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

Example 1 with ExtractionObject

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

the class ProtobufAnnotationSerializer method fromProto.

/**
   * Read a relation mention from its serialized form. Requires the containing sentence to be
   * passed in along with the protocol buffer.
   * @param proto The serialized relation mention.
   * @param sentence The sentence this mention is attached to.
   * @return The relation mention corresponding to the serialized object.
   */
private RelationMention fromProto(CoreNLPProtos.Relation proto, CoreMap sentence) {
    List<ExtractionObject> args = proto.getArgList().stream().map(arg -> fromProto(arg, sentence)).collect(Collectors.toList());
    RelationMention rtn = new RelationMention(proto.hasObjectID() ? proto.getObjectID() : null, sentence, proto.hasExtentStart() ? new Span(proto.getExtentStart(), proto.getExtentEnd()) : null, proto.hasType() ? proto.getType() : null, proto.hasSubtype() ? proto.getSubtype() : null, args);
    if (proto.hasSignature()) {
        rtn.setSignature(proto.getSignature());
    }
    if (proto.getArgNameCount() > 0 || proto.getArgCount() == 0) {
        rtn.setArgNames(proto.getArgNameList());
    }
    return rtn;
}

Example 2 with ExtractionObject

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

the class RothCONLL04Reader method readSentence.

private Annotation readSentence(String docId, Iterator<String> lineIterator) {
    Annotation sentence = new Annotation("");
    sentence.set(CoreAnnotations.DocIDAnnotation.class, docId);
    sentence.set(MachineReadingAnnotations.EntityMentionsAnnotation.class, new ArrayList<>());
    // we'll need to set things like the tokens and textContent after we've
    // fully read the sentence
    // contains the full text that we've read so far
    StringBuilder textContent = new StringBuilder();
    // how many tokens we've seen so far
    int tokenCount = 0;
    List<CoreLabel> tokens = new ArrayList<>();
    // when we've seen two blank lines in a row, this sentence is over (one
    // blank line separates the sentence and the relations
    int numBlankLinesSeen = 0;
    String sentenceID = null;
    // keeps tracks of entities we've seen so far for use by relations
    Map<String, EntityMention> indexToEntityMention = new HashMap<>();
    while (lineIterator.hasNext() && numBlankLinesSeen < 2) {
        String currentLine = lineIterator.next();
        currentLine = currentLine.replace("COMMA", ",");
        List<String> pieces = StringUtils.split(currentLine);
        String identifier;
        int size = pieces.size();
        switch(size) {
            case // blank line between sentences or relations
            1:
                numBlankLinesSeen++;
                break;
            case // relation
            3:
                String type = pieces.get(2);
                List<ExtractionObject> args = new ArrayList<>();
                EntityMention entity1 = indexToEntityMention.get(pieces.get(0));
                EntityMention entity2 = indexToEntityMention.get(pieces.get(1));
                args.add(entity1);
                args.add(entity2);
                Span span = new Span(entity1.getExtentTokenStart(), entity2.getExtentTokenEnd());
                // identifier = "relation" + sentenceID + "-" + sentence.getAllRelations().size();
                identifier = RelationMention.makeUniqueId();
                RelationMention relationMention = new RelationMention(identifier, sentence, span, type, null, args);
                AnnotationUtils.addRelationMention(sentence, relationMention);
                break;
            case // token
            9:
                /*
         * Roth token lines look like this:
         *
         * 19 Peop 9 O NNP/NNP Jamal/Ghosheh O O O
         */
                // Entities may be multiple words joined by '/'; we split these up
                List<String> words = StringUtils.split(pieces.get(5), "/");
                //List<String> postags = StringUtils.split(pieces.get(4),"/");
                String text = StringUtils.join(words, " ");
                identifier = "entity" + pieces.get(0) + '-' + pieces.get(2);
                // entity type of the word/expression
                String nerTag = getNormalizedNERTag(pieces.get(1));
                if (sentenceID == null)
                    sentenceID = pieces.get(0);
                if (!nerTag.equals("O")) {
                    Span extentSpan = new Span(tokenCount, tokenCount + words.size());
                    // Temporarily sets the head span to equal the extent span.
                    // This is so the entity has a head (in particular, getValue() works) even if preprocessSentences isn't called.
                    // The head span is later modified if preprocessSentences is called.
                    EntityMention entity = new EntityMention(identifier, sentence, extentSpan, extentSpan, nerTag, null, null);
                    AnnotationUtils.addEntityMention(sentence, entity);
                    // we can get by using these indices as strings since we only use them
                    // as a hash key
                    String index = pieces.get(2);
                    indexToEntityMention.put(index, entity);
                }
                // int i =0;
                for (String word : words) {
                    CoreLabel label = new CoreLabel();
                    label.setWord(word);
                    //label.setTag(postags.get(i));
                    label.set(CoreAnnotations.TextAnnotation.class, word);
                    label.set(CoreAnnotations.ValueAnnotation.class, word);
                    // we don't set TokenBeginAnnotation or TokenEndAnnotation since we're
                    // not keeping track of character offsets
                    tokens.add(label);
                // i++;
                }
                textContent.append(text);
                textContent.append(' ');
                tokenCount += words.size();
                break;
        }
    }
    sentence.set(CoreAnnotations.TextAnnotation.class, textContent.toString());
    sentence.set(CoreAnnotations.ValueAnnotation.class, textContent.toString());
    sentence.set(CoreAnnotations.TokensAnnotation.class, tokens);
    sentence.set(CoreAnnotations.SentenceIDAnnotation.class, sentenceID);
    return sentence;
}

Example 3 with ExtractionObject

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

the class AceReader method convertAceRelationMention.

private RelationMention convertAceRelationMention(AceRelationMention aceRelationMention, String docId, CoreMap sentence, Map<String, EntityMention> entityMap) {
    List<AceRelationMentionArgument> args = Arrays.asList(aceRelationMention.getArgs());
    List<ExtractionObject> convertedArgs = new ArrayList<>();
    List<String> argNames = new ArrayList<>();
    // the arguments are already stored in semantic order. Make sure we preserve the same ordering!
    int left = Integer.MAX_VALUE;
    int right = Integer.MIN_VALUE;
    for (AceRelationMentionArgument arg : args) {
        ExtractionObject o = entityMap.get(arg.getContent().getId());
        if (o == null) {
            logger.severe("READER ERROR: Failed to find relation argument with id " + arg.getContent().getId());
            logger.severe("This happens because a few relation mentions illegally span multiple sentences. Will ignore this mention.");
            return null;
        }
        convertedArgs.add(o);
        argNames.add(arg.getRole());
        if (o.getExtentTokenStart() < left)
            left = o.getExtentTokenStart();
        if (o.getExtentTokenEnd() > right)
            right = o.getExtentTokenEnd();
    }
    if (argNames.size() != 2 || !argNames.get(0).equalsIgnoreCase("arg-1") || !argNames.get(1).equalsIgnoreCase("arg-2")) {
        logger.severe("READER ERROR: Invalid succession of arguments in relation mention: " + argNames);
        logger.severe("ACE relations must have two arguments. Will ignore this mention.");
        return null;
    }
    RelationMention relation = new RelationMention(aceRelationMention.getId(), sentence, new Span(left, right), aceRelationMention.getParent().getType(), aceRelationMention.getParent().getSubtype(), convertedArgs, null);
    return relation;
}

Example 4 with ExtractionObject

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

the class AceReader method convertAceEventMention.

private EventMention convertAceEventMention(AceEventMention aceEventMention, String docId, CoreMap sentence, Map<String, EntityMention> entityMap, int tokenOffset) {
    Set<String> roleSet = aceEventMention.getRoles();
    List<String> roles = new ArrayList<>();
    for (String role : roleSet) roles.add(role);
    List<ExtractionObject> convertedArgs = new ArrayList<>();
    int left = Integer.MAX_VALUE;
    int right = Integer.MIN_VALUE;
    for (String role : roles) {
        AceEntityMention arg = aceEventMention.getArg(role);
        ExtractionObject o = entityMap.get(arg.getId());
        if (o == null) {
            logger.severe("READER ERROR: Failed to find event argument with id " + arg.getId());
            logger.severe("This happens because a few event mentions illegally span multiple sentences. Will ignore this mention.");
            return null;
        }
        convertedArgs.add(o);
        if (o.getExtentTokenStart() < left)
            left = o.getExtentTokenStart();
        if (o.getExtentTokenEnd() > right)
            right = o.getExtentTokenEnd();
    }
    AceCharSeq anchor = aceEventMention.getAnchor();
    ExtractionObject anchorObject = new ExtractionObject(aceEventMention.getId() + "-anchor", sentence, new Span(anchor.getTokenStart() - tokenOffset, anchor.getTokenEnd() + 1 - tokenOffset), "ANCHOR", null);
    EventMention em = new EventMention(aceEventMention.getId(), sentence, new Span(left, right), aceEventMention.getParent().getType(), aceEventMention.getParent().getSubtype(), anchorObject, convertedArgs, roles);
    return em;
}