public Set<K2> secondKeySet() {
Set<K2> keys = Generics.newHashSet();
for (K1 k1 : map.keySet()) {
keys.addAll(get(k1).firstKeySet());
}
return keys;
}
@Override
public Set<Requirement> requirementsSatisfied() {
Set<Requirement> satisfied = Generics.newHashSet();
for (Annotator annotator : annotators) {
satisfied.addAll(annotator.requirementsSatisfied());
}
return satisfied;
}
public Set<K3> thirdKeySet() {
Set<K3> keys = Generics.newHashSet();
for (K1 k1 : map.keySet()) {
TwoDimensionalMap<K2, K3, V> m = map.get(k1);
for (K2 k2 : m.firstKeySet()) {
keys.addAll(m.get(k2).keySet());
}
}
return keys;
}
public TokensRegexNERAnnotator(String name, Properties properties) {
String prefix = (name != null && !name.isEmpty()) ? name + '.' : "";
String backgroundSymbol =
properties.getProperty(prefix + "backgroundSymbol", DEFAULT_BACKGROUND_SYMBOL);
String[] backgroundSymbols = backgroundSymbol.split("\\s*,\\s*");
String mappingFiles =
properties.getProperty(prefix + "mapping", DefaultPaths.DEFAULT_REGEXNER_RULES);
String[] mappings = mappingFiles.split("\\s*[,;]\\s*");
String validPosRegex = properties.getProperty(prefix + "validpospattern");
this.posMatchType =
PosMatchType.valueOf(
properties.getProperty(prefix + "posmatchtype", DEFAULT_POS_MATCH_TYPE.name()));
String noDefaultOverwriteLabelsProp =
properties.getProperty(prefix + "noDefaultOverwriteLabels");
this.noDefaultOverwriteLabels =
(noDefaultOverwriteLabelsProp != null)
? Collections.unmodifiableSet(
CollectionUtils.asSet(noDefaultOverwriteLabelsProp.split("\\s*,\\s*")))
: Collections.unmodifiableSet(new HashSet<>());
this.ignoreCase = PropertiesUtils.getBool(properties, prefix + "ignorecase", false);
this.verbose = PropertiesUtils.getBool(properties, prefix + "verbose", false);
if (validPosRegex != null && !validPosRegex.isEmpty()) {
validPosPattern = Pattern.compile(validPosRegex);
} else {
validPosPattern = null;
}
entries =
Collections.unmodifiableList(
readEntries(name, noDefaultOverwriteLabels, ignoreCase, verbose, mappings));
IdentityHashMap<SequencePattern<CoreMap>, Entry> patternToEntry = new IdentityHashMap<>();
multiPatternMatcher = createPatternMatcher(patternToEntry);
this.patternToEntry = Collections.unmodifiableMap(patternToEntry);
Set<String> myLabels = Generics.newHashSet();
// Can always override background or none.
Collections.addAll(myLabels, backgroundSymbols);
myLabels.add(null);
// Always overwrite labels
for (Entry entry : entries) myLabels.add(entry.type);
this.myLabels = Collections.unmodifiableSet(myLabels);
}
/**
* Reads a list of Entries from a mapping file and update the given entries. Line numbers start
* from 1.
*
* @return the updated list of Entries
*/
private static List<Entry> readEntries(
String annotatorName,
List<Entry> entries,
TrieMap<String, Entry> seenRegexes,
String mappingFilename,
BufferedReader mapping,
Set<String> noDefaultOverwriteLabels,
boolean ignoreCase,
boolean verbose)
throws IOException {
int origEntriesSize = entries.size();
int isTokensRegex = 0;
int lineCount = 0;
for (String line; (line = mapping.readLine()) != null; ) {
lineCount++;
String[] split = line.split("\t");
if (split.length < 2 || split.length > 5) {
throw new IllegalArgumentException(
"Provided mapping file is in wrong format. This line is bad: " + line);
}
String regex = split[0].trim();
String tokensRegex = null;
String[] regexes = null;
if (regex.startsWith("( ") && regex.endsWith(" )")) {
// Tokens regex (remove start and end parenthesis)
tokensRegex = regex.substring(1, regex.length() - 1).trim();
} else {
regexes = regex.split("\\s+");
}
String[] key = (regexes != null) ? regexes : new String[] {tokensRegex};
if (ignoreCase) {
String[] norm = new String[key.length];
for (int i = 0; i < key.length; i++) {
norm[i] = key[i].toLowerCase();
}
key = norm;
}
String type = split[1].trim();
Set<String> overwritableTypes = Generics.newHashSet();
double priority = 0.0;
if (split.length >= 3) {
overwritableTypes.addAll(Arrays.asList(split[2].trim().split("\\s*,\\s*")));
}
if (split.length >= 4) {
try {
priority = Double.parseDouble(split[3].trim());
} catch (NumberFormatException e) {
throw new IllegalArgumentException(
"ERROR: Invalid priority in line "
+ lineCount
+ " in regexner file "
+ mappingFilename
+ ": \""
+ line
+ "\"!",
e);
}
}
int annotateGroup = 0;
// Get annotate group from input....
if (split.length >= 5) {
// Which group to take (allow for context)
String context = split[4].trim();
try {
annotateGroup = Integer.parseInt(context);
} catch (NumberFormatException e) {
throw new IllegalArgumentException(
"ERROR: Invalid group in line "
+ lineCount
+ " in regexner file "
+ mappingFilename
+ ": \""
+ line
+ "\"!",
e);
}
}
// Print some warning about the type
int commaPos = type.indexOf(',');
if (commaPos > 0) {
// Strip the "," and just take first type
String newType = type.substring(0, commaPos).trim();
logger.warn(
"TokensRegexNERAnnotator "
+ annotatorName
+ ": Entry has multiple types: "
+ line
+ ". Taking type to be "
+ newType);
type = newType;
}
Entry entry =
new Entry(tokensRegex, regexes, type, overwritableTypes, priority, annotateGroup);
if (seenRegexes.containsKey(key)) {
Entry oldEntry = seenRegexes.get(key);
if (priority > oldEntry.priority) {
logger.warn(
"TokensRegexNERAnnotator "
+ annotatorName
+ ": Replace duplicate entry (higher priority): old="
+ oldEntry
+ ", new="
+ entry);
} else {
if (!oldEntry.type.equals(type)) {
if (verbose) {
logger.warn(
"TokensRegexNERAnnotator "
+ annotatorName
+ ": Ignoring duplicate entry: "
+ split[0]
+ ", old type = "
+ oldEntry.type
+ ", new type = "
+ type);
}
// } else {
// if (verbose) {
// logger.warn("TokensRegexNERAnnotator " + annotatorName +
// ": Duplicate entry [ignored]: " + split[0] + ", old type = " +
// oldEntry.type + ", new type = " + type);
// }
}
continue;
}
}
// Print some warning if label belongs to noDefaultOverwriteLabels but there is no
// overwritable types
if (entry.overwritableTypes.isEmpty() && noDefaultOverwriteLabels.contains(entry.type)) {
logger.warn(
"TokensRegexNERAnnotator "
+ annotatorName
+ ": Entry doesn't have overwriteable types "
+ entry
+ ", but entry type is in noDefaultOverwriteLabels");
}
entries.add(entry);
seenRegexes.put(key, entry);
if (entry.tokensRegex != null) isTokensRegex++;
}
logger.log(
"TokensRegexNERAnnotator "
+ annotatorName
+ ": Read "
+ (entries.size() - origEntriesSize)
+ " unique entries out of "
+ lineCount
+ " from "
+ mappingFilename
+ ", "
+ isTokensRegex
+ " TokensRegex patterns.");
return entries;
}