boxtec
/
yacy_search_server
mirror of https://github.com/yacy/yacy_search_server
4
1
Fork 0
Code Issues Projects Releases Wiki Activity

integrated language detection classes into condenser environment

Browse Source 
git-svn-id: https://svn.berlios.de/svnroot/repos/yacy/trunk@5180 6c8d7289-2bf4-0310-a012-ef5d649a1542
pull/1/head
orbiter 17 years ago
parent 2b13705839
commit ce2a7ed116
7 changed files with 126 additions and 81 deletions
Show Stats Download Patch File Download Diff File

6
source/de/anomic/http/httpdProxyCacheEntry.java
Unescape View File

@ -58,7 +58,6 @@ public class httpdProxyCacheEntry implements indexDocumentMetadata {
private byte[] cacheArray; // or the cache as byte-array
private final yacyURL url;
private final String name; // the name of the link, read as anchor from an <a>-tag
private final String language;
private final CrawlProfile.entry profile;
private final String initiator;
private httpRequestHeader requestHeader;
@ -166,7 +165,6 @@ public class httpdProxyCacheEntry implements indexDocumentMetadata {
// in case of proxy usage, the initiator hash is null,
// which distinguishes local crawling from proxy indexing
this.initiator = (initiator == null) ? null : ((initiator.length() == 0) ? null : initiator);
this.language = yacyURL.language(url);
// to be defined later:
this.cacheArray = null;
@ -205,7 +203,9 @@ public class httpdProxyCacheEntry implements indexDocumentMetadata {
}
public String language() {
return this.language;
// please avoid this method if a condenser document is available, because the condenser has a built-in language detection
// this here is only a guess using the TLD
return this.url().language();
}
public CrawlProfile.entry profile() {

167
source/de/anomic/language/identification/Identificator.java
Unescape View File

@ -24,8 +24,10 @@
package de.anomic.language.identification;
import java.util.HashMap;
import java.util.Iterator;
import java.util.List;
import java.util.Map;
import java.util.Vector;
/**
@ -33,12 +35,18 @@ import java.util.Vector;
*/
public final class Identificator {
private final LanguageStatisticsHolder languages;
private static final LanguageStatisticsHolder languages = LanguageStatisticsHolder.getInstance();
private HashMap<Character, Integer> letter;
private int letters;
private String language;
public Identificator() {
languages = LanguageStatisticsHolder.getInstance();
letter = new HashMap<Character, Integer>();
letters = 0;
language = null;
}
/**
* This method tries to return the language a text is written in. The method will only
* use the first 100000 characters of the text which should be enough. Using more
@ -46,7 +54,7 @@ public final class Identificator {
* @param text the text that is to be analyzed
* @return the language or "unknown" if the method was not able to find out the language
*/
public String getLanguage(final String text) {
public static String getLanguage(final String text) {
// only test the first 100000 characters of a text
return getLanguage(text, 100000);
}
@ -56,47 +64,62 @@ public final class Identificator {
* use the number characters defined in the parameter limit.
* @param text the text that is to be analyzed
* @param limit the number of characters that are supposed to be considered
* @return the language or "unknown" if the method was not able to find out the language
* @return the language or null if the method was not able to find out the language
*/
public String getLanguage(final String text, final int limit) {
String ret = null;
public static String getLanguage(final String text, final int limit) {
final LanguageStatistics testStat = new LanguageStatistics("text");
final char[] letter = new char[1];
float letters = 0;
int upperLimit = text.length();
if (upperLimit > limit) {
upperLimit = limit;
}
Identificator id = new Identificator();
// count number of characters in text
for (int i = 0; i < upperLimit; i++) {
text.getChars(i, i + 1, letter, 0);
// only count if character is a letter
if ((letter[0]+"").matches("\\p{L}")) {
letter[0] = Character.toLowerCase(letter[0]);
testStat.put(letter[0], testStat.get(letter[0]) + 1);
letters++;
}
for (int i = 0; i < upperLimit; i++) id.inc(text.charAt(i));
return id.getLanguage();
}
public void inc(final char c) {
if (!Character.isLetter(c)) return;
Character cc = Character.toLowerCase(c);
Integer i = letter.get(cc);
if (i == null) {
letter.put(cc, 1);
} else {
letter.put(cc, i.intValue() + 1);
}
letters++;
}
public void add(String word) {
if (word == null) return;
for (int i = 0; i < word.length(); i++) inc(word.charAt(i));
}
public String getLanguage() {
if (language != null) return language; // don't compute that twice
if (letters == 0) return null; // not enough information available
final LanguageStatistics testStat = new LanguageStatistics("text");
// calculate percentage
Iterator<Character> iter = testStat.keySet().iterator();
Character character;
Character maxChar = null;
float value = 0;
float max = 0;
while (iter.hasNext()) {
character = iter.next();
value = testStat.get(character);
if (value > max) {
int count = 0;
int max = 0;
for (Map.Entry<Character, Integer> e: letter.entrySet()) {
character = e.getKey();
count = e.getValue().intValue();
if (count > max) {
maxChar = character;
max = value;
max = count;
}
testStat.put(character, (value / letters) * 100);
testStat.put(character, ((float) 100) * ((float) count) / ((float) letters));
}
// create list with relevant languages
final List<Integer> relevantLanguages = new Vector <Integer>();
for (int i = 0; i < languages.size(); i++) {
@ -107,56 +130,56 @@ public final class Identificator {
}
}
if (relevantLanguages.size() > 0) {
if (relevantLanguages.size() == 0) return null;
// compare characters in text with characters in statistics
final float[] offsetList = new float[relevantLanguages.size()];
final int[] votesList = new int[relevantLanguages.size()];
iter = testStat.keySet().iterator();
float minimum;
float offset = 0;
float valueCharacter;
int bestLanguage = -1;
// compare characters in text with characters in statistics
final float[] offsetList = new float[relevantLanguages.size()];
final int[] votesList = new int[relevantLanguages.size()];
while (iter.hasNext()) {
minimum = 100.1f;
character = iter.next();
valueCharacter = testStat.get(character);
for (int i = 0; i < relevantLanguages.size(); i++) {
value = languages.get(relevantLanguages.get(i)).get(character);
offset = Math.abs(value - valueCharacter);
offsetList[i] = offsetList[i] + offset;
if (offset < minimum) {
minimum = offset;
bestLanguage = i;
}
}
votesList[bestLanguage] = ++votesList[bestLanguage];
}
// Now we can count how many votes each language got and how far it was away from the stats.
// If 2 languages have the same amount of votes, the one with the smaller offset wins.
int maxVotes = 0;
float minOffset = 100.1f;
for (int i = 0; i < votesList.length; i++) {
if ((votesList[i] == maxVotes && offsetList[i] < minOffset) || (votesList[i] > maxVotes)) {
maxVotes = votesList[i];
minOffset = offsetList[i];
Iterator<Character> iter = testStat.keySet().iterator();
float minimum;
float offset = 0;
float valueCharacter;
int bestLanguage = -1;
float value;
while (iter.hasNext()) {
minimum = 100.1f;
character = iter.next();
valueCharacter = testStat.get(character);
for (int i = 0; i < relevantLanguages.size(); i++) {
value = languages.get(relevantLanguages.get(i)).get(character);
offset = Math.abs(value - valueCharacter);
offsetList[i] = offsetList[i] + offset;
if (offset < minimum) {
minimum = offset;
bestLanguage = i;
}
}
// Only return name of language of offset is smaller than 20%. This
// prevents a language beeing reported that has won the voting, but
// is still not the right language.
if (offset < 20) {
ret = languages.get(relevantLanguages.get(bestLanguage)).getName();
votesList[bestLanguage] = ++votesList[bestLanguage];
}
// Now we can count how many votes each language got and how far it was away from the stats.
// If 2 languages have the same amount of votes, the one with the smaller offset wins.
int maxVotes = 0;
float minOffset = 100.1f;
for (int i = 0; i < votesList.length; i++) {
if ((votesList[i] == maxVotes && offsetList[i] < minOffset) || (votesList[i] > maxVotes)) {
maxVotes = votesList[i];
minOffset = offsetList[i];
bestLanguage = i;
}
}
return ret;
// Only return name of language of offset is smaller than 20%. This
// prevents a language beeing reported that has won the voting, but
// is still not the right language.
if (offset < 20) {
language = languages.get(relevantLanguages.get(bestLanguage)).getName();
return language;
}
return null;
}

5
source/de/anomic/language/identification/LanguageStatistics.java
Unescape View File

@ -89,8 +89,9 @@ public class LanguageStatistics {
* @return the percentage
*/
public final float get(final char letter) {
if (stats.containsKey(letter)) {
return stats.get(letter);
Float f = stats.get(letter);
if (f != null) {
return f.floatValue();
}
return 0;
}

6
source/de/anomic/language/identification/LanguageStatisticsHolder.java
Unescape View File

@ -28,6 +28,8 @@ import java.io.File;
import java.io.FilenameFilter;
import java.util.Vector;
import de.anomic.server.logging.serverLog;
/**
* This class loads and provides several language statistics to the system.
* It has been implemented as a Singleton since it has to access several
@ -64,6 +66,10 @@ public class LanguageStatisticsHolder extends Vector<LanguageStatistics> {
private void addAllLanguagesInDirectory(final String directory) {
final File folder = new File(directory);
if (!folder.exists()) {
serverLog.logSevere("LanguageStatistics", "the language statistics folder " + directory + " cannot be found");
return;
}
final FilenameFilter filter = new LanguageFilenameFilter();
final File[] allLanguageFiles = folder.listFiles(filter);

9
source/de/anomic/plasma/plasmaCondenser.java
Unescape View File

@ -52,6 +52,7 @@ import de.anomic.index.indexRWIEntry;
import de.anomic.index.indexWord;
import de.anomic.kelondro.kelondroBitfield;
import de.anomic.kelondro.kelondroMSetTools;
import de.anomic.language.identification.Identificator;
import de.anomic.yacy.yacyURL;
public final class plasmaCondenser {
@ -110,6 +111,7 @@ public final class plasmaCondenser {
public int RESULT_NUMB_SENTENCES = -1;
public int RESULT_DIFF_SENTENCES = -1;
public kelondroBitfield RESULT_FLAGS = new kelondroBitfield(4);
Identificator languageIdentificator;
public plasmaCondenser(final plasmaParserDocument document, final boolean indexText, final boolean indexMedia) throws UnsupportedEncodingException {
// if addMedia == true, then all the media links are also parsed and added to the words
@ -119,6 +121,8 @@ public final class plasmaCondenser {
this.words = new TreeMap<String, indexWord>();
this.RESULT_FLAGS = new kelondroBitfield(4);
this.languageIdentificator = new Identificator();
//System.out.println("DEBUG: condensing " + document.getMainLongTitle() + ", indexText=" + Boolean.toString(indexText) + ", indexMedia=" + Boolean.toString(indexMedia));
insertTextToWords(document.dc_source().toNormalform(false, true), 0, indexRWIEntry.flag_app_dc_identifier, RESULT_FLAGS);
@ -229,6 +233,7 @@ public final class plasmaCondenser {
int pip = 0;
while (wordenum.hasMoreElements()) {
word = (new String(wordenum.nextElement())).toLowerCase();
languageIdentificator.add(word);
wprop = words.get(word);
if (wprop == null) wprop = new indexWord(0, pip, phrase);
if (wprop.flags == null) wprop.flags = flagstemplate.clone();
@ -264,6 +269,10 @@ public final class plasmaCondenser {
// returns the words as word/indexWord relation map
return words;
}
public String language() {
return this.languageIdentificator.getLanguage();
}
public String intString(final int number, final int length) {
String s = Integer.toString(number);

11
source/de/anomic/plasma/plasmaWordIndex.java
Unescape View File

@ -818,6 +818,13 @@ public final class plasmaWordIndex implements indexRI {
final String dc_title = document.dc_title();
final yacyURL referrerURL = entry.referrerURL();
final Date docDate = entry.getModificationDate();
String language = condenser.language();
if (language == null) {
System.out.println("*** DEBUG LANGUAGE: identification of " + entry.url() + " FAILED, taking TLD");
language = entry.url().language();
} else {
System.out.println("*** DEBUG LANGUAGE: identification of " + entry.url() + " SUCCESS: " + language);
}
// create a new loaded URL db entry
final long ldate = System.currentTimeMillis();
@ -836,7 +843,7 @@ public final class plasmaWordIndex implements indexRI {
condenser.RESULT_NUMB_WORDS, // word count
httpdProxyCacheEntry.docType(document.dc_format()), // doctype
condenser.RESULT_FLAGS, // flags
yacyURL.language(entry.url()), // language
language, // language
document.inboundLinks(), // inbound links
document.outboundLinks(), // outbound links
document.getAudiolinks().size(), // laudio
@ -856,7 +863,7 @@ public final class plasmaWordIndex implements indexRI {
docDate, // document mod date
document, // document content
condenser, // document condenser
yacyURL.language(entry.url()), // document language
language, // document language
httpdProxyCacheEntry.docType(document.dc_format()), // document type
document.inboundLinks(), // inbound links
document.outboundLinks() // outbound links

3
source/de/anomic/yacy/yacyURL.java
Unescape View File

@ -847,9 +847,8 @@ public class yacyURL implements Serializable {
}
// language calculation
public static String language(final yacyURL url) {
public String language() {
String language = "uk";
final String host = url.getHost();
final int pos = host.lastIndexOf(".");
if ((pos > 0) && (host.length() - pos == 3)) language = host.substring(pos + 1).toLowerCase();
return language;

Write Preview
Loading…
Cancel
Save