Skip to content

Commit

Permalink
fix komma, update dict, centralize dict
Browse files Browse the repository at this point in the history
  • Loading branch information
@bnika
bnika committed Feb 6, 2024
1 parent cce946f commit a954a8b
Show file tree
Hide file tree
Showing 10 changed files with 48 additions and 49 deletions.
2 changes: 1 addition & 1 deletion app/src/main/java/com/grammatek/simaromur/frontend/FrontendManager.java
View file
Original file line number Diff line number Diff line change
Expand Up @@ -24,9 +24,9 @@ public class FrontendManager {
private PronunciationVits mPronunciationVits = null;

public FrontendManager(Context context) {
mNormalizationManager = new NormalizationManager(context);
mPronunciation = new Pronunciation(context);
mPronunciationVits = new PronunciationVits(mPronunciation);
mNormalizationManager = new NormalizationManager(context, mPronunciation.GetIpaPronDict());
}

/**
Expand Down
5 changes: 3 additions & 2 deletions app/src/main/java/com/grammatek/simaromur/frontend/NormalizationManager.java
View file
Original file line number Diff line number Diff line change
Expand Up @@ -10,6 +10,7 @@
import java.io.InputStream;
import java.util.ArrayList;
import java.util.List;
import java.util.Map;

/**
* The NormalizationManager controls the normalization process from raw input text to
Expand All @@ -33,9 +34,9 @@ public class NormalizationManager {
private final Tokenizer mTokenizer;
private final TTSNormalizer mTTSNormalizer;

public NormalizationManager(Context context) {
public NormalizationManager(Context context, Map<String, PronDictEntry> pronDict) {
mContext = context;
mUnicodeNormalizer = new TTSUnicodeNormalizer(context);
mUnicodeNormalizer = new TTSUnicodeNormalizer(context, pronDict);
mTokenizer = new Tokenizer(context);
mTTSNormalizer = new TTSNormalizer();
mPosTagger = initPOSTagger();
Expand Down
6 changes: 3 additions & 3 deletions app/src/main/java/com/grammatek/simaromur/frontend/NumberHelper.java
View file
Original file line number Diff line number Diff line change
Expand Up @@ -104,12 +104,12 @@ private NumberHelper() {}
// DIGIT_NUMBERS.put("\\-", " <sil>");
//DIGIT_NUMBERS.put("\\-", " #");
DIGIT_NUMBERS.put("\\+", " plús");
//TODO: if we have more sentences being normalized, this replaces end-of-sentence dot as well. We don't want that
DIGIT_NUMBERS.put("\\.", " punktur");
DIGIT_NUMBERS.put(":", " tvípunktur");
//TODO: converts normal sentence commas, ask what this is supposed to do
//DIGIT_NUMBERS.put(",", " komma");
DIGIT_NUMBERS.put(",", " komma");
DIGIT_NUMBERS.put("\\/", " skástrik");
// when a semicolon occurs, we need the voice to pause
DIGIT_NUMBERS.put(";", ",");
}

public static final Map<String, String> DIGITS_ORD = new HashMap<>();
Expand Down
4 changes: 3 additions & 1 deletion app/src/main/java/com/grammatek/simaromur/frontend/TTSNormalizer.java
View file
Original file line number Diff line number Diff line change
Expand Up @@ -57,6 +57,7 @@ public class TTSNormalizer {
// This matches also inside strings for valid email addresses
private static final Pattern EMAIL_PTRN = Pattern.compile(".+@.+");
private static final Pattern EOS_PTRN = Pattern.compile("[.:?!;]");
private static final Pattern PUNCTUATION_TO_KEEP = Pattern.compile("[.:?!,\"]");
private static final Pattern NUM_OPT_DOT_PTRN = Pattern.compile("\\d+\\.?(\\d+)?");
private static final Pattern ANY_DIGIT_PTRN = Pattern.compile("\\d");
private static final Pattern SPORT_RES_PTRN = Pattern.compile("^\\d{1,2}/\\d{1,2}$");
Expand Down Expand Up @@ -170,7 +171,8 @@ else if (linksPattern != null) {
token = normalizeURL(token);
else if (token.length() > 1 && token.charAt(0) == token.charAt(1))
token = insertSpaces(token);
else if (NormalizationDictionaries.NOT_LETTER.matcher(token).matches())
else if ((token.length() > 1 && NormalizationDictionaries.NOT_LETTER.matcher(token).matches())
|| !PUNCTUATION_TO_KEEP.matcher(token).matches())
token = normalizeDigits(token);
}

Expand Down
28 changes: 6 additions & 22 deletions app/src/main/java/com/grammatek/simaromur/frontend/TTSUnicodeNormalizer.java
View file
Original file line number Diff line number Diff line change
Expand Up @@ -9,8 +9,10 @@
import java.io.InputStream;
import java.io.InputStreamReader;
import java.util.ArrayList;
import java.util.HashMap;
import java.util.HashSet;
import java.util.List;
import java.util.Map;
import java.util.Set;

/**
Expand All @@ -21,7 +23,7 @@
*/
public class TTSUnicodeNormalizer {

public static Set<String> mLexicon = new HashSet<>();
public static Map<String, PronDictEntry> mLexicon = new HashMap<>();

// The Icelandic alphabet, the grapheme set valid for automatic g2p
private final static Set<Character> CHAR_SET = new HashSet<>();
Expand Down Expand Up @@ -64,8 +66,8 @@ public class TTSUnicodeNormalizer {
// string, but delete the unknown character otherwise
private final String DONT_DELETE = "[.,\":?!-]";

public TTSUnicodeNormalizer(Context context) {
mLexicon = initLexicon(context);
public TTSUnicodeNormalizer(Context context, Map<String, PronDictEntry> pronDict) {
mLexicon = pronDict;
}

/**
Expand Down Expand Up @@ -170,7 +172,7 @@ public List<String> normalizeAlphabet(List<String> sentences) {
}

public static boolean inDictionary(String wrd) {
return mLexicon.contains(wrd.toLowerCase());
return mLexicon.containsKey(wrd.toLowerCase());
}

private boolean isTag(String wrd) {
Expand Down Expand Up @@ -206,22 +208,4 @@ private String getIceAlphaReplacement(Character c) {

return "";
}

private Set<String> initLexicon(Context context) {
Set<String> lexicon = new HashSet<>();
Resources res = context.getResources();
String line;
try {
InputStream is = res.openRawResource(R.raw.lexicon_v2201);
BufferedReader reader = new BufferedReader(new InputStreamReader(is));
if (is != null) {
while ((line = reader.readLine()) != null) {
lexicon.add(line.trim());
}
}
} catch (Exception e) {
e.printStackTrace();
}
return lexicon;
}
}
1 change: 1 addition & 0 deletions app/src/main/java/com/grammatek/simaromur/frontend/UnicodeMaps.java
View file
Original file line number Diff line number Diff line change
Expand Up @@ -95,6 +95,7 @@ public class UnicodeMaps {
postDictLookupMap.put('Ä', "E");
postDictLookupMap.put('Ü', "U");
postDictLookupMap.put('Ø', "Ö");
postDictLookupMap.put('Ô', "Ó");
}

// delete in MVP, transliterate later if necessary
Expand Down
7 changes: 2 additions & 5 deletions app/src/main/res/raw/igc_wiki_news_dict_20240107.csv
View file
Original file line number Diff line number Diff line change
Expand Up @@ -1061,7 +1061,6 @@ sigríður s ˈ ɪː ɣ r i ð ʏ r
group k r ˈ u h p
lán l ˈ auː n
árangri ˈ auː r au ŋ k r ɪ
em ˈ ɛː ɛ m
íbúar ˈ iː p u a r
dagskrá t ˈ aː ɣ s k r au
byrjaði p ˈ ɪ r j a ð ɪ
Expand Down Expand Up @@ -5011,7 +5010,6 @@ hleypt l̥ ˈ ei f t
andersen ˈ a n t ɛ r̥ s ɛ n
hálftíma h ˈ au l v tʰ i m a
flýta f l ˈ iː t a
hí h ˈ iː
kynjanna cʰ ˈ ɪ n j a n a
munaði m ˈ ʏː n a ð ɪ
landamæri l ˈ a n t a m ˌ ai r ɪ
Expand Down Expand Up @@ -22263,6 +22261,7 @@ orrustunni ˈ ɔ r ʏ s t ʏ n ɪ
rúmið r ˈ uː m ɪ ð
útfrá ˈ uː t f r au
space s p ˈ eiː s
spacex s p ˈ eiː s ˌ ɛ k s
classic kʰ l ˈ a s ɪ k
hernaðaraðgerðum h ˈ ɛ r t n a ð a r ˌ a ð c ɛ r ð ʏ m
hindrunum h ˈ ɪ n t r ʏ n ʏ m
Expand Down Expand Up @@ -25383,7 +25382,6 @@ vítinu v ˈ iː t ɪ n ʏ
úrskurðarnefndin ˈ uː r s k ʏ r ð a r n ɛ m t ɪ n
ómari ˈ ouː m a r ɪ
ömurlega ˈ œː m ʏ r l ɛ ɣ a
mbl ˈ ɛ m p l
vinnuveitenda v ˈ ɪ n ʏ v ˌ ei t ɛ n t a
hljóðs l̥ j ˈ ou ð s
ágengt ˈ auː c ei ŋ̊ t
Expand Down Expand Up @@ -69363,7 +69361,7 @@ svarendum s v ˈ aː r ɛ n t ʏ m
viðbjóðslegt v ˈ ɪ ð p j ou ð s t l ɛ x t
hljóðmengun l̥ j ˈ ouː ð m ei ŋ k ʏ n
óttans ˈ ou h t a n s
dísella t ˈ iː s ɛ t l a
dísella t ˈ iː s ɛ l a
mönnunarvanda m ˈ œ n ʏ n a r v ˌ a n t a
mælinganna m ˈ aiː l i ŋ k a n a
acoff ˈ aː k ɔ f
Expand Down Expand Up @@ -150437,7 +150435,6 @@ starfshlutföll s t ˈ a r f s l̥ ʏ t f ˌ œ t l̥
húsgrunninum h ˈ uː s k r ʏ n ɪ n ʏ m
aukaútgjöld ˈ œyː k a ˌ u t c œ l t
frístundarheimili f r ˈ iː s t ʏ n t a r h ˌ ei m ɪ l ɪ
adhd ˈ a t n̥ t
tryggingageiranum tʰ r ˈ ɪ c i ŋ k a c ˌ ei r a n ʏ m
þjark θ j ˈ a r̥ k
tryggra tʰ r ˈ ɪ k r a
Expand Down
33 changes: 20 additions & 13 deletions app/src/test/java/com/grammatek/simaromur/NormalizationManagerTest.java
View file
Original file line number Diff line number Diff line change
Expand Up @@ -6,6 +6,8 @@
import androidx.test.core.app.ApplicationProvider;

import com.grammatek.simaromur.frontend.NormalizationManager;
import com.grammatek.simaromur.frontend.PronDictEntry;
import com.grammatek.simaromur.frontend.Pronunciation;

import org.junit.Test;
import org.junit.runner.RunWith;
Expand All @@ -25,21 +27,23 @@
public class NormalizationManagerTest {

private final static Context context = ApplicationProvider.getApplicationContext();
private final static Pronunciation pron = new Pronunciation(context);
private final static Map<String, PronDictEntry> pronDict = pron.GetIpaPronDict();

@Test
public void processTest() {
String input = "www.mbl.is/frettir";
NormalizationManager manager = new NormalizationManager(context);
String input = "Space-X";
NormalizationManager manager = new NormalizationManager(context, pronDict);
String processed = manager.process(input);
System.out.println(processed);

assertEquals("m b l punktur is skástrik frettir .",
assertEquals("space - x .",
processed);
}

@Test
public void processDigitsTest() {
NormalizationManager manager = new NormalizationManager(context);
NormalizationManager manager = new NormalizationManager(context, pronDict);
for (String sent : getDigits().keySet()) {
String processed = manager.process(sent);
assertEquals(getDigits().get(sent), processed);
Expand All @@ -48,7 +52,7 @@ public void processDigitsTest() {

@Test
public void processSymbolsTest() {
NormalizationManager manager = new NormalizationManager(context);
NormalizationManager manager = new NormalizationManager(context, pronDict);
for (String sent : getSymbols().keySet()) {
String processed = manager.process(sent);
assertEquals(getSymbols().get(sent), processed);
Expand All @@ -57,7 +61,7 @@ public void processSymbolsTest() {

@Test
public void processNewIssuesTest() {
NormalizationManager manager = new NormalizationManager(context);
NormalizationManager manager = new NormalizationManager(context, pronDict);
for (String sent : getNewTestSentences().keySet()) {
String processed = manager.process(sent);
assertEquals(getNewTestSentences().get(sent), processed);
Expand All @@ -66,7 +70,7 @@ public void processNewIssuesTest() {

@Test
public void processV14IssuesTest() {
NormalizationManager manager = new NormalizationManager(context);
NormalizationManager manager = new NormalizationManager(context, pronDict);
for (String sent : getV14TestSentences().keySet()) {
String processed = manager.process(sent);
assertEquals(getV14TestSentences().get(sent), processed);
Expand All @@ -75,7 +79,7 @@ public void processV14IssuesTest() {

@Test
public void processListTest() {
NormalizationManager manager = new NormalizationManager(context);
NormalizationManager manager = new NormalizationManager(context, pronDict);
for (String sent : getTestSentences().keySet()) {
String processed = manager.process(sent);
assertEquals(getTestSentences().get(sent), processed);
Expand Down Expand Up @@ -169,6 +173,12 @@ private Map<String, String> getNewTestSentences() {
private Map<String, String> getV14TestSentences() {
// test sentences added for the deployment of v1.4
Map<String, String> sent = new HashMap<>();
sent.put("láta skoðanir sínar í ljós í athugasemdum.“, segir hún.",
"láta skoðanir sínar í ljós í athugasemdum \" . , segir hún .");
sent.put("Ôlafsson, framkvæmdastjóri Stakka víkur ehf., segir",
"ólafsson , framkvæmdastjóri stakka víkur e h f . , segir .");
sent.put("Stimpilgjald af kaupsamningi er 0,8% af heildarfasteignamati hjá einstaklingum.",
"stimpilgjald af kaupsamningi er núll komma átta prósent af heildarfasteignamati hjá einstaklingum .");
sent.put("íbúðin er 145 fm", "íbúðin er hundrað fjörutíu og fimm fermetrar .");
sent.put("margir með ADHD", "margir með a d h d .");
sent.put("Má áætla að þriðji stóri íþróttaviðburðurinn sem horft hafi verið til sé EM " +
Expand All @@ -185,15 +195,12 @@ private Map<String, String> getV14TestSentences() {
"leigubíl fyrir tvö þúsund og fimm hundruð krónur tæpar " +
"þrjátíu og þrjú þúsund íslenskar krónur .");
sent.put("leigubíl fyrir £377 á dag.", "leigubíl fyrir þrjú hundruð sjötíu og sjö pund á dag .");

sent.put("3.7", "þrír punktur sjö .");
sent.put("13.7", "einn þrír punktur sjö .");

sent.put("mbl.is/frettir/innlent/2024/02/02/litlu_hlutirnir_sem_folkid_saknar_helst/",
"m b l punktur is skástrik frettir skástrik innlent skástrik tvö þúsund tuttugu " +
"og fjögur skástrik núll tvö skástrik núll tvö skástrik litlu hlutirnir sem " +
"folkid saknar helst skástrik .");

sent.put("Stjórnendur Rúv gera ráð fyrir því að augýsingatekjur stofnunarinnar hækki " +
"á þessu ári um 17,4% frá fyrra ári og að útvarpsgjald hækki um 3,5%.",
"stjórnendur rúv gera ráð fyrir því að augýsingatekjur stofnunarinnar hækki " +
Expand Down Expand Up @@ -246,7 +253,7 @@ private Map<String, String> getTestSentences() {
//testSentences.put("Stelpurnar Carmen Diljá Guðbjarnardóttir og Elenora Rós Georgsdóttir söfnuðu 7.046 kr.",
// "Stelpurnar Carmen Diljá Guðbjarnardóttir og Elenora Rós Georgsdóttir söfnuðu sjö þúsund fjörutíu og sex krónum .");
testSentences.put("Stelpurnar Carmen Diljá Guðbjarnardóttir og Elenora Rós Georgsdóttir söfnuðu 7.046 kr.",
"Stelpurnar Karmen Diljá Guðbjarnardóttir og Elenora Rós Georgsdóttir söfnuðu sjö þúsund fjörutíu og sex krónur .".toLowerCase());
"Stelpurnar Carmen Diljá Guðbjarnardóttir og Elenora Rós Georgsdóttir söfnuðu sjö þúsund fjörutíu og sex krónur .".toLowerCase());
testSentences.put("Hann skoraði 21 stig og tók 12 fráköst.", "Hann skoraði tuttugu og eitt stig og tók tólf fráköst .".toLowerCase());
testSentences.put("Opna Suðurnesjamótið í pílu fer fram þann 4. desember nk. kl. 13:00 í píluaðstöðu Pílufélags Reykjanesbæjar að Hrannargötu 6. ",
"Opna Suðurnesjamótið í pílu fer fram þann fjórða desember næstkomandi klukkan þrettán núll núll ".toLowerCase() +
Expand Down Expand Up @@ -286,7 +293,7 @@ private Map<String, String> getTestSentences() {
"Hollenska fjárfestingafyrirtækið EsBro hyggst reisa fimmtán hektarar <sil> hundrað og fimmtíu þúsund fermetrar <sil> gróðurhús til framleiðslu á tómötum .".toLowerCase());
//testSentences.put("Við Lindarhvamm í Hafnarfirði er að finna 134 fm efri sérhæð og ris í snyrtilegu tvíbýlishúsi sem reist var árið 1963.",
// "við lindarhvamm í hafnarfirði er að finna hundrað þrjátíu og fjögur fermetrar efri sérhæð og ris í snyrtilegu tvíbýlishúsi sem reist var árið nítján hundruð sextíu og þrjú .".toLowerCase());
testSentences.put("Mynd / [email protected]", "Mynd skástrik e l g hjá v f punktur is .".toLowerCase());
testSentences.put("Mynd / [email protected]", "Mynd skástrik elg hjá v f punktur is .".toLowerCase());
testSentences.put("hefur leikið sjö leiki með U-21 árs liðinu.", "hefur leikið sjö leiki með U - tuttugu og eins árs liðinu .".toLowerCase());
testSentences.put("er þetta í 23. skiptið sem mótið er haldið .", "er þetta í tuttugasta og þriðja skiptið sem mótið er haldið .".toLowerCase());
testSentences.put("Skráning er hafin á http://keflavik.is/fimleikar/ og ef eitthvað er óljóst er hægt að hafa samband í síma 421-6368 eða á [email protected]",
Expand Down
8 changes: 6 additions & 2 deletions app/src/test/java/com/grammatek/simaromur/TTSUnicodeNormalizerTest.java
View file
Original file line number Diff line number Diff line change
Expand Up @@ -3,6 +3,8 @@
import android.content.Context;
import android.os.Build;

import com.grammatek.simaromur.frontend.PronDictEntry;
import com.grammatek.simaromur.frontend.Pronunciation;
import com.grammatek.simaromur.frontend.TTSUnicodeNormalizer;

import org.junit.Test;
Expand All @@ -27,10 +29,12 @@
public class TTSUnicodeNormalizerTest {

private final static Context context = ApplicationProvider.getApplicationContext();
private final static Pronunciation pron = new Pronunciation(context);
private final static Map<String, PronDictEntry> pronDict = pron.GetIpaPronDict();

@Test
public void unicodeNormalizingTest() {
TTSUnicodeNormalizer normalizer = new TTSUnicodeNormalizer(context);
TTSUnicodeNormalizer normalizer = new TTSUnicodeNormalizer(context, pronDict);
String input = "„ Við vorum samheldnir og þéttir og það er gott að innbyrða sigur á útivelli gegn öflugu liði eins og Breiðabliki , “ sagði Willum Þór Þórsson";
String normalized = normalizer.normalizeEncoding(input);
assertEquals("\" Við vorum samheldnir og þéttir og það er gott að innbyrða sigur á útivelli gegn öflugu liði eins og Breiðabliki , \" sagði Willum Þór Þórsson", normalized);
Expand All @@ -42,7 +46,7 @@ public void unicodeNormalizingTest() {

@Test
public void alphabetNormalizingTest() {
TTSUnicodeNormalizer normalizer = new TTSUnicodeNormalizer(context);
TTSUnicodeNormalizer normalizer = new TTSUnicodeNormalizer(context, pronDict);
for (String sent : getTestSentences().keySet()) {
List<String> processed = normalizer.normalizeAlphabet(Arrays.asList(sent));
assertEquals(getTestSentences().get(sent), processed.get(0));
Expand Down
3 changes: 3 additions & 0 deletions app/src/test/java/com/grammatek/simaromur/TokenizerTest.java
View file
Original file line number Diff line number Diff line change
Expand Up @@ -53,6 +53,9 @@ public void tokenizerCommaSepTest() {
input = "14,1% íbúa";
tokenized = tok.detectSentences(input);
assertEquals("14,1 % íbúa .", tokenized.get(0));
input = "kaupsamningi er 0,8% af heildarfasteignamati";
tokenized = tok.detectSentences(input);
assertEquals("kaupsamningi er 0,8 % af heildarfasteignamati .", tokenized.get(0));
}

@Test
Expand Down

0 comments on commit a954a8b

Please sign in to comment.