package edu.stanford.nlp.process;
import java.io.File;
import java.net.URL;
import java.util.*;
import java.util.regex.Pattern;
import edu.stanford.nlp.io.EncodingPrintWriter;
import edu.stanford.nlp.ling.BasicDocument;
import edu.stanford.nlp.ling.Document;
import edu.stanford.nlp.ling.HasWord;
import edu.stanford.nlp.ling.Sentence;
import edu.stanford.nlp.ling.Word;
import edu.stanford.nlp.ling.CoreAnnotations;
import edu.stanford.nlp.util.CoreMap;
import edu.stanford.nlp.util.Generics;
/**
* Transforms a Document of Words into a Document of Sentences by grouping the
* Words. The word stream is assumed to already be adequately tokenized,
* and this class just divides the list into sentences, perhaps discarding
* some separator tokens based on the setting of the following three sets:
* <ul>
* <li>sentenceBoundaryTokens are tokens that are left in a sentence, but are
* to be regarded as ending a sentence. A canonical example is a period.
* If two of these follow each other, the second will be a sentence
* consisting of only the sentenceBoundaryToken.
* <li>sentenceBoundaryFollowers are tokens that are left in a sentence, and
* which can follow a sentenceBoundaryToken while still belonging to
* the previous sentence. They cannot begin a sentence (except at the
* beginning of a document). A canonical example is a close parenthesis
* ')'.
* <li>sentenceBoundaryToDiscard are tokens which separate sentences and
* which should be thrown away. In web documents, a typical example would
* be a '{@code <p>}' tag. If two of these follow each other, they are
* coalesced: no empty Sentence is output. The end-of-file is not
* represented in this Set, but the code behaves as if it were a member.
* <li>sentenceRegionBeginPattern A regular expression for marking the start
* of a sentence region. Not included in the sentence.
* <li>sentenceRegionEndPattern A regular expression for marking the end
* of a sentence region. Not included in the sentence.
* </ul>
*
* @author Joseph Smarr (jsmarr@stanford.edu)
* @author Christopher Manning
* @author Teg Grenager (grenager@stanford.edu)
* @author Sarah Spikes (sdspikes@cs.stanford.edu) (Templatization)
*
* @param <IN> The type of the tokens in the sentences
* @param <L> The type of the labels
* @param <F> The type of the features
*/
public class WordToSentenceProcessor<IN, L, F> extends AbstractListProcessor<IN, List<IN>, L, F> {
private static final boolean DEBUG = false;
/**
* Set of tokens (Strings) that qualify as sentence-final tokens.
*/
private Set<String> sentenceBoundaryTokens;
/**
* Set of tokens (Strings) that qualify as tokens that can follow
* what normally counts as an end of sentence token, and which are
* attributed to the preceding sentence. For example ")" coming after
* a period.
*/
private Set<String> sentenceBoundaryFollowers;
/**
* Set of tokens (Strings) that are sentence boundaries to be discarded.
*/
private Set<String> sentenceBoundaryToDiscard;
private Pattern sentenceRegionBeginPattern;
private Pattern sentenceRegionEndPattern;
/**
* Returns a List of Sentences where each element is built from a run
* of Words in the input Document. Specifically, reads through each word in
* the input document and breaks off a sentence after finding a valid
* sentence boundary token or end of file.
* Note that for this to work, the words in the
* input document must have been tokenized with a tokenizer that makes
* sentence boundary tokens their own tokens (e.g., {@link PTBTokenizer}).
*
* @param words A list of already tokenized words (must implement HasWord or be a String)
* @return A list of Sentence
* @see #WordToSentenceProcessor(Set, Set, Set)
* @see edu.stanford.nlp.ling.Sentence
*/
public List<List<IN>> process(List<IN> words) {
List<List<IN>> sentences = Generics.newArrayList();
List<IN> currentSentence = null;
List<IN> lastSentence = null;
boolean insideRegion = false;
for (IN o: words) {
String word;
if (o instanceof HasWord) {
HasWord h = (HasWord) o;
word = h.word();
} else if (o instanceof String) {
word = (String) o;
} else if (o instanceof CoreMap) {
word = ((CoreMap)o).get(CoreAnnotations.WordAnnotation.class);
} else {
throw new RuntimeException("Expected token to be either Word or String.");
}
if (DEBUG) {
EncodingPrintWriter.err.println("Word is " + word, "UTF-8");
}
if (currentSentence == null) {
currentSentence = new ArrayList<IN>();
}
if (sentenceRegionBeginPattern != null && ! insideRegion) {
if (sentenceRegionBeginPattern.matcher(word).matches()) {
insideRegion = true;
}
if (DEBUG) {
System.err.println(" outside region");
}
continue;
}
if (sentenceBoundaryFollowers.contains(word) && lastSentence != null && currentSentence.isEmpty()) {
lastSentence.add(o);
if (DEBUG) {
System.err.println(" added to last");
}
} else {
boolean newSent = false;
if (sentenceBoundaryToDiscard.contains(word)) {
newSent = true;
} else if (sentenceRegionEndPattern != null && sentenceRegionEndPattern.matcher(word).matches()) {
insideRegion = false;
newSent = true;
} else if (sentenceBoundaryTokens.contains(word)) {
currentSentence.add(o);
if (DEBUG) {
System.err.println(" is sentence boundary; added to current");
}
newSent = true;
} else {
currentSentence.add(o);
if (DEBUG) {
System.err.println(" added to current");
}
}
if (newSent && currentSentence.size() > 0) {
if (DEBUG) {
System.err.println(" beginning new sentence");
}
sentences.add(currentSentence);
// adds this sentence now that it's complete
lastSentence = currentSentence;
currentSentence = null; // clears the current sentence
}
}
}
// add any words at the end, even if there isn't a sentence
// terminator at the end of file
if (currentSentence != null && currentSentence.size() > 0) {
sentences.add(currentSentence); // adds last sentence
}
return sentences;
}
/**
* Create a <code>WordToSentenceProcessor</code> using a sensible default
* list of tokens to split on. The default set is: {".","?","!"}.
*/
public WordToSentenceProcessor() {
this(new HashSet<String>(Arrays.asList(".", "?", "!")));
}
/**
* Flexibly set the set of acceptable sentence boundary tokens, but with
* a default set of allowed boundary following tokens (based on English
* and Penn Treebank encoding).
* The allowed set of boundary followers is:
* {")","]","\"","\'", "''", "-RRB-", "-RSB-", "-RCB-"}.
*
* @param boundaryTokens The set of boundary tokens
*/
public WordToSentenceProcessor(Set<String> boundaryTokens) {
this(boundaryTokens, Generics.newHashSet(Arrays.asList(")", "]", "\"", "\'", "''", "-RRB-", "-RSB-", "-RCB-")));
}
/**
* Flexibly set the set of acceptable sentence boundary tokens and
* also the set of tokens commonly following sentence boundaries, and
* the set of discarded separator tokens.
* The default set of discarded separator tokens is: {"\n"}.
*/
public WordToSentenceProcessor(Set<String> boundaryTokens, Set<String> boundaryFollowers) {
this(boundaryTokens, boundaryFollowers, Collections.singleton("\n"));
}
/**
* Flexibly set the set of acceptable sentence boundary tokens,
* the set of tokens commonly following sentence boundaries, and also
* the set of tokens that are sentences boundaries that should be
* discarded.
*/
public WordToSentenceProcessor(Set<String> boundaryTokens, Set<String> boundaryFollowers,
Set<String> boundaryToDiscard) {
this(boundaryTokens, boundaryFollowers, boundaryToDiscard, null, null);
}
public WordToSentenceProcessor(Pattern regionBeginPattern, Pattern regionEndPattern) {
this(Collections.<String>emptySet(), Collections.<String>emptySet(),
Collections.<String>emptySet(), regionBeginPattern, regionEndPattern);
}
/**
* Flexibly set the set of acceptable sentence boundary tokens,
* the set of tokens commonly following sentence boundaries, and also
* the set of tokens that are sentences boundaries that should be
* discarded.
* This is private because it is a dangerous constructor. It's not clear what the semantics
* should be if there are both boundary token sets, and patterns to match.
*/
private WordToSentenceProcessor(Set<String> boundaryTokens, Set<String> boundaryFollowers, Set<String> boundaryToDiscard, Pattern regionBeginPattern, Pattern regionEndPattern) {
sentenceBoundaryTokens = boundaryTokens;
sentenceBoundaryFollowers = boundaryFollowers;
sentenceBoundaryToDiscard = boundaryToDiscard;
sentenceRegionBeginPattern = regionBeginPattern;
sentenceRegionEndPattern = regionEndPattern;
if (DEBUG) {
EncodingPrintWriter.err.println("WordToSentenceProcessor: boundaryTokens=" + boundaryTokens, "UTF-8");
EncodingPrintWriter.err.println(" boundaryFollowers=" + boundaryFollowers, "UTF-8");
EncodingPrintWriter.err.println(" boundaryToDiscard=" + boundaryToDiscard, "UTF-8");
}
}
/* -- for testing only
private void printSet(Set s) {
for (Iterator i = s.iterator(); i.hasNext();) {
System.out.print(i.next() + " ");
}
System.out.println();
}
-- */
/**
* This will print out as sentences some text. It can be used to
* test sentence division. <br>
* Usage: java edu.stanford.nlp.process.WordToSentenceProcessor fileOrUrl+
*
* @param args Command line argument: files or URLs
*/
public static void main(String[] args) {
if (args.length == 0) {
System.out.println("usage: java edu.stanford.nlp.process.WordToSentenceProcessor fileOrUrl");
System.exit(0);
}
try {
for (String filename : args) {
Document<?, Word, Word> d; // always initialized below
if (filename.startsWith("http://")) {
Document dpre = new BasicDocument().init(new URL(filename));
DocumentProcessor notags = new StripTagsProcessor();
d = notags.processDocument(dpre);
} else {
d = new BasicDocument().init(new File(filename));
}
WordToSentenceProcessor proc = new WordToSentenceProcessor();
List<Sentence> sentd = proc.processDocument(d);
for (Sentence sent : sentd) {
System.out.println(sent);
}
}
} catch (Exception e) {
e.printStackTrace();
}
}
}
|