Home

Awesome

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

Contents

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

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.

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.

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:

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.

Notes on pronoun substitute annotation