/**
* Summarize text.
*
* @param sentences Tokenized sentences to summarize.
* @param summarySentences Maximum number of sentences to return in the summary.
* @return Summary of the input text.
*/
public <T extends Comparable> List<Integer> summarize(
List<List<T>> sentences, int summarySentences) {
// Get word counts ignoring stop words.
Map<String, WordCountAndSentences> wordCounts =
WordCountUtils.countWordsInSentences(sentences, new BuckleyAndSaltonStopWords());
// Sort the counts into descending
// order by count.
List<WordCountAndSentences> wcsData = new SortedArrayList<WordCountAndSentences>();
Iterator<String> iterator = wordCounts.keySet().iterator();
while (iterator.hasNext()) {
wcsData.add(wordCounts.get(iterator.next()));
}
// Holds summary sentence indices.
Set<Integer> summarySentencesSet = new TreeSet<Integer>();
// Use up to 100 most commonly used words.
int maxWords = Math.min(100, wcsData.size());
// For each commonly word word,
// find the first sentence in which
// that word appears, and add it to the
// summary sentences collection.
for (int i = 0; (i < wcsData.size()) && (summarySentencesSet.size() < summarySentences); i++) {
WordCountAndSentences wcs = wcsData.get(i);
if (CharUtils.isNumber(wcs.word.toString())) continue;
if (CharUtils.hasDigit(wcs.word.toString())) continue;
Integer[] sentenceNumbers =
(Integer[]) wcs.sentences.toArray(new Integer[wcs.sentences.size()]);
summarySentencesSet.add(sentenceNumbers[0]);
}
// Return indices of selected
// summary sentences.
return new ArrayList<Integer>(summarySentencesSet);
}
