due to incorrect EOS detection. Adjusting entity extent.");
        extStart = 0;
    }
    if (extEnd > sentence.get(CoreAnnotations.TokensAnnotation.class).size()) {
        logger.severe("READER ERROR: Invalid extent end " + extEnd + " for entity mention " + entityMention.getId() + " in document " + docId + " in sentence " + sentence);
        logger.severe("This may happen due to incorrect EOS detection. Adjusting entity extent.");
        extEnd = sentence.get(CoreAnnotations.TokensAnnotation.class).size();
    }
    int headStart = head.getTokenStart() - tokenOffset;
    int headEnd = head.getTokenEnd() - tokenOffset + 1;
    if (headStart < 0) {
        logger.severe("READER ERROR: Invalid head start " + headStart + " for entity mention " + entityMention.getId() + " in document " + docId + " in sentence " + sentence);
        logger.severe("This may happen due to incorrect EOS detection. Adjusting entity head span.");
        headStart = 0;
    }
    if (headEnd > sentence.get(CoreAnnotations.TokensAnnotation.class).size()) {
        logger.severe("READER ERROR: Invalid head end " + headEnd + " for entity mention " + entityMention.getId() + " in document " + docId + " in sentence " + sentence);
        logger.severe("This may happen due to incorrect EOS detection. Adjusting entity head span.");
        headEnd = sentence.get(CoreAnnotations.TokensAnnotation.class).size();
    }
    // must adjust due to possible incorrect EOS detection
    if (headStart < extStart) {
        headStart = extStart;
    }
    if (headEnd > extEnd) {
        headEnd = extEnd;
    }
    assert (headStart < headEnd);
    // note: the ace.reader stores absolute token offset from the beginning of the document, but
    //       we need token offsets from the beginning of the sentence => adjust by tokenOffset
    // note: in ace.reader the end token position is inclusive, but
    //       in our setup the end token position is exclusive => add 1 to end
    EntityMention converted = new EntityMention(entityMention.getId(), sentence, new Span(extStart, extEnd), new Span(headStart, headEnd), entityMention.getParent().getType(), entityMention.getParent().getSubtype(), entityMention.getLdctype());
    return converted;
}

Example 14 with EntityMention

use of edu.stanford.nlp.ie.machinereading.structure.EntityMention in project cogcomp-nlp by CogComp.

the class StanfordRelationsHandler method addView.

@Override
protected void addView(TextAnnotation ta) throws AnnotatorException {
    Annotation document = new Annotation(ta.text);
    pipeline.annotate(document);
    SpanLabelView vu = new SpanLabelView(viewName, ta);
    for (CoreMap sentence : document.get(CoreAnnotations.SentencesAnnotation.class)) {
        for (RelationMention rm : sentence.get(MachineReadingAnnotations.RelationMentionsAnnotation.class)) {
            if (rm.getType().equals("_NR"))
                continue;
            Map<String, Double> scores = new HashMap<>();
            for (String label : rm.getTypeProbabilities().keySet()) scores.put(label, rm.getTypeProbabilities().getCount(label));
            Constituent c1 = createConstituentGivenMention(rm.getEntityMentionArgs().get(0), ta);
            Constituent c2 = createConstituentGivenMention(rm.getEntityMentionArgs().get(1), ta);
            Relation r = new Relation(scores, c1, c2);
            vu.addRelation(r);
            if (!vu.containsConstituent(c1))
                vu.addConstituent(c1);
            if (!vu.containsConstituent(c2))
                vu.addConstituent(c2);
        }
    }
    for (CoreMap sentence : document.get(CoreAnnotations.SentencesAnnotation.class)) {
        for (EntityMention rm : sentence.get(MachineReadingAnnotations.EntityMentionsAnnotation.class)) {
            Constituent c = createConstituentGivenMention(rm, ta);
            if (!vu.containsConstituent(c))
                vu.addConstituent(c);
        }
    }
    ta.addView(viewName, vu);
}