Examples of org.apache.stanbol.enhancer.engines.sentiment.api.SentimentClassifier

Package org.apache.stanbol.enhancer.engines.sentiment.api

Examples of org.apache.stanbol.enhancer.engines.sentiment.api.SentimentClassifier

org.apache.stanbol.enhancer.engines.sentiment.api.SentimentClassifier
A simple classifier that determines for a given word a positive or negative value between -1 and 1. Unknown words will return the value 0.
This Interface need to be implemented by Sentiment frameworks so that they can be used with this engine. Implementations need to be {@link BundleContext#registerService(String,Object,java.util.Dictionary) registered as OSGI service}. @see LexicalCategoryClassifier @author Sebastian Schaffert @author Rupert Westenthaler

    /** bind method for {@link #classifiers} */
    protected void bindClassifier(SentimentClassifier classifier){
        log.info("  ... bind Sentiment Classifier {} for language {}",
            classifier.getClass().getSimpleName(),classifier.getLanguage());
        synchronized (classifiers) {
            SentimentClassifier old = classifiers.put(classifier.getLanguage(), classifier);
            if(old != null){
                log.warn("Replaced Sentiment Classifier for language {} (old: {}, new: {}",
                    new Object[]{old.getLanguage(),old,classifier});
            }
        }
    }

View Full Code Here

    }
    /** unbind method for {@link #classifiers} */
    protected void unbindClassifier(SentimentClassifier classifier){
        String lang = classifier.getLanguage();
        synchronized (classifiers) {
            SentimentClassifier current = classifiers.remove(lang);
            if(!classifier.equals(current) //the current is not the parsed one
                    && current != null){
                classifiers.put(lang,current); //re-add the value
            } else {
                log.info("  ... unbind Sentiment Classifier {} for language {}",

View Full Code Here

     */
    @Override
    public void computeEnhancements(ContentItem ci) throws EngineException {
        AnalysedText analysedText = getAnalysedText(this,ci, true);
        String language = getLanguage(this, ci, true);
        SentimentClassifier classifier = classifiers.get(language);
        if(classifier == null){
            throw new IllegalStateException("Sentiment Classifier for language '"
                + language +"' not available. As this is also checked in "
                + " canEnhance this may indicate an Bug in the used "
                + "EnhancementJobManager!");
        }
        //TODO: locking for AnalysedText not yet defined
//        ci.getLock().writeLock().lock();
//        try {
        Iterator<Token> tokens = analysedText.getTokens();
        while(tokens.hasNext()){
            Token token = tokens.next();
            Set<LexicalCategory> cats = null;
            boolean process = false;
            if(!adjectivesOnly){
                process = true;
                Value<PosTag> posTag = token.getAnnotation(NlpAnnotations.POS_ANNOTATION);
                if(posTag != null && posTag.probability() == Value.UNKNOWN_PROBABILITY
                        || posTag.probability() >= (minPOSConfidence/2.0)){
                    cats = classifier.getCategories(posTag.value());
                } else { //no POS tags or probability to low
                    cats = Collections.emptySet();
                }
            } else { //check PosTags if we need to lookup this word
                Iterator<Value<PosTag>> posTags = token.getAnnotations(NlpAnnotations.POS_ANNOTATION).iterator();
                boolean ignore = false;
                while(!ignore && !process && posTags.hasNext()) {
                    Value<PosTag> value = posTags.next();
                    PosTag tag = value.value();
                    cats = classifier.getCategories(tag);
                    boolean state = cats.contains(LexicalCategory.Adjective) 
                            || cats.contains(LexicalCategory.Noun);
                    ignore = !state && (value.probability() == Value.UNKNOWN_PROBABILITY ||
                            value.probability() >= minPOSConfidence);
                    process = state && (value.probability() == Value.UNKNOWN_PROBABILITY ||
                            value.probability() >= (minPOSConfidence/2.0));
                }
            } //else process all tokens ... no POS tag checking needed
            if(process){
                String word = token.getSpan();
                double sentiment = 0.0;
                if(cats.isEmpty()){
                    sentiment = classifier.classifyWord(null, word);
                } else { //in case of multiple Lexical Cats
                    //we build the average over NOT NULL sentiments for the word
                    int catSentNum = 0;
                    for(LexicalCategory cat : cats){
                        double catSent = classifier.classifyWord(cat, word);
                        if(catSent != 0.0){
                            catSentNum++;
                            sentiment = sentiment + catSent;
                        }
                    }

View Full Code Here

    /** bind method for {@link #classifiers} */
    protected void bindClassifier(SentimentClassifier classifier){
        log.info("  ... bind Sentiment Classifier {} for language {}",
            classifier.getClass().getSimpleName(),classifier.getLanguage());
        synchronized (classifiers) {
            SentimentClassifier old = classifiers.put(classifier.getLanguage(), classifier);
            if(old != null){
                log.warn("Replaced Sentiment Classifier for language {} (old: {}, new: {}",
                    new Object[]{old.getLanguage(),old,classifier});
            }
        }
    }

View Full Code Here

    }
    /** unbind method for {@link #classifiers} */
    protected void unbindClassifier(SentimentClassifier classifier){
        String lang = classifier.getLanguage();
        synchronized (classifiers) {
            SentimentClassifier current = classifiers.remove(lang);
            if(!classifier.equals(current) //the current is not the parsed one
                    && current != null){
                classifiers.put(lang,current); //re-add the value
            } else {
                log.info("  ... unbind Sentiment Classifier {} for language {}",

View Full Code Here

     */
    @Override
    public void computeEnhancements(ContentItem ci) throws EngineException {
        AnalysedText analysedText = getAnalysedText(this,ci, true);
        String language = getLanguage(this, ci, true);
        SentimentClassifier classifier = classifiers.get(language);
        if(classifier == null){
            throw new IllegalStateException("Sentiment Classifier for language '"
                + language +"' not available. As this is also checked in "
                + " canEnhance this may indicate an Bug in the used "
                + "EnhancementJobManager!");
        }
        //TODO: locking for AnalysedText not yet defined
//        ci.getLock().writeLock().lock();
//        try {
        Iterator<Token> tokens = analysedText.getTokens();
        while(tokens.hasNext()){
            Token token = tokens.next();
            boolean process = !adjectivesOnly;
            if(!process){ //check POS types
                Iterator<Value<PosTag>> posTags = token.getAnnotations(NlpAnnotations.POS_ANNOTATION).iterator();
                boolean ignore = false;
                while(!ignore && !process && posTags.hasNext()) {
                    Value<PosTag> value = posTags.next();
                    PosTag tag = value.value();
                    boolean state = classifier.isAdjective(tag) || classifier.isNoun(tag);
                    ignore = !state && value.probability() >= minPOSConfidence;
                    process = state && value.probability() >= (minPOSConfidence/2.0);
                }
            } //else process all tokens ... no POS tag checking needed
            if(process){
                double sentiment = classifier.classifyWord(token.getSpan());
                if(sentiment != 0.0){
                    token.addAnnotation(SENTIMENT_ANNOTATION, new Value<Double>(sentiment));
                } //else do not set sentiments with 0.0
            }
        }

View Full Code Here

TOP

Related Classes of org.apache.stanbol.enhancer.engines.sentiment.api.SentimentClassifier

org.apache.stanbol.enhancer.engines.sentiment.services.SentimentEngine

All source code are property of their respective owners. Java is a trademark of Sun Microsystems, Inc and owned by ORACLE Inc. Contact coftware#gmail.com.