TUFS Asian Language Parallel Corpus (TALPCo)

Introduction

The TUFS Asian Language Parallel Corpus (TALPCo) is an open parallel corpus consisting of Japanese sentences and their translations into Korean, Burmese (Myanmar; the official language of the Republic of the Union of Myanmar), Malay (the national language of Malaysia, Singapore and Brunei), Indonesian, Thai, Vietnamese and English. TALPCo is licensed under a Creative Commons Attribution 4.0 International (CC BY 4.0) license. See the paper below for the details of TALPCo.

How to cite

• (For Korean, Burmese, Malay, Indonesian and English translations)
  Nomoto, Hiroki, Kenji Okano, David Moeljadi and Hideo Sawada. 2018. TUFS Asian Language Parallel Corpus (TALPCo). Proceedings of the Twenty-Fourth Annual Meeting of the Association for Natural Language Processing, 436-439.
• (For Thai and Vietnamese translations, and interpersonal meaning annotations)
  Nomoto, Hiroki, Kenji Okano, Sunisa Wittayapanyanon and Junta Nomura. 2019. Interpersonal meaning annotation for Asian language corpora: The case of TUFS Asian Language Parallel Corpus (TALPCo). Proceedings of the Twenty-Fifth Annual Meeting of the Association for Natural Language Processing, 846-849. Supplement
• (For Javanese translations)
  Lestari, Sri Buti and Yoshimi Miyake. 2022. Jawago ni mirareru taishouhyougen: Shootomuubii karano youreibuseki [Second person expressions in Javanese: An analysis of examples from short movies]. Indonesia: Gengo to Bunka 28: 65-84.
• (For Malay and Indonesian constituency tree annotations)
  Nomoto, Hiroki. 2022. Kyokushoushugi ni motoduku heiretsu tsuriibanku no kouchiku [Building a parallel treebank based on minimalism]. Proceedings of the Twenty-Eighth Annual Meeting of the Association for Natural Language Processing, 103-107.

Contents

• data_jpn.txt Japanese (raw sentences)
• data_jpn-sound.txt Japanese (links to sound files)
• data_jpn-token.txt Japanese (tokenized sentences)
• data_jpn-IPSpkr.csv Japanese (interpersonal meaning annotation, speaker)
• data_jpn-IPAddr.csv Japanese (interpersonal meaning annotation, addressee)
• data_jpn-IPLex.csv Japanese (interpersonal meaning annotation, lexical)
• data_jpn-prosub.jsonl Japanese (pronoun substitute annotation)
• data_jpn-prosub.txt Japanese (pronoun substitute annotation)

---

• data_kor.txt Korean (raw sentences)
• data_kor-sound.txt Korean (links to sound files)
• data_kor-token.txt Korean (tokenized sentences)
• data_kor-prosub.jsonl Korean (pronoun substitute annotation)
• data_kor-prosub.txt Korean (pronoun substitute annotation)

---

• data_zsm.txt Malay (raw sentences)
• data_zsm-sound.txt Malay (links to sound files)
• data_zsm-token.txt Malay (tokenized sentences)
• data_zsm-MWE.txt Malay (multiword expression list)
• data_zsm-tree.txt Malay (constituency tree annotation)
• data_zsm.jpn-zsm Malay (partial Japanese-Malay alignment)
• data_zsm-IPSpkr.csv Malay (interpersonal meaning annotation, speaker)
• data_zsm-IPAddr.csv Malay (interpersonal meaning annotation, addressee)
• data_zsm-IPLex.csv Malay (interpersonal meaning annotation, lexical)
• data_zsm-prosub.jsonl Malay (pronoun substitute annotation)
• data_zsm-prosub.txt Malay (pronoun substitute annotation)

---

• data_ind.txt Indonesian (raw sentences)
• data_ind-sound.txt Indonesian (links to sound files)
• data_ind-token.txt Indonesian (tokenized sentences)
• data_ind-MWE.txt Indonesian (multiword expression list)
• data_ind-tree.txt Indonesian (constituency tree annotation)
• data_ind.jpn-ind Indonesian (partial Japanese-Indonesian alignment)
• data_ind-IPSpkr.csv Indonesian (interpersonal meaning annotation, speaker)
• data_ind-IPAddr.csv Indonesian (interpersonal meaning annotation, addressee)
• data_ind-IPLex.csv Indonesian (interpersonal meaning annotation, lexical)
• data_ind-prosub.jsonl Indonesian (pronoun substitute annotation)
• data_ind-prosub.txt Indonesian (pronoun substitute annotation)

---

• data_jav.txt Javanese (raw sentences)
• data_jav-prosub.jsonl Javanese (pronoun substitute annotation)
• data_jav-prosub.txt Javanese (pronoun substitute annotation)

---

• data_tha.txt Thai (raw sentences)
• data_tha-sound.txt Thai (links to sound files)
• data_tha-token.txt Thai (tokenized sentences)
• data_tha.jpn-tha Thai (partial Japanese-Thai alignment)
• data_tha-IPSpkr.csv Thai (interpersonal meaning annotation, speaker)
• data_tha-IPAddr.csv Thai (interpersonal meaning annotation, addressee)
• data_tha-IPLex.csv Thai (interpersonal meaning annotation, lexical)
• data_tha-prosub.jsonl Thai (pronoun substitute annotation)
• data_tha-prosub.txt Thai (pronoun substitute annotation)

---

• data_vie.txt Vietnamese (raw sentences)
• data_vie-sound.txt Vietnamese (links to sound files)
• data_vie-token.txt Vietnamese (tokenized sentences)
• data_vie-MWE.txt Vietnamese (multi-syllable expression list)
• data_vie.jpn-vie Vietnamese (partial Japanese-Vietnamese alignment)
• data_vie-IPSpkr.csv Vietnamese (interpersonal meaning annotation, speaker)
• data_vie-IPAddr.csv Vietnamese (interpersonal meaning annotation, addressee)
• data_vie-IPLex.csv Vietnamese (interpersonal meaning annotation, lexical)
• data_vie-prosub.jsonl Vietnamese (pronoun substitute annotation)
• data_vie-prosub.txt Vietnamese (pronoun substitute annotation)

---

• data_myn.txt Burmese (raw sentences)
• data_myn-sound.txt Burmese (links to sound files)
• data_myn-token.txt Burmese (tokenized sentences)
• data_myn-ps.txt Burmese (POS-tagged sentences)
• data_myn-prosub.jsonl Burmese (pronoun substitute annotation)
• data_myn-prosub.txt Burmese (pronoun substitute annotation)

---

• data_eng.txt English (raw sentences)
• data_eng-US.txt English (US) (raw sentences) [by courtesy of Charles Kelly]

---

• readme.me (this document)

Format

All files are encoded in UTF-8 with DOS format.

Raw sentences

Sentence_ID [TAB] Sentence

1176	田中さんは 学生では ありません。
1176	Mr Tanaka is not a student.

Links to sound files

Sentence_ID [TAB] URL

Tokenized sentences

Sentence_ID [LINEBREAK] token [LINEBREAK] token [LINEBREAK] <EOS>

3627
Buku
ini
mempunyai
se-
ratus
dua
puluh
muka surat
.
<EOS>

Burmese POS-tagged sentences

Sentence_ID [TAB] Sentence

• White space: Phrasal boundary

• Dash: Morpheme boundary

  1176 n-pr-postp n pref-v-suf-suf

Constituency tree annotation

Sentence_ID.n [TAB] bracketing (for the n-th sentence for Sentence_ID)

3695.2	[S [CP [Conj Tapi] [CP [C *C_decl*] [TP [DP_a [D saya]] [T'[T *T*] [AP [AP [AdvP [Adv agak]] [AP [DP *t*<a>] [A letih]]] [AdvP [Adv sedikit]]]]]]] [PU .]]

Alignment

Sentence_ID [TAB] Japanese_token_index-target_language_token_index

1176	0-1 1-0 3-3 8-2 9-4

Interpersonal meaning annotation

See the second paper above and its supplement for the details of the interpersonal meaning feature system.

Speaker, Addressee

Sentence_ID, Gender, Marital status, Honour, Age, Social status, Role, Group, Formality, Number

3243,female,,,,neutral,,,,sg

Lexical

Token_index, token, Gender, Marital status, Honour, Age, Social status, Role, Group, Formality, Number

3845,,,,,,,,,,
0,Cô,female,,,elder.parents_younger_sibling,,parents_sibling.paternal,,,
1,tôi,,,,,neutral,,,,sg
2,làm việc,,,,,,,,,
3,ở,,,,,,,,,
4,cửa hàng,,,,,,,,,
5,hoa,,,,,,,,,
6,.,,,,,,,,,
<EOS>,,,,,,,,,,

Notes on sound files

The sound files for the following sentences come from TUFS Open Language Resources.

• Japanese: All sentences except sentences 1176, 1178, 1180, 1194, 1222, 1229, 1233, 1244, 1245, 1246, 1247, 1248, 1249, 1250, 1251, 1254, 1255, 1257, 1258, 1259, 1264, 1268, 1274, 1275, 1276, 1280, 1281, 1286, 1291, 1296, 1297, 1299, 1302, 1305, 1308, 1311, 1312, 1319, 1320, 1322, 1323, 1326, 1330, 1333, 1334, 1335, 1341, 1348, 1349, 1350, 1356, 1367, 1369, 1371, 1378, 1379, 1380, 1382, 1386, 1388, 1389, 1392, 1393, 1396, 1397, 1398, 1401, 1402, 1403, 1408, 1409, 1410, 1414, 1418, 1419, 1423, 1426, 1427, 1428, 1431, 1433, 1434, 1435, 1446, 1449, 1450, 1451, 1454, 1455, 1457, 1459, 1460, 1473, 1474, 1476, 1477, 1478, 1481, 1482, 1484, 1485, 1491, 1492, 1494, 1495, 1501, 1507, 1512, 1513, 1515, 1520, 1527, 1528, 1530, 1531, 1532, 1533, 1534, 1535, 1537, 1542, 1544, 1545, 1546, 1547, 1548, 1549, 1551, 1553, 1556, 1557, 1558, 1562, 1586, 1588, 1592, 1599, 1601, 1605, 1609, 1614, 1619, 1625, 1630, 1634, 1636, 1637, 1638, 1640, 1645, 1646, 1647, 1650, 1654, 1660, 1667, 1668, 1671, 1672, 1675, 1679, 1682, 1683, 1692, 1693, 1694, 1696, 1710, 1714, 1715, 1720, 1727, 1728, 1734, 1738, 1745, 1748, 1768, 1778, 1794, 1799, 1801, 1810, 1811, 1813, 1817, 1818, 1824, 1826, 1831, 1836, 1837, 1848, 1849, 1860, 1862, 1879, 1881, 1883, 1885, 1900, 1919, 1921, 1928, 1931, 1935, 1942, 1947, 1965, 1967, 1971, 1974, 1990, 1993, 2004, 2005, 2007, 2011, 2017, 2018, 2019, 2023, 2024, 2030, 2041, 2042, 2043, 2047, 2055, 2056, 2060, 2061, 2064, 2065, 2068, 2069, 2079, 2086, 2088, 2090, 2092, 2093, 2097, 2099, 2100, 2104, 2115, 2120, 2123, 2130, 2153, 2154, 2158, 2170, 2178, 2179, 2180, 2187, 2189, 2192, 2205, 2206, 2213, 2220, 2227, 2229, 2232, 2234, 2247, 2249, 2332, 2369, 2383, 2391, 2396, 2409, 2415, 2429, 2495, 2497, 2498, 2520, 2548, 2562, 2599, 2637, 2651, 2672, 2734, 2744, 2908, 3309, 3322 and 3751.
• Malay: All sentences except sentences 1371, 1446, 1459, 2030, 2092, 2154, 2800, 2812, 3103, 3107, 3156, 3241, 3268, 3336, 3342, 3516, 3578, 3730, 3731, 3811, 3840, 3869 and 3888.
• Burmese: All sentences.

Notes on tokenization

Malay/Indonesian

The Malay and Indonesian sentences were tokenized manually by Hiroki Nomoto and David Moeljadi, respectively. All clitics (i.e. -nya, -lah, -kah) were tokenized. In addition, the instances of the prefix se- were tokenized if they were cardinal numerals. Note that the suffix -nya and the non-numeral instances of se- were not tokenized. The following dictionaries were consulted when it was not immediately obvious whether a word sequence constituted a multiword expression.

• KBBI5. 2016. Kamus Besar Bahasa Indonesia (edisi kelima). Jakarta: Badan Pengembangan dan Pembinaan Bahasa.
• Nomoto, Hiroki. 2016. Pootaburu Nichi-Maree-Ei, Maree-Nichi-Ei Jiten [Japanese-Malay-English, Malay-Japanese-English Portable Dictionary]. Tokyo: Sanshusha.

Thai

The sentences were tokenized using the tokenize function of Deepcut and then checked by Sunisa Wittayapanyanon and Yuka Sato. The principle adopted for the manual correction is:

• Tokenize a sequence consisting of two or more syllables if and only if all constituent syllables have a meaning that contributes to the meaning of the whole phrase/sentence.
  • Do not tokenize a sequence if it contains a meaningless syllable.
  • Do not tokenize a sequence if tokenizing it will change the meaning of the whole phrase/sentence.

Vietnamese

The sentences were tokenized using the word_tokenize function of the Undersea - Vietnamese NLP Project and then checked by Junta Nomura and Hiroki Nomoto. The following dictionary was consulted when it was not immediately obvious whether a syllable sequence constituted a multi-syllable expression.

• Hoàng, Phê, ed. 2003. Từ Điển Tiếng Việt. Đà Nẵng: Nhà Xuất Bản Đà Nẵng.

Notes on pronoun substitute annotation

• See the pronoun substitute project page.
• One can modify the annotation, create a summary table and visualize the annotation by feeding the raw sentences and pronoun subtitute (prosub) annotation .txt files to ETA: Easy Text Annotator.