2018-10-02 10:37:54 +02:00
|
|
|
package eu.dnetlib.pace.common;
|
|
|
|
|
|
|
|
import com.google.common.base.Joiner;
|
2018-12-12 16:30:03 +01:00
|
|
|
import com.google.common.base.Splitter;
|
2018-10-02 10:37:54 +02:00
|
|
|
import com.google.common.collect.Iterables;
|
|
|
|
import com.google.common.collect.Lists;
|
|
|
|
import com.google.common.collect.Sets;
|
|
|
|
import eu.dnetlib.pace.clustering.NGramUtils;
|
|
|
|
import eu.dnetlib.pace.model.Field;
|
|
|
|
import eu.dnetlib.pace.model.FieldList;
|
|
|
|
import eu.dnetlib.pace.model.FieldListImpl;
|
2018-11-12 12:46:06 +01:00
|
|
|
import org.apache.commons.collections.CollectionUtils;
|
|
|
|
import org.apache.commons.io.IOUtils;
|
|
|
|
import org.apache.commons.lang.StringUtils;
|
|
|
|
|
2019-10-08 14:53:52 +02:00
|
|
|
import java.io.IOException;
|
|
|
|
import java.io.StringWriter;
|
2018-11-12 12:46:06 +01:00
|
|
|
import java.text.Normalizer;
|
|
|
|
import java.util.*;
|
2020-01-24 10:01:11 +01:00
|
|
|
import java.util.regex.Matcher;
|
2019-11-28 16:54:44 +01:00
|
|
|
import java.util.regex.Pattern;
|
2019-07-19 17:10:29 +02:00
|
|
|
import java.util.stream.Collectors;
|
2018-10-02 10:37:54 +02:00
|
|
|
|
|
|
|
/**
|
2019-11-07 12:47:12 +01:00
|
|
|
* Set of common functions for the framework
|
2018-10-02 10:37:54 +02:00
|
|
|
*
|
|
|
|
* @author claudio
|
|
|
|
*
|
|
|
|
*/
|
|
|
|
public abstract class AbstractPaceFunctions {
|
|
|
|
|
2019-11-07 12:47:12 +01:00
|
|
|
//city map to be used when translating the city names into codes
|
2019-07-19 17:10:29 +02:00
|
|
|
private static Map<String,String> cityMap = AbstractPaceFunctions.loadMapFromClasspath("/eu/dnetlib/pace/config/city_map.csv");
|
|
|
|
|
2019-11-07 12:47:12 +01:00
|
|
|
//list of stopwords in different languages
|
2019-03-21 14:27:27 +01:00
|
|
|
protected static Set<String> stopwords_en = loadFromClasspath("/eu/dnetlib/pace/config/stopwords_en.txt");
|
|
|
|
protected static Set<String> stopwords_de = loadFromClasspath("/eu/dnetlib/pace/config/stopwords_de.txt");
|
|
|
|
protected static Set<String> stopwords_es = loadFromClasspath("/eu/dnetlib/pace/config/stopwords_es.txt");
|
|
|
|
protected static Set<String> stopwords_fr = loadFromClasspath("/eu/dnetlib/pace/config/stopwords_fr.txt");
|
|
|
|
protected static Set<String> stopwords_it = loadFromClasspath("/eu/dnetlib/pace/config/stopwords_it.txt");
|
|
|
|
protected static Set<String> stopwords_pt = loadFromClasspath("/eu/dnetlib/pace/config/stopwords_pt.txt");
|
2018-10-02 10:37:54 +02:00
|
|
|
|
2019-11-07 12:47:12 +01:00
|
|
|
//blacklist of ngrams: to avoid generic keys
|
2018-10-02 10:37:54 +02:00
|
|
|
protected static Set<String> ngramBlacklist = loadFromClasspath("/eu/dnetlib/pace/config/ngram_blacklist.txt");
|
|
|
|
|
|
|
|
private static final String alpha = "abcdefghijklmnopqrstuvwxyzABCDEFGHIJKLMNOPQRSTUVWXYZ0123456789 ";
|
2019-11-07 12:47:12 +01:00
|
|
|
private static final String aliases_from = "⁰¹²³⁴⁵⁶⁷⁸⁹⁺⁻⁼⁽⁾ⁿ₀₁₂₃₄₅₆₇₈₉₊₋₌₍₎àáâäæãåāèéêëēėęəîïíīįìôöòóœøōõûüùúūßśšłžźżçćčñń";
|
|
|
|
private static final String aliases_to = "0123456789+-=()n0123456789+-=()aaaaaaaaeeeeeeeeiiiiiioooooooouuuuussslzzzcccnn";
|
2018-10-02 10:37:54 +02:00
|
|
|
|
2019-11-07 12:47:12 +01:00
|
|
|
//doi prefix for normalization
|
2019-07-08 09:44:02 +02:00
|
|
|
public final String DOI_PREFIX = "(https?:\\/\\/dx\\.doi\\.org\\/)|(doi:)";
|
|
|
|
|
2019-11-28 16:54:44 +01:00
|
|
|
private Pattern numberPattern = Pattern.compile("-?\\d+(\\.\\d+)?");
|
|
|
|
|
2020-01-24 10:01:11 +01:00
|
|
|
private Pattern hexUnicodePattern = Pattern.compile("\\\\u(\\p{XDigit}{4})");
|
|
|
|
|
2018-10-02 10:37:54 +02:00
|
|
|
protected final static FieldList EMPTY_FIELD = new FieldListImpl();
|
|
|
|
|
|
|
|
protected String concat(final List<String> l) {
|
|
|
|
return Joiner.on(" ").skipNulls().join(l);
|
|
|
|
}
|
|
|
|
|
|
|
|
protected String cleanup(final String s) {
|
2020-01-24 10:01:11 +01:00
|
|
|
final String s0 = unicodeNormalization(s.toLowerCase());
|
2019-03-21 14:27:27 +01:00
|
|
|
final String s1 = fixAliases(s0);
|
|
|
|
final String s2 = nfd(s1);
|
2018-10-02 10:37:54 +02:00
|
|
|
final String s3 = s2.replaceAll("–", " ");
|
|
|
|
final String s4 = s3.replaceAll("&", " ");
|
|
|
|
final String s5 = s4.replaceAll(""", " ");
|
|
|
|
final String s6 = s5.replaceAll("−", " ");
|
|
|
|
final String s7 = s6.replaceAll("([0-9]+)", " $1 ");
|
2019-11-07 12:47:12 +01:00
|
|
|
final String s8 = s7.replaceAll("[^\\p{ASCII}]", "");
|
|
|
|
final String s9 = s8.replaceAll("[\\p{Punct}]", " ");
|
|
|
|
final String s10 = s9.replaceAll("\\n", " ");
|
|
|
|
final String s11 = s10.replaceAll("(?m)\\s+", " ");
|
|
|
|
final String s12 = s11.trim();
|
|
|
|
return s12;
|
2018-10-02 10:37:54 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
protected boolean checkNumbers(final String a, final String b) {
|
|
|
|
final String numbersA = getNumbers(a);
|
|
|
|
final String numbersB = getNumbers(b);
|
|
|
|
final String romansA = getRomans(a);
|
|
|
|
final String romansB = getRomans(b);
|
|
|
|
return !numbersA.equals(numbersB) || !romansA.equals(romansB);
|
|
|
|
}
|
|
|
|
|
|
|
|
protected String getRomans(final String s) {
|
|
|
|
final StringBuilder sb = new StringBuilder();
|
|
|
|
for (final String t : s.split(" ")) {
|
|
|
|
sb.append(isRoman(t) ? t : "");
|
|
|
|
}
|
|
|
|
return sb.toString();
|
|
|
|
}
|
|
|
|
|
|
|
|
protected boolean isRoman(final String s) {
|
|
|
|
return s.replaceAll("^M{0,4}(CM|CD|D?C{0,3})(XC|XL|L?X{0,3})(IX|IV|V?I{0,3})$", "qwertyuiop").equals("qwertyuiop");
|
|
|
|
}
|
|
|
|
|
|
|
|
protected String getNumbers(final String s) {
|
2019-11-28 16:54:44 +01:00
|
|
|
final StringBuilder sb = new StringBuilder();
|
|
|
|
for (final String t : s.split(" ")) {
|
|
|
|
sb.append(isNumber(t)? t : "");
|
|
|
|
}
|
|
|
|
return sb.toString();
|
|
|
|
}
|
|
|
|
|
|
|
|
public boolean isNumber(String strNum) {
|
|
|
|
if (strNum == null) {
|
|
|
|
return false;
|
|
|
|
}
|
|
|
|
return numberPattern.matcher(strNum).matches();
|
2018-10-02 10:37:54 +02:00
|
|
|
}
|
|
|
|
|
2018-11-05 17:22:59 +01:00
|
|
|
protected static String fixAliases(final String s) {
|
2018-10-02 10:37:54 +02:00
|
|
|
final StringBuilder sb = new StringBuilder();
|
|
|
|
for (final char ch : Lists.charactersOf(s)) {
|
|
|
|
final int i = StringUtils.indexOf(aliases_from, ch);
|
|
|
|
sb.append(i >= 0 ? aliases_to.charAt(i) : ch);
|
|
|
|
}
|
|
|
|
return sb.toString();
|
|
|
|
}
|
|
|
|
|
|
|
|
protected String removeSymbols(final String s) {
|
|
|
|
final StringBuilder sb = new StringBuilder();
|
|
|
|
|
|
|
|
for (final char ch : Lists.charactersOf(s)) {
|
|
|
|
sb.append(StringUtils.contains(alpha, ch) ? ch : " ");
|
|
|
|
}
|
|
|
|
return sb.toString().replaceAll("\\s+", " ");
|
|
|
|
}
|
|
|
|
|
|
|
|
protected String getFirstValue(final Field values) {
|
|
|
|
return (values != null) && !Iterables.isEmpty(values) ? Iterables.getFirst(values, EMPTY_FIELD).stringValue() : "";
|
|
|
|
}
|
|
|
|
|
|
|
|
protected boolean notNull(final String s) {
|
|
|
|
return s != null;
|
|
|
|
}
|
|
|
|
|
|
|
|
protected String normalize(final String s) {
|
2020-01-24 10:01:11 +01:00
|
|
|
return nfd(unicodeNormalization(s))
|
2019-11-07 12:47:12 +01:00
|
|
|
.toLowerCase()
|
2018-10-02 10:37:54 +02:00
|
|
|
// do not compact the regexes in a single expression, would cause StackOverflowError in case of large input strings
|
2019-11-07 12:47:12 +01:00
|
|
|
.replaceAll("[^ \\w]+", "")
|
|
|
|
.replaceAll("(\\p{InCombiningDiacriticalMarks})+", "")
|
2018-10-02 10:37:54 +02:00
|
|
|
.replaceAll("(\\p{Punct})+", " ")
|
|
|
|
.replaceAll("(\\d)+", " ")
|
|
|
|
.replaceAll("(\\n)+", " ")
|
|
|
|
.trim();
|
|
|
|
}
|
|
|
|
|
2019-11-07 12:47:12 +01:00
|
|
|
public String nfd(final String s) {
|
2018-10-02 10:37:54 +02:00
|
|
|
return Normalizer.normalize(s, Normalizer.Form.NFD);
|
|
|
|
}
|
|
|
|
|
2020-01-24 10:01:11 +01:00
|
|
|
public String unicodeNormalization(final String s) {
|
|
|
|
|
|
|
|
Matcher m = hexUnicodePattern.matcher(s);
|
|
|
|
StringBuffer buf = new StringBuffer(s.length());
|
|
|
|
while (m.find()) {
|
|
|
|
String ch = String.valueOf((char) Integer.parseInt(m.group(1), 16));
|
|
|
|
m.appendReplacement(buf, Matcher.quoteReplacement(ch));
|
|
|
|
}
|
|
|
|
m.appendTail(buf);
|
|
|
|
return buf.toString();
|
|
|
|
}
|
|
|
|
|
2018-10-02 10:37:54 +02:00
|
|
|
protected String filterStopWords(final String s, final Set<String> stopwords) {
|
|
|
|
final StringTokenizer st = new StringTokenizer(s);
|
|
|
|
final StringBuilder sb = new StringBuilder();
|
|
|
|
while (st.hasMoreTokens()) {
|
|
|
|
final String token = st.nextToken();
|
|
|
|
if (!stopwords.contains(token)) {
|
|
|
|
sb.append(token);
|
|
|
|
sb.append(" ");
|
|
|
|
}
|
|
|
|
}
|
|
|
|
return sb.toString().trim();
|
|
|
|
}
|
|
|
|
|
2019-08-06 17:06:05 +02:00
|
|
|
public String filterAllStopWords(String s) {
|
2019-03-21 14:27:27 +01:00
|
|
|
|
|
|
|
s = filterStopWords(s, stopwords_en);
|
|
|
|
s = filterStopWords(s, stopwords_de);
|
|
|
|
s = filterStopWords(s, stopwords_it);
|
|
|
|
s = filterStopWords(s, stopwords_fr);
|
|
|
|
s = filterStopWords(s, stopwords_pt);
|
|
|
|
s = filterStopWords(s, stopwords_es);
|
|
|
|
|
|
|
|
return s;
|
|
|
|
}
|
|
|
|
|
2018-10-02 10:37:54 +02:00
|
|
|
protected Collection<String> filterBlacklisted(final Collection<String> set, final Set<String> ngramBlacklist) {
|
|
|
|
final Set<String> newset = Sets.newLinkedHashSet();
|
|
|
|
for (final String s : set) {
|
|
|
|
if (!ngramBlacklist.contains(s)) {
|
|
|
|
newset.add(s);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
return newset;
|
|
|
|
}
|
|
|
|
|
|
|
|
public static Set<String> loadFromClasspath(final String classpath) {
|
|
|
|
final Set<String> h = Sets.newHashSet();
|
|
|
|
try {
|
|
|
|
for (final String s : IOUtils.readLines(NGramUtils.class.getResourceAsStream(classpath))) {
|
|
|
|
h.add(s);
|
|
|
|
}
|
|
|
|
} catch (final Throwable e) {
|
|
|
|
return Sets.newHashSet();
|
|
|
|
}
|
|
|
|
return h;
|
|
|
|
}
|
|
|
|
|
2018-11-05 17:22:59 +01:00
|
|
|
public static Map<String, String> loadMapFromClasspath(final String classpath) {
|
|
|
|
final Map<String, String> m = new HashMap<>();
|
|
|
|
try {
|
2019-08-06 17:06:05 +02:00
|
|
|
for (final String s: IOUtils.readLines(AbstractPaceFunctions.class.getResourceAsStream(classpath))) {
|
2018-11-05 17:22:59 +01:00
|
|
|
//string is like this: code;word1;word2;word3
|
|
|
|
String[] line = s.split(";");
|
|
|
|
String value = line[0];
|
2019-08-06 17:06:05 +02:00
|
|
|
for (int i=1; i<line.length;i++){
|
|
|
|
m.put(line[i].toLowerCase(),value);
|
2018-11-05 17:22:59 +01:00
|
|
|
}
|
|
|
|
}
|
|
|
|
} catch (final Throwable e){
|
|
|
|
return new HashMap<>();
|
|
|
|
}
|
|
|
|
return m;
|
|
|
|
}
|
|
|
|
|
2019-07-19 17:10:29 +02:00
|
|
|
public String removeKeywords(String s, Set<String> keywords) {
|
2018-12-20 09:54:41 +01:00
|
|
|
|
2019-07-19 17:10:29 +02:00
|
|
|
s = " " + s + " ";
|
|
|
|
for (String k: keywords ) {
|
|
|
|
s = s.replaceAll(k.toLowerCase(), "");
|
2018-12-20 09:54:41 +01:00
|
|
|
}
|
|
|
|
|
2019-07-19 17:10:29 +02:00
|
|
|
return s.trim();
|
2018-12-20 09:54:41 +01:00
|
|
|
}
|
|
|
|
|
2019-11-20 10:45:00 +01:00
|
|
|
public double commonElementsPercentage(Set<String> s1, Set<String> s2){
|
|
|
|
|
|
|
|
int longer = (s1.size()>s2.size())?s1.size():s2.size();
|
|
|
|
|
|
|
|
return (double)CollectionUtils.intersection(s1,s2).size()/(double)longer;
|
|
|
|
}
|
|
|
|
|
2019-07-19 17:10:29 +02:00
|
|
|
//convert the set of keywords to codes
|
|
|
|
public Set<String> toCodes(Set<String> keywords, Map<String, String> translationMap) {
|
|
|
|
return keywords.stream().map(s -> translationMap.get(s)).collect(Collectors.toSet());
|
|
|
|
}
|
2018-11-05 17:22:59 +01:00
|
|
|
|
2019-10-08 16:49:07 +02:00
|
|
|
public Set<String> keywordsToCodes(Set<String> keywords, Map<String, String> translationMap) {
|
2019-07-19 17:10:29 +02:00
|
|
|
return toCodes(keywords, translationMap);
|
|
|
|
}
|
2018-11-05 17:22:59 +01:00
|
|
|
|
2019-07-19 17:10:29 +02:00
|
|
|
public Set<String> citiesToCodes(Set<String> keywords) {
|
|
|
|
return toCodes(keywords, cityMap);
|
2018-11-05 17:22:59 +01:00
|
|
|
}
|
|
|
|
|
2018-12-12 16:30:03 +01:00
|
|
|
protected String firstLC(final String s) {
|
|
|
|
return StringUtils.substring(s, 0, 1).toLowerCase();
|
|
|
|
}
|
|
|
|
|
|
|
|
protected Iterable<String> tokens(final String s, final int maxTokens) {
|
|
|
|
return Iterables.limit(Splitter.on(" ").omitEmptyStrings().trimResults().split(s), maxTokens);
|
|
|
|
}
|
|
|
|
|
2019-07-08 09:44:02 +02:00
|
|
|
public String normalizePid(String pid) {
|
|
|
|
return pid.toLowerCase().replaceAll(DOI_PREFIX, "");
|
|
|
|
}
|
|
|
|
|
2019-11-07 12:47:12 +01:00
|
|
|
//get the list of keywords into the input string
|
2019-07-19 17:10:29 +02:00
|
|
|
public Set<String> getKeywords(String s1, Map<String, String> translationMap, int windowSize){
|
2019-07-08 09:44:02 +02:00
|
|
|
|
2019-08-06 17:06:05 +02:00
|
|
|
String s = s1;
|
2019-07-08 09:44:02 +02:00
|
|
|
|
|
|
|
List<String> tokens = Arrays.asList(s.toLowerCase().split(" "));
|
|
|
|
|
2019-07-08 11:01:49 +02:00
|
|
|
Set<String> codes = new HashSet<>();
|
2019-07-08 09:44:02 +02:00
|
|
|
|
|
|
|
if (tokens.size()<windowSize)
|
|
|
|
windowSize = tokens.size();
|
|
|
|
|
|
|
|
int length = windowSize;
|
|
|
|
|
|
|
|
while (length != 0) {
|
|
|
|
|
|
|
|
for (int i = 0; i<=tokens.size()-length; i++){
|
2019-11-07 12:47:12 +01:00
|
|
|
String candidate = concat(tokens.subList(i, i + length));
|
2019-07-08 09:44:02 +02:00
|
|
|
if (translationMap.containsKey(candidate)) {
|
2019-07-19 17:10:29 +02:00
|
|
|
codes.add(candidate);
|
2019-11-07 12:47:12 +01:00
|
|
|
s = s.replace(candidate, "").trim();
|
2019-07-08 09:44:02 +02:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
tokens = Arrays.asList(s.split(" "));
|
|
|
|
length-=1;
|
|
|
|
}
|
|
|
|
|
|
|
|
return codes;
|
|
|
|
}
|
2018-12-12 16:30:03 +01:00
|
|
|
|
2019-07-19 17:10:29 +02:00
|
|
|
public Set<String> getCities(String s1, int windowSize) {
|
|
|
|
return getKeywords(s1, cityMap, windowSize);
|
|
|
|
}
|
|
|
|
|
2019-10-08 14:53:52 +02:00
|
|
|
public static <T> String readFromClasspath(final String filename, final Class<T> clazz) {
|
|
|
|
final StringWriter sw = new StringWriter();
|
|
|
|
try {
|
|
|
|
IOUtils.copy(clazz.getResourceAsStream(filename), sw);
|
|
|
|
return sw.toString();
|
|
|
|
} catch (final IOException e) {
|
|
|
|
throw new RuntimeException("cannot load resource from classpath: " + filename);
|
|
|
|
}
|
|
|
|
}
|
2018-10-02 10:37:54 +02:00
|
|
|
}
|