Resource: GlobalPhone Swahili Pronunciation Dictionary

Reference GlobalPhone Swahili Pronunciation Dictionary
Date of Submission Jan. 27, 2016, 3:07 p.m.
Status accepted
ISLRN 010-360-238-702-2
Resource Type Lexicon
Media Type Audio
Source
Language Swahili
Description

The GlobalPhone pronunciation dictionaries, created within the framework of the multilingual speech and language corpus GlobalPhone, were developed in collaboration with the Karlsruhe Institute of Technology (KIT).

The GlobalPhone pronunciation dictionaries contain the pronunciations of all word forms found in the transcription data of the GlobalPhone speech & text database. The pronunciation dictionaries are currently available in 20 languages: Arabic (29230 entries/27059 words), Bulgarian (20193 entries), Chinese-Mandarin (73388 pronunciations), Croatian (23497 entries/20628 words), Czech (33049 entries/32942 words), French (36837 entries/20710 words), German (48979 entries/46035 words), Hausa (42662 entries/42079 words), Japanese (18094 entries), Korean (3500 entries syllable-based, 97493 entries/81602 words word-based), Polish (36484 entries), Portuguese (Brazilian) (58803 entries/58787 words), Russian (28818 entries/27667 words), Spanish (Latin American) (43264 entries/33960 words), Swahili (10664 entries), Swedish (25401 entries/25356 words), Thai (small set with 12420 entries and larger set with 25570 entries/22462 words), Turkish (31330 entries/31087 words), Ukrainian (7748 entries/7740 words), and Vietnamese (38504 entries/29974 words).

1) Dictionary Encoding:
The GlobalPhone pronunciation dictionary entries consist of full word forms and are either given in the original script of that language, mostly in UTF-8 encoding (Bulgarian, Chinese-Mandarin, Croatian, Czech, French, Japanese, Korean, Polish, Portuguese, Russian, Spanish, Swahili, Turkish, Thai, Ukrainian, Vietnamese) corresponding to the trl-files of the GlobalPhone transcriptions or in a Romanized versions encoded in ASCII/ISO-8859 encoding to fit the rmn-files of the GlobalPhone transcriptions (Arabic, German, Hausa (simplified boko), Swedish). In some languages both versions exist. Romanization was performed by reversible mappings, which are documented in most cases. Furthermore, in several languages, alternative versions are available, e.g. Chinese-Mandarin is provided in both, UTF-8 for Hanzi character-based dictionary (trl) and Pinyin version in ASCII (rmn); Korean is provided in both, UTF-8 for Eojeol- and Hangul-based dictionary (trl) and ASCII for a Romanized version in which a data-driven algorithm was performed to merge syllable units into a reasonable set of word-like units (rmn).

2) Dictionary Phone set:
The phone sets for each language were derived individually from the literature following best practices for automatic speech processing. Each phone set is explained and described in the documentation using the international standards of the International Phonetic Alphabet (IPA). A language independent GlobalPhone naming convention for the phone sets is used (indicated by "M_") to support the sharing of phones across languages to build multilingual pronunciation dictionaries or acoustic models. For historical reasons, some dictionaries still use language dependent phone names. For most of those dictionaries, the documentation provides a mapping to the GlobalPhone phone names.

3) Dictionary Generation:
Whenever the grapheme-to-phoneme relationship allowed, the dictionaries were created semi-automatically. In the first step handcrafted grapheme-to-phoneme rules were applied to generate initial pronunciations from all word forms appearing in the GlobalPhone transcriptions. The number of rules highly depends on the language. In the second step, the generated pronunciations were manually checked by native speakers, correcting potential errors of the automatic pronunciation generation process. In the third step, most of the dictionaries were enriched by special entries such as acronyms, foreign words, pronunciation variants, numbers, or partial words and cross-checked by the native speakers. Most of the dictionaries have been applied to large vocabulary speech recognition. In many cases the GlobalPhone dictionaries were compared to straight-forward grapheme-based speech recognition and to alternative sources, such as Wiktionary and usually demonstrated to be superior in terms of quality, coverage, and accuracy.

4) Format:
The format of the dictionaries is the same across languages and is straight-forward. Each line consists of one word form and its pronunciation separated by blank. The pronunciation consists of a concatenation of phone symbols separated by blanks. Both, words and their pronunciations are given in tcl-script list format, i.e. enclosed in "{}", since phones can carry tags, indicating the tone, length or stress of a vowel or the palatalization of consonants, or the word boundary tag "WB", indicating the boundary of a dictionary unit. The WB tag can for example be included as a standard question in the decision tree questions for capturing crossword models in context-dependent modeling. Pronunciation variants are indicated by (<n>) with n = 2, 3, 4,... showing the number of variants per word. The order in which variants occur in the dictionary is not necessarily related to their frequency in the corpus.
Example: {word} {{w WB} o r {d WB}}

5) Documentation: The pronunciation dictionaries for each language are complemented by a documentation that describes the format of the dictionary, the phone set including its mapping to the International Phonetic Alphabet (IPA), and the frequency distribution of the phones in the dictionary. Most of the pronunciation dictionaries have been successfully applied to large vocabulary speech recognition. Experimental results and general information about the GlobalPhone corpus were published widely in conference or journal papers and partially referenced in the documentation.

A good summary of the pronunciation dictionaries is provided in:
Tanja Schultz and Tim Schlippe (2014) GlobalPhone: Pronunciation Dictionaries in 20 Languages, Proceedings of the 9th edition of the Language Resources and Evaluation Conference (LREC), Reykjavik, Iceland, 2014.

Version 1.0
Distributor ELRA