Examples with RawTextNoSentenceBreakMarker com.joliciel.talismane.rawText.RawTextMarker.RawTextNoSentenceBreakMarker used on opensource projects

Example 1 with RawTextNoSentenceBreakMarker

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

the class SentenceDetector method detectSentences.

/**
 * Detect sentences within an annotated text. Sentences are added in the form
 * of an Annotation around a {@link SentenceBoundary}, with the start position
 * (relative to the start of the annotated text) at the start of the sentence
 * and the end position immediately after the end of the sentence. <br>
 * <br>
 * Sentence boundaries will not be detected within any annotation of type
 * {@link RawTextNoSentenceBreakMarker}, nor will they be detected before or
 * after the {@link AnnotatedText#getAnalysisStart()} and
 * {@link AnnotatedText#getAnalysisEnd()} respectively. <br>
 * <br>
 * If the text contained existing {@link SentenceBoundary} annotations before
 * analysis start, the first sentence will begin where the last existing
 * annotation ended. Otherwise, the first boundary will begin at position 0.
 * <br>
 * <br>
 * If the text's analysis end is equal to the text length, it is assumed that
 * the text end is a sentence boundary. In this case, an additional sentence
 * is added starting at the final detected boundary and ending at text end.
 *
 * @param text
 *          the annotated text in which we need to detect sentences.
 * @return in addition to the annotations added, we return a List of integers
 *         marking the end position of each sentence boundary.
 */
public List<Integer> detectSentences(AnnotatedText text, String... labels) throws TalismaneException {
    LOG.debug("detectSentences");
    List<Annotation<RawTextNoSentenceBreakMarker>> noSentenceBreakMarkers = text.getAnnotations(RawTextNoSentenceBreakMarker.class);
    Matcher matcher = possibleBoundaryPattern.matcher(text.getText());
    List<Integer> possibleBoundaries = new ArrayList<>();
    while (matcher.find()) {
        if (matcher.start() >= text.getAnalysisStart() && matcher.start() < text.getAnalysisEnd()) {
            boolean noSentences = false;
            int position = matcher.start();
            for (Annotation<RawTextNoSentenceBreakMarker> noSentenceBreakMarker : noSentenceBreakMarkers) {
                if (noSentenceBreakMarker.getStart() <= position && position < noSentenceBreakMarker.getEnd()) {
                    noSentences = true;
                    break;
                }
            }
            if (!noSentences)
                possibleBoundaries.add(position);
        }
    }
    // collect all deterministic sentence boundaries
    List<Annotation<RawTextSentenceBreakMarker>> sentenceBreakMarkers = text.getAnnotations(RawTextSentenceBreakMarker.class);
    Set<Integer> guessedBoundaries = new TreeSet<>(sentenceBreakMarkers.stream().filter(f -> f.getEnd() >= text.getAnalysisStart()).map(f -> f.getEnd()).collect(Collectors.toList()));
    // Share one token sequence for all possible boundaries, to avoid tokenising
    // multiple times
    Sentence sentence = new Sentence(text.getText(), sessionId);
    TokenSequence tokenSequence = new TokenSequence(sentence, sessionId);
    List<PossibleSentenceBoundary> boundaries = new ArrayList<>();
    for (int possibleBoundary : possibleBoundaries) {
        PossibleSentenceBoundary boundary = new PossibleSentenceBoundary(tokenSequence, possibleBoundary);
        if (LOG.isTraceEnabled()) {
            LOG.trace("Testing boundary: " + boundary);
            LOG.trace(" at position: " + possibleBoundary);
        }
        List<FeatureResult<?>> featureResults = new ArrayList<>();
        for (SentenceDetectorFeature<?> feature : features) {
            RuntimeEnvironment env = new RuntimeEnvironment();
            FeatureResult<?> featureResult = feature.check(boundary, env);
            if (featureResult != null)
                featureResults.add(featureResult);
        }
        if (LOG.isTraceEnabled()) {
            SortedSet<String> featureResultSet = featureResults.stream().map(f -> f.toString()).collect(Collectors.toCollection(() -> new TreeSet<String>()));
            for (String featureResultString : featureResultSet) {
                LOG.trace(featureResultString);
            }
        }
        List<Decision> decisions = this.decisionMaker.decide(featureResults);
        if (LOG.isTraceEnabled()) {
            for (Decision decision : decisions) {
                LOG.trace(decision.getOutcome() + ": " + decision.getProbability());
            }
        }
        if (decisions.get(0).getOutcome().equals(SentenceDetectorOutcome.IS_BOUNDARY.name())) {
            if (LOG.isTraceEnabled()) {
                LOG.trace("Adding boundary: " + possibleBoundary + 1);
            }
            guessedBoundaries.add(possibleBoundary + 1);
            boundaries.add(boundary);
        }
    }
    if (LOG.isTraceEnabled()) {
        LOG.trace("context: " + text.getText().toString().replace('\n', '¶').replace('\r', '¶'));
        for (PossibleSentenceBoundary boundary : boundaries) LOG.trace("boundary: " + boundary.toString());
    }
    if (LOG.isDebugEnabled())
        LOG.debug("guessedBoundaries : " + guessedBoundaries.toString());
    List<Annotation<SentenceBoundary>> newBoundaries = new ArrayList<>();
    int lastBoundary = 0;
    List<Annotation<SentenceBoundary>> existingBoundaries = text.getAnnotations(SentenceBoundary.class);
    if (existingBoundaries.size() > 0) {
        lastBoundary = existingBoundaries.get(existingBoundaries.size() - 1).getEnd();
    }
    // advance boundary start until a non space character is encountered
    while (lastBoundary < text.getAnalysisEnd() && Character.isWhitespace(text.getText().charAt(lastBoundary))) {
        lastBoundary++;
    }
    for (int guessedBoundary : guessedBoundaries) {
        if (guessedBoundary > lastBoundary) {
            Annotation<SentenceBoundary> sentenceBoundary = new Annotation<>(lastBoundary, guessedBoundary, new SentenceBoundary(), labels);
            newBoundaries.add(sentenceBoundary);
            if (LOG.isTraceEnabled()) {
                LOG.trace("Added boundary: " + sentenceBoundary);
            }
            lastBoundary = guessedBoundary;
        }
    }
    if (text.getAnalysisEnd() == text.getText().length()) {
        if (text.getAnalysisEnd() > lastBoundary) {
            Annotation<SentenceBoundary> sentenceBoundary = new Annotation<>(lastBoundary, text.getAnalysisEnd(), new SentenceBoundary(), labels);
            newBoundaries.add(sentenceBoundary);
            if (LOG.isTraceEnabled()) {
                LOG.trace("Added final boundary: " + sentenceBoundary);
            }
        }
    }
    text.addAnnotations(newBoundaries);
    return new ArrayList<>(guessedBoundaries);
}

Example 2 with RawTextNoSentenceBreakMarker

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

the class RawTextTest method testNoSentenceAnnotationLocation.

@Test
public void testNoSentenceAnnotationLocation() throws Exception {
    System.setProperty("config.file", "src/test/resources/test.conf");
    ConfigFactory.invalidateCaches();
    final Config config = ConfigFactory.load();
    final String sessionId = "test";
    String[] labels = new String[0];
    String text = "Mr. Jones and <skip/>Mrs. Smith.";
    RawText textBlock = new RawText(text, true, sessionId);
    List<Annotation<RawTextNoSentenceBreakMarker>> noSentenceBreaks = new ArrayList<>();
    System.out.println("we add no sentence break annotations (as if they were added by a filter)");
    noSentenceBreaks.add(new Annotation<>("".length(), "Mr.".length(), new RawTextNoSentenceBreakMarker("me"), labels));
    noSentenceBreaks.add(new Annotation<>("Mr. Jones and <skip/>".length(), "Mr. Jones and <skip/>Mrs.".length(), new RawTextNoSentenceBreakMarker("me"), labels));
    textBlock.addAnnotations(noSentenceBreaks);
    List<Annotation<RawTextSkipMarker>> skips = new ArrayList<>();
    skips.add(new Annotation<>("Mr. Jones and ".length(), "Mr. Jones and <skip/>".length(), new RawTextSkipMarker("me"), labels));
    textBlock.addAnnotations(skips);
    System.out.println("textBlock text: " + textBlock.getText());
    System.out.println("textBlock annotations: " + textBlock.getAnnotations().toString());
    AnnotatedText processedTextBlock = textBlock.getProcessedText();
    assertEquals("Mr. Jones and Mrs. Smith.", processedTextBlock.getText());
    // ensure that the no sentence break text got added at the right place
    // in the processed text
    noSentenceBreaks = processedTextBlock.getAnnotations(RawTextNoSentenceBreakMarker.class);
    System.out.println("Processed annotations: " + noSentenceBreaks);
    assertEquals(2, noSentenceBreaks.size());
    assertEquals("".length(), noSentenceBreaks.get(0).getStart());
    assertEquals("Mr.".length(), noSentenceBreaks.get(0).getEnd());
    assertEquals("Mr. Jones and ".length(), noSentenceBreaks.get(1).getStart());
    assertEquals("Mr. Jones and Mrs.".length(), noSentenceBreaks.get(1).getEnd());
}

Example 3 with RawTextNoSentenceBreakMarker

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

the class RollingTextBlockTest method testNoSentenceAnnotationLocation.

@Test
public void testNoSentenceAnnotationLocation() throws Exception {
    System.setProperty("config.file", "src/test/resources/test.conf");
    ConfigFactory.invalidateCaches();
    final Config config = ConfigFactory.load();
    final String sessionId = "test";
    String[] labels = new String[0];
    // String text = "I see Mr. Jones and <skip/>Mrs. Smith.";
    RollingTextBlock textBlock = new RollingTextBlock(true, null, sessionId);
    textBlock = textBlock.roll("I see ");
    textBlock = textBlock.roll("Mr. Jones ");
    textBlock = textBlock.roll("and <sk");
    AnnotatedText rawText = textBlock.getRawTextBlock();
    System.out.println("rawText text: " + rawText.getText());
    List<Annotation<RawTextNoSentenceBreakMarker>> noSentenceBreaks = new ArrayList<>();
    System.out.println("we add no sentence break annotations (as if they were added by a filter)");
    noSentenceBreaks.add(new Annotation<>("".length(), "Mr.".length(), new RawTextNoSentenceBreakMarker("me"), labels));
    rawText.addAnnotations(noSentenceBreaks);
    System.out.println("textBlock text: " + textBlock.getText());
    System.out.println("textBlock annotations: " + textBlock.getAnnotations().toString());
    textBlock = textBlock.roll("ip/>Mrs.");
    rawText = textBlock.getRawTextBlock();
    System.out.println("rawText text: " + rawText.getText());
    List<Annotation<RawTextSkipMarker>> skips = new ArrayList<>();
    skips.add(new Annotation<>("and ".length(), "and <skip/>".length(), new RawTextSkipMarker("me"), labels));
    rawText.addAnnotations(skips);
    AnnotatedText processedTextBlock = textBlock.getProcessedText();
    assertEquals("I see Mr. Jones and ", processedTextBlock.getText());
    // ensure that the no sentence break text got added at the right place
    // in the processed text
    noSentenceBreaks = processedTextBlock.getAnnotations(RawTextNoSentenceBreakMarker.class);
    System.out.println("Processed annotations: " + noSentenceBreaks);
    assertEquals(1, noSentenceBreaks.size());
    assertEquals("I see ".length(), noSentenceBreaks.get(0).getStart());
    assertEquals("I see Mr.".length(), noSentenceBreaks.get(0).getEnd());
    textBlock = textBlock.roll(" Smith.");
    rawText = textBlock.getRawTextBlock();
    System.out.println("rawText text: " + rawText.getText());
    noSentenceBreaks = new ArrayList<>();
    noSentenceBreaks.add(new Annotation<>("ip/>".length(), "ip/>Mrs.".length(), new RawTextNoSentenceBreakMarker("me"), labels));
    rawText.addAnnotations(noSentenceBreaks);
    System.out.println("textBlock text: " + textBlock.getText());
    System.out.println("textBlock annotations: " + textBlock.getAnnotations().toString());
    textBlock = textBlock.roll("");
    System.out.println("textBlock text: " + textBlock.getText());
    System.out.println("textBlock annotations: " + textBlock.getAnnotations().toString());
    processedTextBlock = textBlock.getProcessedText();
    assertEquals("and Mrs. Smith.", processedTextBlock.getText());
    // ensure that the no sentence break text got added at the right place
    // in the processed text
    noSentenceBreaks = processedTextBlock.getAnnotations(RawTextNoSentenceBreakMarker.class);
    System.out.println("Processed annotations: " + noSentenceBreaks);
    assertEquals(1, noSentenceBreaks.size());
    assertEquals("and ".length(), noSentenceBreaks.get(0).getStart());
    assertEquals("and Mrs.".length(), noSentenceBreaks.get(0).getEnd());
}

Example 4 with RawTextNoSentenceBreakMarker

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

the class RawTextRegexAnnotator method annotate.

@Override
public void annotate(AnnotatedText textBlock, String... labels) throws MatchTooLargeException {
    if (LOG.isTraceEnabled()) {
        LOG.trace("Matching " + regex.replace('\n', '¶').replace('\r', '¶'));
    }
    List<Annotation<RawTextMarker>> rawTextMarkers = new ArrayList<>();
    List<Annotation<TokenAttribute<?>>> tokenAttributes = new ArrayList<>();
    Matcher matcher = pattern.matcher(textBlock.getText());
    while (matcher.find()) {
        int matcherStart = 0;
        int matcherEnd = 0;
        if (groupIndex == 0) {
            matcherStart = matcher.start();
            matcherEnd = matcher.end();
        } else {
            matcherStart = matcher.start(groupIndex);
            matcherEnd = matcher.end(groupIndex);
        }
        CharSequence matchText = textBlock.getText().subSequence(matcher.start(), matcher.end());
        if (LOG.isTraceEnabled()) {
            LOG.trace("Next match: " + matchText.toString().replace('\n', '¶').replace('\r', '¶'));
            if (matcher.start() != matcherStart || matcher.end() != matcherEnd) {
                LOG.trace("But matching group: " + textBlock.getText().subSequence(matcherStart, matcherEnd).toString().replace('\n', '¶').replace('\r', '¶'));
            }
            LOG.trace("matcher.start()=" + matcher.start() + ", matcher.end()=" + matcher.end() + ", matcherStart=" + matcherStart + ", matcherEnd=" + matcherEnd + ", analysisStart=" + textBlock.getAnalysisStart() + ", analysisEnd=" + textBlock.getAnalysisEnd());
        }
        if (blockSize > 0 && matcherEnd - matcherStart > blockSize) {
            String errorString = "Match size (" + (matcherEnd - matcherStart) + ") bigger than block size (" + blockSize + "). " + "Increase blockSize or change filter. " + "Maybe you need to change a greedy quantifier (e.g. .*) to a reluctant quantifier (e.g. .*?)? " + "Regex: " + regex + ". Text: " + matchText;
            throw new MatchTooLargeException(errorString);
        }
        if (matcherStart >= textBlock.getAnalysisStart() && matcherStart < textBlock.getAnalysisEnd()) {
            if (LOG.isTraceEnabled()) {
                LOG.trace("Start in range: analysisStart " + textBlock.getAnalysisStart() + ">= matcherStart [[" + matcherStart + "]] < analysisEnd " + textBlock.getAnalysisEnd());
            }
            for (RawTextMarkType filterType : filterTypes) {
                switch(filterType) {
                    case REPLACE:
                        {
                            String insertionText = RegexUtils.getReplacement(replacement, textBlock.getText(), matcher);
                            if (LOG.isTraceEnabled()) {
                                LOG.trace("Setting replacement to: " + insertionText);
                            }
                            RawTextMarker marker = new RawTextReplaceMarker(this.toString(), insertionText);
                            Annotation<RawTextMarker> annotation = new Annotation<>(matcherStart, matcherEnd, marker, labels);
                            rawTextMarkers.add(annotation);
                            break;
                        }
                    case SENTENCE_BREAK:
                        {
                            RawTextMarker marker = new RawTextSentenceBreakMarker(this.toString());
                            Annotation<RawTextMarker> annotation = new Annotation<>(matcherStart, matcherEnd, marker, labels);
                            rawTextMarkers.add(annotation);
                            break;
                        }
                    case NO_SENTENCE_BREAK:
                        {
                            RawTextMarker marker = new RawTextNoSentenceBreakMarker(this.toString());
                            Annotation<RawTextMarker> annotation = new Annotation<>(matcherStart, matcherEnd, marker, labels);
                            rawTextMarkers.add(annotation);
                            break;
                        }
                    case SKIP:
                        {
                            RawTextMarker marker = new RawTextSkipMarker(this.toString());
                            Annotation<RawTextMarker> annotation = new Annotation<>(matcherStart, matcherEnd, marker, labels);
                            rawTextMarkers.add(annotation);
                            break;
                        }
                    case TAG:
                        {
                            Annotation<TokenAttribute<?>> annotation = new Annotation<TokenAttribute<?>>(matcherStart, matcherEnd, this.attribute, labels);
                            tokenAttributes.add(annotation);
                            break;
                        }
                    default:
                        {
                            RawTextMarker marker = new RawTextMarker(filterType, this.toString());
                            Annotation<RawTextMarker> annotation = new Annotation<>(matcherStart, matcherEnd, marker, labels);
                            rawTextMarkers.add(annotation);
                            break;
                        }
                }
            }
        } else {
            if (LOG.isTraceEnabled()) {
                LOG.trace("Start out of range: analysisStart " + textBlock.getAnalysisStart() + ">= matcherStart [[" + matcherStart + "]] < analysisEnd " + textBlock.getAnalysisEnd());
            }
        }
    }
    if (rawTextMarkers.size() > 0) {
        if (LOG.isDebugEnabled()) {
            LOG.debug("For regex: " + this.regex.replace('\n', '¶').replace('\r', '¶'));
            LOG.debug("Added annotations: " + rawTextMarkers);
        }
    }
    if (rawTextMarkers.size() > 0)
        textBlock.addAnnotations(rawTextMarkers);
    if (tokenAttributes.size() > 0)
        textBlock.addAnnotations(tokenAttributes);
}

Example 5 with RawTextNoSentenceBreakMarker

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

the class SentenceDetectorTest method testDetectSentences2.

@Test
public void testDetectSentences2() throws Exception {
    System.setProperty("config.file", "src/test/resources/test.conf");
    ConfigFactory.invalidateCaches();
    final Config config = ConfigFactory.load();
    final String sessionId = "test";
    DecisionMaker decisionMaker = new DecisionMaker() {

        @Override
        public ScoringStrategy<ClassificationSolution> getDefaultScoringStrategy() {
            return new GeometricMeanScoringStrategy();
        }

        @Override
        public List<Decision> decide(List<FeatureResult<?>> featureResults) {
            List<Decision> decisions = new ArrayList<>();
            Decision decision = new Decision(SentenceDetectorOutcome.IS_BOUNDARY.name(), 1.0);
            decisions.add(decision);
            return decisions;
        }
    };
    String[] labels = new String[0];
    Set<SentenceDetectorFeature<?>> features = new HashSet<>();
    SentenceDetector sentenceDetector = new SentenceDetector(decisionMaker, features, sessionId);
    String text = "Before analysis. Hello Mr. Jones\nHow are you, Mr. Jones? After";
    AnnotatedText annotatedText = new AnnotatedText(text, "Before analysis. ".length(), text.length());
    List<Annotation<RawTextNoSentenceBreakMarker>> noSentenceBreakMarkers = new ArrayList<>();
    noSentenceBreakMarkers.add(new Annotation<>("Before analysis. Hello ".length(), "Before analysis. Hello Mr.".length(), new RawTextNoSentenceBreakMarker("me"), labels));
    noSentenceBreakMarkers.add(new Annotation<>("Before analysis. Hello Mr. Jones\nHow are you, ".length(), "Before analysis. Hello Mr. Jones\nHow are you, Mr.".length(), new RawTextNoSentenceBreakMarker("me"), labels));
    annotatedText.addAnnotations(noSentenceBreakMarkers);
    List<Annotation<SentenceBoundary>> existingBoundaries = new ArrayList<>();
    existingBoundaries.add(new Annotation<>("".length(), "Before analysis.".length(), new SentenceBoundary(), labels));
    annotatedText.addAnnotations(existingBoundaries);
    List<Annotation<RawTextSentenceBreakMarker>> sentenceBreaks = new ArrayList<>();
    sentenceBreaks.add(new Annotation<>("Before analysis. Hello Mr. Jones".length(), "Before analysis. Hello Mr. Jones\n".length(), new RawTextSentenceBreakMarker("me"), labels));
    annotatedText.addAnnotations(sentenceBreaks);
    List<Integer> guessedBoundaries = sentenceDetector.detectSentences(annotatedText);
    assertEquals(2, guessedBoundaries.size());
    assertEquals("Before analysis. Hello Mr. Jones\n".length(), guessedBoundaries.get(0).intValue());
    assertEquals("Before analysis. Hello Mr. Jones\nHow are you, Mr. Jones?".length(), guessedBoundaries.get(1).intValue());
    List<Annotation<SentenceBoundary>> sentenceBoundaries = annotatedText.getAnnotations(SentenceBoundary.class);
    System.out.println(sentenceBoundaries.toString());
    assertEquals(4, sentenceBoundaries.size());
    assertEquals("".length(), sentenceBoundaries.get(0).getStart());
    assertEquals("Before analysis.".length(), sentenceBoundaries.get(0).getEnd());
    assertEquals("Before analysis. ".length(), sentenceBoundaries.get(1).getStart());
    assertEquals("Before analysis. Hello Mr. Jones\n".length(), sentenceBoundaries.get(1).getEnd());
    assertEquals("Before analysis. Hello Mr. Jones\n".length(), sentenceBoundaries.get(2).getStart());
    assertEquals("Before analysis. Hello Mr. Jones\nHow are you, Mr. Jones?".length(), sentenceBoundaries.get(2).getEnd());
    assertEquals("Before analysis. Hello Mr. Jones\nHow are you, Mr. Jones?".length(), sentenceBoundaries.get(3).getStart());
    assertEquals("Before analysis. Hello Mr. Jones\nHow are you, Mr. Jones? After".length(), sentenceBoundaries.get(3).getEnd());
}