Examples with RawTextAnnotator com.joliciel.talismane.rawText.RawTextAnnotator used on opensource projects

Example 1 with RawTextAnnotator

use of com.joliciel.talismane.rawText.RawTextAnnotator in project talismane by joliciel-informatique.

the class Talismane method analyse.

/**
 * Analyse the data provided by this reader, as specified by the
 * configuration.
 *
 * @param reader
 * @throws IOException
 * @throws ReflectiveOperationException
 * @throws TalismaneException
 *           if it's impossible to read a sentence from an annotated corpus
 */
public void analyse(Reader reader) throws IOException, ReflectiveOperationException, TalismaneException {
    long startTime = System.currentTimeMillis();
    try {
        TokeniserAnnotatedCorpusReader tokenCorpusReader = null;
        PosTagAnnotatedCorpusReader posTagCorpusReader = null;
        if (this.startModule.equals(Module.posTagger)) {
            tokenCorpusReader = TokeniserAnnotatedCorpusReader.getCorpusReader(reader, config.getConfig("talismane.core." + sessionId + ".tokeniser.input"), sessionId);
        }
        if (this.startModule.equals(Module.parser)) {
            posTagCorpusReader = PosTagAnnotatedCorpusReader.getCorpusReader(reader, config.getConfig("talismane.core." + sessionId + ".pos-tagger.input"), sessionId);
        }
        LinkedList<String> textSegments = new LinkedList<String>();
        LinkedList<Sentence> sentences = new LinkedList<Sentence>();
        TokenSequence tokenSequence = null;
        PosTagSequence posTagSequence = null;
        StringBuilder stringBuilder = new StringBuilder();
        boolean finished = false;
        int sentenceCount = 0;
        CurrentFileProvider currentFileProvider = reader instanceof CurrentFileProvider ? (CurrentFileProvider) reader : null;
        RollingTextBlock rollingTextBlock = new RollingTextBlock(this.processByDefault, currentFileProvider, sessionId);
        int endBlockCharacterCount = 0;
        URI currentURI = null;
        File currentFile = null;
        while (!finished) {
            if (this.startModule.equals(Module.sentenceDetector) || this.startModule.equals(Module.tokeniser)) {
                // Note SentenceDetector and Tokeniser start modules treated
                // identically,
                // except that for SentenceDetector we apply a probabilistic
                // sentence detector
                // whereas for Tokeniser we assume all sentence breaks are
                // marked by filters
                // read characters from the reader, one at a time
                char c;
                int r = -1;
                try {
                    r = reader.read();
                } catch (IOException e) {
                    LogUtils.logError(LOG, e);
                }
                if (r == -1) {
                    finished = true;
                    c = '\n';
                } else {
                    c = (char) r;
                }
                // Jump out if we have 3 consecutive end-block characters.
                if (c == TalismaneSession.get(sessionId).getEndBlockCharacter()) {
                    endBlockCharacterCount++;
                    if (endBlockCharacterCount == 3) {
                        LOG.info("Three consecutive end-block characters. Exiting.");
                        finished = true;
                    }
                } else {
                    endBlockCharacterCount = 0;
                }
                // have sentence detector
                if (finished || (Character.isWhitespace(c) && c != '\r' && c != '\n' && stringBuilder.length() > TalismaneSession.get(sessionId).getBlockSize()) || c == TalismaneSession.get(sessionId).getEndBlockCharacter()) {
                    if (c == TalismaneSession.get(sessionId).getEndBlockCharacter())
                        stringBuilder.append(c);
                    if (stringBuilder.length() > 0) {
                        String textSegment = stringBuilder.toString();
                        stringBuilder = new StringBuilder();
                        textSegments.add(textSegment);
                    }
                    // is the current block > 0 characters?
                    if (c == TalismaneSession.get(sessionId).getEndBlockCharacter()) {
                        textSegments.addLast("");
                    }
                }
                if (finished) {
                    if (stringBuilder.length() > 0) {
                        textSegments.addLast(stringBuilder.toString());
                        stringBuilder = new StringBuilder();
                    }
                    // add three final text segments to roll everything
                    // through processing
                    textSegments.addLast("");
                    textSegments.addLast("");
                    textSegments.addLast("");
                }
                if (c != TalismaneSession.get(sessionId).getEndBlockCharacter())
                    stringBuilder.append(c);
                while (textSegments.size() > 0) {
                    // roll in a new block 4, and roll the other blocks
                    // leftwards
                    String nextText = textSegments.removeFirst();
                    rollingTextBlock = rollingTextBlock.roll(nextText);
                    // annotate block 3 with raw text filters
                    AnnotatedText rawTextBlock = rollingTextBlock.getRawTextBlock();
                    for (RawTextAnnotator textAnnotator : TalismaneSession.get(sessionId).getTextAnnotators()) {
                        textAnnotator.annotate(rawTextBlock);
                    }
                    // detect sentences in block 2 using the sentence
                    // detector
                    AnnotatedText processedText = rollingTextBlock.getProcessedText();
                    if (LOG.isTraceEnabled()) {
                        LOG.trace("processedText: " + processedText.getText().toString().replace('\n', '¶').replace('\r', '¶'));
                    }
                    if (this.startModule.equals(Module.sentenceDetector)) {
                        sentenceDetector.detectSentences(processedText);
                    }
                    // get the sentences detected in block 2
                    List<Sentence> theSentences = rollingTextBlock.getDetectedSentences();
                    for (Sentence sentence : theSentences) {
                        sentences.add(sentence);
                        sentenceCount++;
                    }
                    if (this.sentenceCount > 0 && sentenceCount >= this.sentenceCount) {
                        finished = true;
                    }
                }
            // we have at least one text segment to process
            } else if (this.startModule.equals(Module.posTagger)) {
                if (tokenCorpusReader.hasNextSentence()) {
                    tokenSequence = tokenCorpusReader.nextTokenSequence();
                } else {
                    tokenSequence = null;
                    finished = true;
                }
            } else if (this.startModule.equals(Module.parser)) {
                if (posTagCorpusReader.hasNextSentence()) {
                    posTagSequence = posTagCorpusReader.nextPosTagSequence();
                } else {
                    posTagSequence = null;
                    finished = true;
                }
            }
            // which start module?
            boolean needToProcess = false;
            if (this.startModule.equals(Module.sentenceDetector) || this.startModule.equals(Module.tokeniser))
                needToProcess = !sentences.isEmpty();
            else if (this.startModule.equals(Module.posTagger))
                needToProcess = tokenSequence != null;
            else if (this.startModule.equals(Module.parser))
                needToProcess = posTagSequence != null;
            while (needToProcess) {
                Sentence sentence = null;
                if (this.startModule.compareTo(Module.tokeniser) <= 0 && this.endModule.compareTo(Module.sentenceDetector) >= 0) {
                    sentence = sentences.poll();
                    LOG.debug("Sentence: " + sentence);
                    for (SentenceAnnotator annotator : TalismaneSession.get(sessionId).getSentenceAnnotators()) annotator.annotate(sentence);
                    if (sentence.getFileURI() != null && !sentence.getFileURI().equals(currentURI)) {
                        currentURI = sentence.getFileURI();
                        currentFile = sentence.getFile();
                        LOG.debug("Setting current file to " + currentFile.getPath());
                        if (writer instanceof CurrentFileObserver)
                            ((CurrentFileObserver) writer).onNextFile(currentFile);
                        for (SentenceProcessor processor : sentenceProcessors) if (processor instanceof CurrentFileObserver)
                            ((CurrentFileObserver) processor).onNextFile(currentFile);
                        for (TokenSequenceProcessor processor : tokenSequenceProcessors) if (processor instanceof CurrentFileObserver)
                            ((CurrentFileObserver) processor).onNextFile(currentFile);
                        for (PosTagSequenceProcessor processor : posTagSequenceProcessors) if (processor instanceof CurrentFileObserver)
                            ((CurrentFileObserver) processor).onNextFile(currentFile);
                        for (ParseConfigurationProcessor processor : parseConfigurationProcessors) if (processor instanceof CurrentFileObserver)
                            ((CurrentFileObserver) processor).onNextFile(currentFile);
                    }
                    if (sentence.getLeftoverOriginalText().length() > 0) {
                        writer.append(sentence.getLeftoverOriginalText() + "\n");
                    }
                    for (SentenceProcessor sentenceProcessor : sentenceProcessors) {
                        sentenceProcessor.onNextSentence(sentence);
                    }
                }
                // need to read next sentence
                List<TokenSequence> tokenSequences = null;
                if (this.needsTokeniser()) {
                    tokenSequences = tokeniser.tokenise(sentence);
                    tokenSequence = tokenSequences.get(0);
                    for (TokenSequenceProcessor tokenSequenceProcessor : tokenSequenceProcessors) {
                        tokenSequenceProcessor.onNextTokenSequence(tokenSequence);
                    }
                }
                // need to tokenise ?
                List<PosTagSequence> posTagSequences = null;
                if (this.needsPosTagger()) {
                    posTagSequence = null;
                    if (tokenSequences == null) {
                        tokenSequences = new ArrayListNoNulls<>();
                        tokenSequences.add(tokenSequence);
                    }
                    if (posTagger instanceof NonDeterministicPosTagger) {
                        NonDeterministicPosTagger nonDeterministicPosTagger = (NonDeterministicPosTagger) posTagger;
                        posTagSequences = nonDeterministicPosTagger.tagSentence(tokenSequences);
                        posTagSequence = posTagSequences.get(0);
                    } else {
                        posTagSequence = posTagger.tagSentence(tokenSequence);
                    }
                    for (PosTagSequenceProcessor posTagSequenceProcessor : this.posTagSequenceProcessors) {
                        posTagSequenceProcessor.onNextPosTagSequence(posTagSequence);
                    }
                    tokenSequence = null;
                }
                if (this.needsParser()) {
                    if (posTagSequences == null) {
                        posTagSequences = new ArrayListNoNulls<>();
                        posTagSequences.add(posTagSequence);
                    }
                    ParseConfiguration parseConfiguration = null;
                    List<ParseConfiguration> parseConfigurations = null;
                    try {
                        if (parser instanceof NonDeterministicParser) {
                            NonDeterministicParser nonDeterministicParser = (NonDeterministicParser) parser;
                            parseConfigurations = nonDeterministicParser.parseSentence(posTagSequences);
                            parseConfiguration = parseConfigurations.get(0);
                        } else {
                            parseConfiguration = parser.parseSentence(posTagSequence);
                        }
                        for (ParseConfigurationProcessor parseConfigurationProcessor : this.parseConfigurationProcessors) {
                            parseConfigurationProcessor.onNextParseConfiguration(parseConfiguration);
                        }
                    } catch (Exception e) {
                        LogUtils.logError(LOG, e);
                        if (stopOnError)
                            throw new RuntimeException(e);
                    }
                    posTagSequence = null;
                }
                if (this.startModule.equals(Module.sentenceDetector) || this.startModule.equals(Module.tokeniser))
                    needToProcess = !sentences.isEmpty();
                else if (this.startModule.equals(Module.posTagger))
                    needToProcess = tokenSequence != null;
                else if (this.startModule.equals(Module.parser))
                    needToProcess = posTagSequence != null;
            }
        // next sentence
        }
        // Check if there's any leftover output to output!
        if (rollingTextBlock.getLeftoverOriginalText().length() > 0)
            writer.append(rollingTextBlock.getLeftoverOriginalText());
    } finally {
        IOException exception = null;
        try {
            reader.close();
            writer.flush();
        } catch (IOException e) {
            LogUtils.logError(LOG, e);
            exception = e;
        }
        for (SentenceProcessor processor : this.sentenceProcessors) try {
            processor.close();
        } catch (IOException e) {
            LogUtils.logError(LOG, e);
            exception = e;
        }
        for (TokenSequenceProcessor processor : this.tokenSequenceProcessors) try {
            processor.close();
        } catch (IOException e) {
            LogUtils.logError(LOG, e);
            exception = e;
        }
        for (PosTagSequenceProcessor processor : this.posTagSequenceProcessors) {
            try {
                processor.onCompleteAnalysis();
                processor.close();
            } catch (IOException e) {
                LogUtils.logError(LOG, e);
                exception = e;
            }
        }
        for (ParseConfigurationProcessor processor : this.parseConfigurationProcessors) {
            try {
                processor.onCompleteParse();
                processor.close();
            } catch (IOException e) {
                LogUtils.logError(LOG, e);
                exception = e;
            }
        }
        long endTime = System.currentTimeMillis();
        long totalTime = endTime - startTime;
        LOG.debug("Total time for Talismane.process(): " + totalTime);
        try {
            writer.close();
        } catch (IOException e) {
            LogUtils.logError(LOG, e);
            exception = e;
        }
        if (exception != null)
            throw exception;
    }
}

Example 2 with RawTextAnnotator

use of com.joliciel.talismane.rawText.RawTextAnnotator in project talismane by joliciel-informatique.

the class TalismaneAPIExamples method example2.

/**
 * Similar to example1, but begins with filtering and sentence detection.
 */
public static void example2(String sessionId) throws Exception {
    String text = "Les gens qui voient de travers pensent que les bancs verts qu'on voit sur les trottoirs " + "sont faits pour les impotents ou les ventripotents. " + "Mais c'est une absurdité, car, à la vérité, ils sont là, c'est notoire, " + "pour accueillir quelque temps les amours débutants.";
    RawText rawText = new RawText(text, true, sessionId);
    // issues (e.g. replace " with ")
    for (RawTextAnnotator filter : TalismaneSession.get(sessionId).getTextAnnotators()) {
        filter.annotate(rawText);
    }
    // retrieve the processed text after filters have been applied
    AnnotatedText processedText = rawText.getProcessedText();
    // detect sentences
    SentenceDetector sentenceDetector = SentenceDetector.getInstance(sessionId);
    sentenceDetector.detectSentences(processedText);
    // the detected sentences can be retrieved directly from the raw text
    // this allows annotations made on the sentences to get reflected in the
    // raw text
    List<Sentence> sentences = rawText.getDetectedSentences();
    for (Sentence sentence : sentences) {
        // assignment for a given word)
        for (SentenceAnnotator annotator : TalismaneSession.get(sessionId).getSentenceAnnotators()) {
            annotator.annotate(sentence);
        }
        // tokenise the text
        Tokeniser tokeniser = Tokeniser.getInstance(sessionId);
        TokenSequence tokenSequence = tokeniser.tokeniseSentence(sentence);
        // pos-tag the token sequence
        PosTagger posTagger = PosTaggers.getPosTagger(sessionId);
        PosTagSequence posTagSequence = posTagger.tagSentence(tokenSequence);
        System.out.println(posTagSequence);
        // parse the pos-tag sequence
        Parser parser = Parsers.getParser(sessionId);
        ParseConfiguration parseConfiguration = parser.parseSentence(posTagSequence);
        System.out.println(parseConfiguration);
        ParseTree parseTree = new ParseTree(parseConfiguration, true);
        System.out.println(parseTree);
    }
}