|
|
/**
|
|
|
* DateDetection
|
|
|
* Copyright 2014 by Michael Peter Christen
|
|
|
* First released 12.12.2014 at http://yacy.net
|
|
|
*
|
|
|
* This library is free software; you can redistribute it and/or
|
|
|
* modify it under the terms of the GNU Lesser General Public
|
|
|
* License as published by the Free Software Foundation; either
|
|
|
* version 2.1 of the License, or (at your option) any later version.
|
|
|
*
|
|
|
* This library is distributed in the hope that it will be useful,
|
|
|
* but WITHOUT ANY WARRANTY; without even the implied warranty of
|
|
|
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
|
|
|
* Lesser General Public License for more details.
|
|
|
*
|
|
|
* You should have received a copy of the GNU Lesser General Public License
|
|
|
* along with this program in the file lgpl21.txt
|
|
|
* If not, see <http://www.gnu.org/licenses/>.
|
|
|
*/
|
|
|
|
|
|
package net.yacy.document;
|
|
|
|
|
|
import java.text.ParseException;
|
|
|
import java.text.SimpleDateFormat;
|
|
|
import java.util.Date;
|
|
|
import java.util.HashMap;
|
|
|
import java.util.HashSet;
|
|
|
import java.util.LinkedHashMap;
|
|
|
import java.util.LinkedHashSet;
|
|
|
import java.util.Locale;
|
|
|
import java.util.Map;
|
|
|
import java.util.Set;
|
|
|
import java.util.TimeZone;
|
|
|
import java.util.TreeMap;
|
|
|
import java.util.regex.Matcher;
|
|
|
import java.util.regex.Pattern;
|
|
|
|
|
|
import net.yacy.cora.date.AbstractFormatter;
|
|
|
import net.yacy.cora.date.GenericFormatter;
|
|
|
|
|
|
/**
|
|
|
* The purpose of this class exceeds the demands on simple date parsing using a SimpleDateFormat
|
|
|
* because it tries to
|
|
|
* - discover where in a text a date is given
|
|
|
* - recognize human ways of date description and get it into a context, like 'next friday'
|
|
|
* - enrich partially given dates, i.e. when the year is omitted
|
|
|
* - understand different languages
|
|
|
*/
|
|
|
public class DateDetection {
|
|
|
|
|
|
private static final TimeZone TZ_GMT = TimeZone.getTimeZone("GMT");
|
|
|
private static final String CONPATT = "yyyy/MM/dd";
|
|
|
private static final SimpleDateFormat CONFORM = new SimpleDateFormat(CONPATT, Locale.US);
|
|
|
private static final LinkedHashMap<Language, String[]> Weekdays = new LinkedHashMap<>();
|
|
|
private static final LinkedHashMap<Language, String[]> Months = new LinkedHashMap<>();
|
|
|
private static final int[] MaxDaysInMonth = new int[]{31,29,31,30,31,30,31,31,30,31,30,31};
|
|
|
|
|
|
// to assign names for days and months, we must know what language is used to express that time
|
|
|
public static enum Language {
|
|
|
GERMAN, ENGLISH, FRENCH, SPANISH, ITALIAN;
|
|
|
}
|
|
|
|
|
|
static {
|
|
|
CONFORM.setTimeZone(TZ_GMT);
|
|
|
// all names must be lowercase because compared strings are made to lowercase as well
|
|
|
Weekdays.put(Language.GERMAN, new String[]{"montag", "dienstag", "mittwoch", "donnerstag", "freitag", "samstag" /*oder: "sonnabend"*/, "sonntag"});
|
|
|
Weekdays.put(Language.ENGLISH, new String[]{"monday", "tuesday", "wednesday", "thursday", "friday", "saturday", "sunday"});
|
|
|
Weekdays.put(Language.FRENCH, new String[]{"lundi", "mardi", "mercredi", "jeudi", "vendredi", "samedi", "dimanche"});
|
|
|
Weekdays.put(Language.SPANISH, new String[]{"lunes", "martes", "miércoles", "jueves", "viernes", "sábado", "domingo"});
|
|
|
Weekdays.put(Language.ITALIAN, new String[]{"lunedì", "martedì", "mercoledì", "giovedì", "venerdì", "sabato", "domenica"});
|
|
|
Months.put(Language.GERMAN, new String[]{"januar", "februar", "märz", "april", "mai", "juni", "juli", "august", "september", "oktober", "november", "dezember"});
|
|
|
Months.put(Language.ENGLISH, new String[]{"january", "february", "march", "april", "may", "june", "july", "august", "september", "october", "november", "december"});
|
|
|
Months.put(Language.FRENCH, new String[]{"janvier", "février", "mars", "avril", "mai", "juin", "juillet", "août", "septembre", "octobre", "novembre", "décembre"});
|
|
|
Months.put(Language.SPANISH, new String[]{"enero", "febrero", "marzo", "abril", "mayo", "junio", "julio", "agosto", "septiembre", "octubre", "noviembre", "diciembre"});
|
|
|
Months.put(Language.ITALIAN, new String[]{"gennaio", "febbraio", "marzo", "aprile", "maggio", "giugno", "luglio", "agosto", "settembre", "ottobre", "novembre", "dicembre"});
|
|
|
|
|
|
}
|
|
|
|
|
|
// RFC 822 day and month specification as a norm for date formats. This is needed to reconstruct the actual date later
|
|
|
public static enum Weekday {
|
|
|
Mon(Weekdays, 0),
|
|
|
Tue(Weekdays, 1),
|
|
|
Wed(Weekdays, 2),
|
|
|
Thu(Weekdays, 3),
|
|
|
Fri(Weekdays, 4),
|
|
|
Sat(Weekdays, 5),
|
|
|
Sun(Weekdays, 6);
|
|
|
|
|
|
private final Map<String, Language> inLanguages; // a map from the word to the language
|
|
|
public final int offset; // the day offset in the week, monday = 0
|
|
|
private Weekday(final LinkedHashMap<Language, String[]> weekdayMap, final int offset) {
|
|
|
this.inLanguages = new HashMap<>();
|
|
|
this.offset = offset;
|
|
|
for (Map.Entry<Language, String[]> entry: weekdayMap.entrySet()) {
|
|
|
this.inLanguages.put(entry.getValue()[offset], entry.getKey());
|
|
|
}
|
|
|
}
|
|
|
}
|
|
|
|
|
|
public static enum Month {
|
|
|
Jan( 1), Feb( 2), Mar( 3), Apr( 4), May( 5), Jun( 6),
|
|
|
Jul( 7), Aug( 8), Sep( 9), Oct(10), Nov(11), Dec(12);
|
|
|
//private final Map<String, Language> inLanguages;
|
|
|
private final int count;
|
|
|
private Month(final int count) {
|
|
|
this.count = count;
|
|
|
}
|
|
|
}
|
|
|
|
|
|
public static enum EntityType {
|
|
|
YEAR(new LinkedHashMap<Language, String[]>()),
|
|
|
MONTH(Months),
|
|
|
DAY(new LinkedHashMap<Language, String[]>()),
|
|
|
WEEKDAYS(Weekdays);
|
|
|
LinkedHashMap<Language, String[]> languageTerms;
|
|
|
EntityType(LinkedHashMap<Language, String[]> languageTerms) {
|
|
|
this.languageTerms = languageTerms;
|
|
|
}
|
|
|
}
|
|
|
|
|
|
private final static Date TODAY = new Date();
|
|
|
private final static int CURRENT_YEAR = Integer.parseInt(CONFORM.format(TODAY).substring(0, 4)); // we need that to parse dates without given years, see the ShortStyle class
|
|
|
private final static int CURRENT_MONTH = Integer.parseInt(CONFORM.format(TODAY).substring(5, 7)); // wee need that to generate recurring dates, see RecurringStyle class
|
|
|
|
|
|
private final static String BODNCG = "(?:\\b|^)"; // begin of date non-capturing group
|
|
|
private final static String EODNCG = "(?:[).:;! ]|$)"; // end of date non-capturing group
|
|
|
private final static String SEPARATORNCG = "(?:/|-| - |\\.\\s|,\\s|\\.|,|\\s)"; // separator non-capturing group
|
|
|
private final static String DAYCAPTURE = "(\\d{1,2})";
|
|
|
private final static String YEARCAPTURE = "(\\d{2}|\\d{4})";
|
|
|
private final static String MONTHCAPTURE = "(\\p{L}{3,}|\\d{1,2})";
|
|
|
|
|
|
public static class HolidayMap extends TreeMap<String, Date[]>{
|
|
|
private static final long serialVersionUID = 1L;
|
|
|
public HolidayMap() {
|
|
|
super(String.CASE_INSENSITIVE_ORDER);
|
|
|
}
|
|
|
}
|
|
|
|
|
|
public static HolidayMap Holidays = new HolidayMap();
|
|
|
public static Map<Pattern, Date[]> HolidayPattern = new HashMap<>();
|
|
|
|
|
|
static {
|
|
|
try {
|
|
|
// German
|
|
|
Holidays.put("Neujahr", sameDayEveryYear(1, 1));
|
|
|
Holidays.put("Heilige Drei Könige", sameDayEveryYear(1, 6));
|
|
|
Holidays.put("Valentinstag", sameDayEveryYear(2, 14));
|
|
|
Holidays.put("Weiberfastnacht", new Date[]{CONFORM.parse("2014/02/27"), CONFORM.parse("2015/02/12"), CONFORM.parse("2016/02/04")});
|
|
|
Holidays.put("Weiberfasching", Holidays.get("Weiberfastnacht"));
|
|
|
Holidays.put("Rosenmontag", new Date[]{CONFORM.parse("2014/03/03"), CONFORM.parse("2015/03/16"), CONFORM.parse("2016/02/08")});
|
|
|
Holidays.put("Faschingsdienstag", new Date[]{CONFORM.parse("2014/03/04"), CONFORM.parse("2015/03/17"), CONFORM.parse("2016/02/09")});
|
|
|
Holidays.put("Fastnacht", new Date[]{CONFORM.parse("2014/03/04"), CONFORM.parse("2015/03/17"), CONFORM.parse("2016/02/09")});
|
|
|
Holidays.put("Aschermittwoch", new Date[]{CONFORM.parse("2014/03/05"), CONFORM.parse("2015/03/18"), CONFORM.parse("2016/02/10")});
|
|
|
Holidays.put("Palmsonntag", new Date[]{CONFORM.parse("2014/04/13"), CONFORM.parse("2015/03/29"), CONFORM.parse("2016/04/20")});
|
|
|
Holidays.put("Gründonnerstag", new Date[]{CONFORM.parse("2014/04/17"), CONFORM.parse("2015/04/02"), CONFORM.parse("2016/04/24")});
|
|
|
Holidays.put("Karfreitag", new Date[]{CONFORM.parse("2014/04/18"), CONFORM.parse("2015/04/03"), CONFORM.parse("2016/04/25")});
|
|
|
Holidays.put("Karsamstag", new Date[]{CONFORM.parse("2014/04/19"), CONFORM.parse("2015/04/04"), CONFORM.parse("2016/04/26")});
|
|
|
Holidays.put("Ostersonntag", new Date[]{CONFORM.parse("2014/04/20"), CONFORM.parse("2015/04/05"), CONFORM.parse("2016/04/27")});
|
|
|
Holidays.put("Ostermontag", new Date[]{CONFORM.parse("2014/04/21"), CONFORM.parse("2015/04/06"), CONFORM.parse("2016/04/28")});
|
|
|
Holidays.put("Ostern", new Date[]{CONFORM.parse("2014/04/20"), CONFORM.parse("2015/04/05"), CONFORM.parse("2016/04/27"),
|
|
|
CONFORM.parse("2014/04/21"), CONFORM.parse("2015/04/06"), CONFORM.parse("2016/04/28")});
|
|
|
Holidays.put("Walpurgisnacht", sameDayEveryYear(4, 30));
|
|
|
Holidays.put("Tag der Arbeit", sameDayEveryYear(5, 1));
|
|
|
Holidays.put("Muttertag", new Date[]{CONFORM.parse("2014/05/11"), CONFORM.parse("2015/05/10"), CONFORM.parse("2016/05/08")});
|
|
|
Holidays.put("Christi Himmelfahrt", new Date[]{CONFORM.parse("2014/05/29"), CONFORM.parse("2015/05/14"), CONFORM.parse("2016/05/05")});
|
|
|
Holidays.put("Pfingstsonntag", new Date[]{CONFORM.parse("2014/06/08"), CONFORM.parse("2015/05/24"), CONFORM.parse("2016/05/15")});
|
|
|
Holidays.put("Pfingstmontag", new Date[]{CONFORM.parse("2014/06/09"), CONFORM.parse("2015/05/25"), CONFORM.parse("2016/05/16")});
|
|
|
Holidays.put("Fronleichnam", new Date[]{CONFORM.parse("2014/06/19"), CONFORM.parse("2015/06/04"), CONFORM.parse("2016/05/25")});
|
|
|
Holidays.put("Mariä Himmelfahrt", sameDayEveryYear(8, 15));
|
|
|
Holidays.put("Tag der Deutschen Einheit", sameDayEveryYear(10, 3));
|
|
|
Holidays.put("Reformationstag", sameDayEveryYear(10, 31));
|
|
|
Holidays.put("Allerheiligen", sameDayEveryYear(11, 1));
|
|
|
Holidays.put("Allerseelen", sameDayEveryYear(11, 2));
|
|
|
Holidays.put("Martinstag", sameDayEveryYear(11, 11));
|
|
|
Holidays.put("St. Martin", Holidays.get("Martinstag"));
|
|
|
Holidays.put("Volkstrauertag", new Date[]{CONFORM.parse("2014/11/16"), CONFORM.parse("2015/11/15"), CONFORM.parse("2016/11/13")});
|
|
|
Holidays.put("Buß- und Bettag", new Date[]{CONFORM.parse("2014/11/19"), CONFORM.parse("2015/11/18"), CONFORM.parse("2016/11/16")});
|
|
|
Holidays.put("Totensonntag", new Date[]{CONFORM.parse("2014/11/23"), CONFORM.parse("2015/11/22"), CONFORM.parse("2016/11/20")});
|
|
|
Holidays.put("Nikolaus", sameDayEveryYear(12, 6));
|
|
|
Holidays.put("Heiligabend", sameDayEveryYear(12, 24));
|
|
|
Holidays.put("1. Weihnachtsfeiertag", sameDayEveryYear(12, 25));
|
|
|
Holidays.put("2. Weihnachtsfeiertag", sameDayEveryYear(12, 26));
|
|
|
Holidays.put("1. Advent", new Date[]{CONFORM.parse("2014/11/30"), CONFORM.parse("2015/11/29"), CONFORM.parse("2016/11/27")});
|
|
|
Holidays.put("2. Advent", new Date[]{CONFORM.parse("2014/12/07"), CONFORM.parse("2015/12/06"), CONFORM.parse("2016/12/04")});
|
|
|
Holidays.put("3. Advent", new Date[]{CONFORM.parse("2014/12/14"), CONFORM.parse("2015/12/13"), CONFORM.parse("2016/12/11")});
|
|
|
Holidays.put("4. Advent", new Date[]{CONFORM.parse("2014/12/21"), CONFORM.parse("2015/12/20"), CONFORM.parse("2016/12/18")});
|
|
|
Holidays.put("Silvester", sameDayEveryYear(12, 26));
|
|
|
|
|
|
// English
|
|
|
Holidays.put("Eastern", Holidays.get("Ostern"));
|
|
|
Holidays.put("New Year's Day", Holidays.get("Neujahr"));
|
|
|
Holidays.put("Epiphany", Holidays.get("Heilige Drei Könige"));
|
|
|
Holidays.put("Valentine's Day", Holidays.get("Valentinstag"));
|
|
|
Holidays.put("Orthodox Christmas", sameDayEveryYear(1, 7));
|
|
|
Holidays.put("St. Patrick's Day", sameDayEveryYear(3, 17));
|
|
|
Holidays.put("April Fools' Day", sameDayEveryYear(4, 1));
|
|
|
Holidays.put("Independence Day", sameDayEveryYear(7, 4));
|
|
|
Holidays.put("Halloween", Holidays.get("Reformationstag"));
|
|
|
Holidays.put("Immaculate Conception of the Virgin Mary", sameDayEveryYear(12, 8));
|
|
|
Holidays.put("Christmas Eve", Holidays.get("Heiligabend"));
|
|
|
Holidays.put("Christmas Day", Holidays.get("1. Weihnachtsfeiertag"));
|
|
|
Holidays.put("Boxing Day", Holidays.get("2. Weihnachtsfeiertag"));
|
|
|
Holidays.put("New Year's Eve", Holidays.get("Silvester"));
|
|
|
} catch (ParseException e) {}
|
|
|
|
|
|
|
|
|
for (Map.Entry<String, Date[]> holiday: Holidays.entrySet()) {
|
|
|
HolidayPattern.put(Pattern.compile(BODNCG + holiday.getKey() + EODNCG), holiday.getValue());
|
|
|
}
|
|
|
}
|
|
|
|
|
|
private static Date[] sameDayEveryYear(int month, int day) {
|
|
|
Date[] r = new Date[4];
|
|
|
String d = "/" + (month < 10 ? "0" + month : month) + "/ "+ (day < 10 ? "0" + day : day);
|
|
|
for (int y = 0; y < 4; y++) try {r[y] = CONFORM.parse((CURRENT_YEAR + y - 1) + d);} catch (ParseException e) {}
|
|
|
return r;
|
|
|
}
|
|
|
|
|
|
/**
|
|
|
* The language recognition subclass understands date description parts in different languages.
|
|
|
* It can also be used to identify the language of a text, if that text uses words from a date vocabulary.
|
|
|
*/
|
|
|
public static class LanguageRecognition {
|
|
|
|
|
|
private final Pattern weekdayMatch, monthMatch;
|
|
|
private final Set<Language> usedInLanguages;
|
|
|
private final Map<String, Integer> weekdayIndex, monthIndex, monthIndexAbbrev;
|
|
|
|
|
|
public LanguageRecognition(Language[] languages) {
|
|
|
this.usedInLanguages = new HashSet<Language>();
|
|
|
// prepare a month index for the languages that this notion supports
|
|
|
this.weekdayIndex = new HashMap<>();
|
|
|
this.monthIndex = new HashMap<>();
|
|
|
this.monthIndexAbbrev = new HashMap<>();
|
|
|
StringBuilder weekdayMatchString = new StringBuilder();
|
|
|
StringBuilder monthMatchString = new StringBuilder();
|
|
|
for (Language language: languages) {
|
|
|
this.usedInLanguages.add(language);
|
|
|
|
|
|
String[] weekdays = Weekdays.get(language);
|
|
|
if (weekdays != null) {
|
|
|
assert weekdays.length == 7;
|
|
|
for (int i = 0; i < 7; i++) {
|
|
|
this.weekdayIndex.put(weekdays[i], i);
|
|
|
weekdayMatchString.append("|(?:").append(BODNCG).append(weekdays[i]).append(SEPARATORNCG).append(EODNCG).append(')');
|
|
|
}
|
|
|
}
|
|
|
|
|
|
String[] months = Months.get(language);
|
|
|
if (months != null) {
|
|
|
assert months.length == 12;
|
|
|
for (int i = 0; i < 12; i++) {
|
|
|
monthIndex.put(months[i], i + 1);
|
|
|
monthMatchString.append("|(?:").append(BODNCG).append(months[i]).append(SEPARATORNCG).append(EODNCG).append(')');
|
|
|
String abbrev = months[i].substring(0, 3);
|
|
|
if (monthIndexAbbrev.containsKey(abbrev) && monthIndexAbbrev.get(abbrev).intValue() != i + 1)
|
|
|
monthIndexAbbrev.put(abbrev, -1); // ambiguous months get a -1
|
|
|
else
|
|
|
monthIndexAbbrev.put(abbrev, i + 1);
|
|
|
}
|
|
|
}
|
|
|
}
|
|
|
this.weekdayMatch = Pattern.compile(weekdayMatchString.length() > 0 ? weekdayMatchString.substring(1) : "");
|
|
|
this.monthMatch = Pattern.compile(monthMatchString.length() > 0 ? monthMatchString.substring(1) : "");
|
|
|
}
|
|
|
|
|
|
/**
|
|
|
* this is an expensive check that looks if any of the words from the date expressions (month and weekday expressions)
|
|
|
* appear in the text. This should only be used to verify a parse result if the result was ambiguous
|
|
|
* @param text
|
|
|
* @return true if one of the month and weekday expressions appear in the text
|
|
|
*/
|
|
|
public boolean usesLanguageOfNotion(String text) {
|
|
|
return this.weekdayMatch.matcher(text).matches() || this.monthMatch.matcher(text).matches();
|
|
|
}
|
|
|
|
|
|
/**
|
|
|
* parse a part of a date
|
|
|
* @param entity
|
|
|
* @param object
|
|
|
* @return a scalar value associated with this date part
|
|
|
*/
|
|
|
public int parseEntity(EntityType entity, String object) {
|
|
|
if (entity == EntityType.YEAR) {
|
|
|
try {
|
|
|
int i = Integer.parseInt(object);
|
|
|
if (i < 100) i += 2000; // yes that makes it possible to parse the years 0-99 and it will be incorrect in the year 2100 when that is abbreviated with 00
|
|
|
if (i > CURRENT_YEAR + 10) return -1; // there are very rarely dates in the future that far
|
|
|
return i;
|
|
|
} catch (NumberFormatException e) {
|
|
|
return -1;
|
|
|
}
|
|
|
}
|
|
|
if (entity == EntityType.MONTH) {
|
|
|
try {
|
|
|
int i = Integer.parseInt(object);
|
|
|
if (i >= 1 && i <= 12) return i;
|
|
|
return -1; // no reason to try in a different way, its just a wrong number
|
|
|
} catch (NumberFormatException e) {
|
|
|
// this may be the name of a month
|
|
|
if (object.length() == 3) {
|
|
|
// try RFC 822 names
|
|
|
object = object.substring(0, 1).toUpperCase() + object.substring(1).toLowerCase();
|
|
|
try {
|
|
|
Month m = Month.valueOf(object);
|
|
|
return m.count;
|
|
|
} catch (IllegalArgumentException | NoClassDefFoundError ee) {} // just ignore this, that was just a try to shorten things..
|
|
|
}
|
|
|
// try the collection of names for each language
|
|
|
object = object.toLowerCase(); // the stored month names are all lowercase
|
|
|
Integer i = this.monthIndex.get(object);
|
|
|
if (i != null) return i.intValue();
|
|
|
// try an abbreviation
|
|
|
if (object.length() == 3) {
|
|
|
i = this.monthIndexAbbrev.get(object.substring(0, 3));
|
|
|
if (i != null) return i.intValue(); // may also be -1!
|
|
|
}
|
|
|
return -1;
|
|
|
}
|
|
|
}
|
|
|
if (entity == EntityType.DAY) {
|
|
|
try {
|
|
|
int i = Integer.parseInt(object);
|
|
|
if (i < 1 || i > 31) return -1;
|
|
|
return i;
|
|
|
} catch (NumberFormatException e) {
|
|
|
return -1;
|
|
|
}
|
|
|
}
|
|
|
return -1;
|
|
|
}
|
|
|
|
|
|
}
|
|
|
|
|
|
private final static LanguageRecognition ENGLISH_LANGUAGE = new LanguageRecognition(new Language[]{Language.ENGLISH});
|
|
|
private final static LanguageRecognition GERMAN_LANGUAGE = new LanguageRecognition(new Language[]{Language.GERMAN});
|
|
|
private final static LanguageRecognition FRENCH_LANGUAGE = new LanguageRecognition(new Language[]{Language.FRENCH});
|
|
|
private final static LanguageRecognition ENGLISH_GERMAN_LANGUAGE = new LanguageRecognition(new Language[]{Language.GERMAN, Language.ENGLISH});
|
|
|
private final static LanguageRecognition ENGLISH_GERMAN_FRENCH_SPANISH_ITALIAN_LANGUAGE = new LanguageRecognition(new Language[]{Language.GERMAN, Language.ENGLISH, Language.FRENCH, Language.SPANISH, Language.ITALIAN});
|
|
|
|
|
|
public static interface StyleParser {
|
|
|
/**
|
|
|
* get all dates in the text
|
|
|
* @param text
|
|
|
* @return a set of dates, ordered by occurrence.
|
|
|
*/
|
|
|
public LinkedHashSet<Date> parse(String text);
|
|
|
}
|
|
|
|
|
|
/**
|
|
|
* Regular expressions for various types of date writings.
|
|
|
* Uses terminology and data taken from:
|
|
|
* http://en.wikipedia.org/wiki/Date_format_by_country
|
|
|
*/
|
|
|
public static enum EndianStyle implements StyleParser {
|
|
|
YMD(EntityType.YEAR, EntityType.MONTH, EntityType.DAY, // Big-endian (year, month, day), e.g. 1996-04-22
|
|
|
ENGLISH_GERMAN_LANGUAGE, // GERMAN: 'official standard date format', ENGLISH: used in UK
|
|
|
BODNCG + YEARCAPTURE + SEPARATORNCG + MONTHCAPTURE + SEPARATORNCG + DAYCAPTURE + EODNCG
|
|
|
),
|
|
|
DMY(EntityType.DAY, EntityType.MONTH, EntityType.YEAR, // Little-endian (day, month, year), e.g. 22.04.96 or 22/04/96 or 22 April 1996
|
|
|
ENGLISH_GERMAN_FRENCH_SPANISH_ITALIAN_LANGUAGE, // GERMAN: traditional, ENGLISH: used in UK
|
|
|
BODNCG + DAYCAPTURE + SEPARATORNCG + MONTHCAPTURE + SEPARATORNCG + YEARCAPTURE + EODNCG
|
|
|
),
|
|
|
MDY(EntityType.MONTH, EntityType.DAY, EntityType.YEAR, // Middle-endian (month, day, year), e.g. 04/22/96 or April 22, 1996
|
|
|
ENGLISH_LANGUAGE, // ENGLISH: used in USA
|
|
|
BODNCG + MONTHCAPTURE + SEPARATORNCG + DAYCAPTURE + SEPARATORNCG + YEARCAPTURE + EODNCG
|
|
|
);
|
|
|
|
|
|
private final Pattern pattern;
|
|
|
private final EntityType firstEntity, secondEntity, thirdEntity;
|
|
|
public final LanguageRecognition languageParser;
|
|
|
EndianStyle(EntityType firstEntity, EntityType secondEntity, EntityType thirdEntity, LanguageRecognition languageParser, String patternString) {
|
|
|
this.firstEntity = firstEntity;
|
|
|
this.secondEntity = secondEntity;
|
|
|
this.thirdEntity = thirdEntity;
|
|
|
this.pattern = Pattern.compile(patternString);
|
|
|
this.languageParser = languageParser;
|
|
|
}
|
|
|
|
|
|
/**
|
|
|
* get all dates in the text
|
|
|
* @param text
|
|
|
* @return a set of dates, ordered by occurrence.
|
|
|
*/
|
|
|
@Override
|
|
|
public LinkedHashSet<Date> parse(final String text) {
|
|
|
LinkedHashSet<Date> dates = new LinkedHashSet<>();
|
|
|
Matcher matcher = this.pattern.matcher(text);
|
|
|
while (matcher.find()) {
|
|
|
if (!(matcher.groupCount() == 3)) continue;
|
|
|
String entity1 = matcher.group(1); if (entity1 == null) continue;
|
|
|
String entity2 = matcher.group(2); if (entity2 == null) continue;
|
|
|
String entity3 = matcher.group(3); if (entity3 == null) continue;
|
|
|
//System.out.println("FRAGMENTS: entity1=" + entity1 + ", entity2=" + entity2 + ", entity3=" + entity3); // DEBUG
|
|
|
int i1 = languageParser.parseEntity(this.firstEntity, entity1);
|
|
|
if (i1 < 0) continue;
|
|
|
int i2 = languageParser.parseEntity(this.secondEntity, entity2);
|
|
|
if (i2 < 0) continue;
|
|
|
int i3 = languageParser.parseEntity(this.thirdEntity, entity3);
|
|
|
if (i3 < 0) continue;
|
|
|
int day = this.firstEntity == EntityType.DAY ? i1 : this.secondEntity == EntityType.DAY ? i2 : i3;
|
|
|
int month = this.firstEntity == EntityType.MONTH ? i1 : this.secondEntity == EntityType.MONTH ? i2 : i3;
|
|
|
if (day > MaxDaysInMonth[month - 1]) continue; // validity check of the day number
|
|
|
int year = this.firstEntity == EntityType.YEAR ? i1 : this.secondEntity == EntityType.YEAR ? i2 : i3;
|
|
|
synchronized (CONFORM) {try {
|
|
|
dates.add(CONFORM.parse(year + "/" + (month < 10 ? "0" : "") + month + "/" + (day < 10 ? "0" : "") + day));
|
|
|
} catch (ParseException e) {
|
|
|
continue;
|
|
|
}}
|
|
|
if (dates.size() > 100) {dates.clear(); break;} // that does not make sense
|
|
|
}
|
|
|
return dates;
|
|
|
}
|
|
|
|
|
|
}
|
|
|
|
|
|
public static enum ShortStyle implements StyleParser {
|
|
|
MD_ENGLISH(EntityType.MONTH, EntityType.DAY, // Big-endian (month, day), e.g. "from october 1st to september 13th"
|
|
|
ENGLISH_LANGUAGE,
|
|
|
BODNCG + "on " + MONTHCAPTURE + SEPARATORNCG + DAYCAPTURE + EODNCG
|
|
|
),
|
|
|
DM_GERMAN(EntityType.DAY, EntityType.MONTH, // Little-endian (day, month), e.g. "am 1. April"
|
|
|
GERMAN_LANGUAGE,
|
|
|
BODNCG + "am " + DAYCAPTURE + SEPARATORNCG + MONTHCAPTURE + EODNCG
|
|
|
),
|
|
|
DM_FRENCH(EntityType.DAY, EntityType.MONTH, // Little-endian (day, month), e.g. "le 29 Septembre,"
|
|
|
FRENCH_LANGUAGE,
|
|
|
BODNCG + "le " + DAYCAPTURE + " " + MONTHCAPTURE + EODNCG
|
|
|
),
|
|
|
DM_ITALIAN(EntityType.DAY, EntityType.MONTH, // Little-endian (day, month), e.g. "il 29 settembre,"
|
|
|
FRENCH_LANGUAGE,
|
|
|
BODNCG + "il " + DAYCAPTURE + " " + MONTHCAPTURE + EODNCG
|
|
|
),
|
|
|
DM_SPANISH(EntityType.DAY, EntityType.MONTH, // Little-endian (day, month), e.g. "el 29 de septiembre,"
|
|
|
FRENCH_LANGUAGE,
|
|
|
BODNCG + "el " + DAYCAPTURE + " de " + MONTHCAPTURE + EODNCG
|
|
|
);
|
|
|
public final Pattern pattern;
|
|
|
private final EntityType firstEntity, secondEntity;
|
|
|
public final LanguageRecognition languageParser;
|
|
|
ShortStyle(EntityType firstEntity, EntityType secondEntity, LanguageRecognition languageParser, String patternString) {
|
|
|
this.firstEntity = firstEntity;
|
|
|
this.secondEntity = secondEntity;
|
|
|
this.pattern = Pattern.compile(patternString);
|
|
|
this.languageParser = languageParser;
|
|
|
}
|
|
|
|
|
|
/**
|
|
|
* get all dates in the text
|
|
|
* @param text
|
|
|
* @return a set of dates, ordered by occurrence.
|
|
|
*/
|
|
|
@Override
|
|
|
public LinkedHashSet<Date> parse(final String text) {
|
|
|
LinkedHashSet<Date> dates = new LinkedHashSet<>();
|
|
|
Matcher matcher = this.pattern.matcher(text);
|
|
|
//ConcurrentLog.info("DateDetection", "applying matcher: " + matcher.toString());
|
|
|
while (matcher.find()) {
|
|
|
if (!(matcher.groupCount() == 2)) continue;
|
|
|
String entity1 = matcher.group(1); if (entity1 == null) continue;
|
|
|
String entity2 = matcher.group(2); if (entity2 == null) continue;
|
|
|
//System.out.println("FRAGMENTS: entity1=" + entity1 + ", entity2=" + entity2 + ", entity3=" + entity3); // DEBUG
|
|
|
int i1 = languageParser.parseEntity(this.firstEntity, entity1);
|
|
|
if (i1 < 0) continue;
|
|
|
int i2 = languageParser.parseEntity(this.secondEntity, entity2);
|
|
|
if (i2 < 0) continue;
|
|
|
int day = this.firstEntity == EntityType.DAY ? i1 : i2;
|
|
|
int month = this.firstEntity == EntityType.MONTH ? i1 : i2;
|
|
|
if (day > MaxDaysInMonth[month - 1]) continue; // validity check of the day number
|
|
|
int thisyear = CURRENT_YEAR;
|
|
|
int nextyear = CURRENT_YEAR + 1;
|
|
|
synchronized (CONFORM) {try {
|
|
|
String datestub = "/" + (month < 10 ? "0" : "") + month + "/" + (day < 10 ? "0" : "") + day;
|
|
|
Date atThisYear = CONFORM.parse(thisyear + datestub);
|
|
|
Date atNextYear = CONFORM.parse(nextyear + datestub);
|
|
|
dates.add(atThisYear);
|
|
|
dates.add(atNextYear);
|
|
|
//dates.add(atThisYear.after(TODAY) ? atThisYear : atNextYear); // we consider these kind of dates as given for the future
|
|
|
} catch (ParseException e) {
|
|
|
continue;
|
|
|
}}
|
|
|
if (dates.size() > 100) {dates.clear(); break;} // that does not make sense
|
|
|
}
|
|
|
return dates;
|
|
|
}
|
|
|
|
|
|
}
|
|
|
|
|
|
private static final HashMap<String, Long> specialDayOffset = new HashMap<>();
|
|
|
static {
|
|
|
specialDayOffset.put("today", 0L); specialDayOffset.put("heute", 0L);
|
|
|
specialDayOffset.put("tomorrow", AbstractFormatter.dayMillis); specialDayOffset.put("morgen", AbstractFormatter.dayMillis);
|
|
|
specialDayOffset.put("dayaftertomorrow", 2 * AbstractFormatter.dayMillis); specialDayOffset.put("uebermorgen", 2 * AbstractFormatter.dayMillis);
|
|
|
specialDayOffset.put("yesterday", -AbstractFormatter.dayMillis); specialDayOffset.put("gestern", -AbstractFormatter.dayMillis);
|
|
|
}
|
|
|
|
|
|
/**
|
|
|
* get all dates in the text
|
|
|
* @param text
|
|
|
* @return a set of dates, ordered by time. first date in the ordered set is the oldest time.
|
|
|
*/
|
|
|
public static LinkedHashSet<Date> parse(String text) {
|
|
|
Long offset;
|
|
|
if ((offset = specialDayOffset.get(text)) != null) {
|
|
|
LinkedHashSet<Date> dates = new LinkedHashSet<>(); dates.add(new Date((System.currentTimeMillis() / AbstractFormatter.dayMillis) * AbstractFormatter.dayMillis + offset.longValue())); return dates;
|
|
|
}
|
|
|
LinkedHashSet<Date> dates = parseRawDate(text);
|
|
|
for (Map.Entry<Pattern, Date[]> entry: HolidayPattern.entrySet()) {
|
|
|
if (entry.getKey().matcher(text).matches()) {
|
|
|
for (Date d: entry.getValue()) dates.add(d);
|
|
|
}
|
|
|
}
|
|
|
return dates;
|
|
|
}
|
|
|
|
|
|
public static Date parseLine(String text) {
|
|
|
Date d = null;
|
|
|
try {d = CONFORM.parse(text);} catch (ParseException e) {}
|
|
|
//if (d == null) try {d = GenericFormatter.FORMAT_SHORT_DAY.parse(text);} catch (ParseException e) {} // did not work well and fired for wrong formats; do not use
|
|
|
if (d == null) try {d = GenericFormatter.FORMAT_RFC1123_SHORT.parse(text);} catch (ParseException e) {}
|
|
|
if (d == null) try {d = GenericFormatter.FORMAT_ANSIC.parse(text);} catch (ParseException e) {}
|
|
|
|
|
|
if (d == null) {
|
|
|
Set<Date> dd = parse(text);
|
|
|
if (dd.size() >= 1) d = dd.iterator().next();
|
|
|
}
|
|
|
return d;
|
|
|
}
|
|
|
|
|
|
private static LinkedHashSet<Date> parseRawDate(String text) {
|
|
|
// get parse alternatives for different date styles; we consider that one document uses only one style
|
|
|
LinkedHashSet<Date> DMYDates = EndianStyle.DMY.parse(text);
|
|
|
ShortStyle[] shortStyleCheck = new ShortStyle[]{ShortStyle.DM_GERMAN, ShortStyle.DM_FRENCH, ShortStyle.DM_ITALIAN, ShortStyle.DM_SPANISH};
|
|
|
LinkedHashSet<Date> DMDates = new LinkedHashSet<>();
|
|
|
for (ShortStyle shortStyle: shortStyleCheck) {
|
|
|
DMDates.addAll(shortStyle.parse(text));
|
|
|
if (DMDates.size() > 0) break;
|
|
|
}
|
|
|
DMYDates.addAll(DMDates);
|
|
|
|
|
|
LinkedHashSet<Date> MDYDates = DMYDates.size() == 0 ? EndianStyle.MDY.parse(text) : new LinkedHashSet<Date>(0);
|
|
|
LinkedHashSet<Date> MDDates = DMYDates.size() == 0 ? ShortStyle.MD_ENGLISH.parse(text) : new LinkedHashSet<Date>(0);
|
|
|
MDYDates.addAll(MDDates);
|
|
|
|
|
|
LinkedHashSet<Date> YMDDates = DMYDates.size() == 0 && MDYDates.size() == 0 ? EndianStyle.YMD.parse(text) : new LinkedHashSet<Date>(0);
|
|
|
|
|
|
// if either one of them contains any and the other contain no date, chose that one (we don't want to mix them)
|
|
|
if (YMDDates.size() > 0 && DMYDates.size() == 0 && MDYDates.size() == 0) return YMDDates;
|
|
|
if (YMDDates.size() == 0 && DMYDates.size() > 0 && MDYDates.size() == 0) return DMYDates;
|
|
|
if (YMDDates.size() == 0 && DMYDates.size() == 0 && MDYDates.size() > 0) return MDYDates;
|
|
|
|
|
|
// if we have several sets, check if we can detect the language from month or weekday expressions
|
|
|
// we sort out such sets, which do not contain any of these languages
|
|
|
boolean usesLanguageOfYMD = YMDDates.size() > 0 ? false : EndianStyle.YMD.languageParser.usesLanguageOfNotion(text);
|
|
|
boolean usesLanguageOfDMY = DMYDates.size() > 0 ? false : EndianStyle.DMY.languageParser.usesLanguageOfNotion(text);
|
|
|
boolean usesLanguageOfMDY = MDYDates.size() > 0 ? false : EndianStyle.MDY.languageParser.usesLanguageOfNotion(text);
|
|
|
|
|
|
// now check again
|
|
|
if (usesLanguageOfYMD && !usesLanguageOfDMY && !usesLanguageOfMDY) return YMDDates;
|
|
|
if (!usesLanguageOfYMD && usesLanguageOfDMY && !usesLanguageOfMDY) return DMYDates;
|
|
|
if (!usesLanguageOfYMD && !usesLanguageOfDMY && usesLanguageOfMDY) return MDYDates;
|
|
|
|
|
|
// if this fails, we return only the DMY format since that has the most chances to be right (it is mostly used)
|
|
|
// we choose DMYDates even if it is empty to avoid false positives.
|
|
|
return DMYDates;
|
|
|
}
|
|
|
|
|
|
public static void main(String[] args) {
|
|
|
String fill = ""; for (int i = 0; i < 1000; i++) fill += 'x';
|
|
|
String[] test = new String[]{
|
|
|
"\n laden die Stadtwerke \n X am Rosenmontag und am \n Faschingsdienstag zur Disko auf die \n",
|
|
|
"kein Datum im Text",
|
|
|
" Fastnacht am 4. März noch",
|
|
|
" Fastnacht am 4. April noch",
|
|
|
"heute 12. Dezember 2014. ",
|
|
|
"heute 12. Dezember 2014",
|
|
|
"12. Dezember 2014. ",
|
|
|
"heute 12. Dezember 2014 ",
|
|
|
"heute 12. Dezember 2014. ",
|
|
|
"Donnerstag, 18. Dezember 2014 xyz",
|
|
|
"Donnerstag, 18 Dezember 2014 xyz",
|
|
|
"Donnerstag, 18.Dezember 2014 xyz",
|
|
|
"Montag, 8. Dezember 2014 xyz",
|
|
|
"Montag, 8.Dezember 2014 xyz",
|
|
|
"Donnerstag, 18.12.2014 xyz",
|
|
|
"Montag, 8.12.2014 xyz",
|
|
|
"Donnerstag, 18.12.14 xyz",
|
|
|
"Montag, 8.12.14 xyz",
|
|
|
"Mitglied seit: 13. Januar 2007 xyz",
|
|
|
"Im Dezember 2014 xyz",
|
|
|
"11.12.2014",
|
|
|
"11. September 2001",
|
|
|
"12.12.2014 08:43",
|
|
|
"immer am 1. Dezember abends",
|
|
|
"immer am 31. Dezember abends",
|
|
|
"immer am 31. dezember abends",
|
|
|
"on october 20 every year",
|
|
|
" on october 20 every year",
|
|
|
"on September 29,",
|
|
|
"am Karfreitag um 15:00 Uhr"
|
|
|
};
|
|
|
long t = System.currentTimeMillis();
|
|
|
for (String s: test) {
|
|
|
String parsed = parse(fill + " " + s + " " + fill).toString();
|
|
|
System.out.println("SOURCE: " + s);
|
|
|
System.out.println("DATE : " + parsed);
|
|
|
System.out.println();
|
|
|
}
|
|
|
System.out.println("Runtime: " + (System.currentTimeMillis() - t) + " milliseconds.");
|
|
|
}
|
|
|
|
|
|
}
|