Huteson (2003) Rukai Survey
This corpus contains the 29 Appendix B imitation-test sentences in Greg Huteson's 2003 survey: 14 Maga (Maolin) and 15 Tona (Dona) examples. Each has a natural source FORM, analyzed W forms, English translations, Ortho113 forms and generated phonology. Fifteen sentences have source morpheme analysis; fourteen have word glosses without an invented M tier. There are 103 W and 70 M. Eight word-gloss cells are blank in the source: seven stay unglossed and one carries a gloss inferred from the corpus's own usage, marked as inferred in the data. No audio is included.
Corpus Statistics
Rukai Dona
Rukai Maolin
Word count
55
48
Total audio
0
0
Transcribed
0
0
Untranscribed
0
0
Translated words
English
55
48
Mandarin
0
0
Japanese
0
0
Dutch
0
0
Morphologically segmented
29
23
Glossed words
29
23
Notes and Issues
Four things a user should know before comparing this corpus with FormosanBank's other Rukai data. All are properties of the 2003 source, not defects in the port.
Maga (Maolin) does not distinguish schwa. Ortho113 gives Maolin a seven-vowel system, writing [e] as é and [ə] as e. Huteson's Maga transcription uses six vowel symbols and never writes ə at all, though he writes it 60 times in the Tona chapter. His e is therefore read as [e] throughout and standardizes to é. Where the modern orthography has [ə] in these words — isierkɨ 'sleep', spelled sierkɨ in ePark, is the clear case — the distinction is simply not recorded in this source, and the standard tier follows the author rather than reconciling him.
Word-final vowels are written double. All eleven doubled vowels in the Maga data are word-final (ikee, knee, mamaa, broo …); the Tona data has none in that position. The doubling is preserved as the author's, so the standard tier writes ikéé where the bank's other Maolin corpus writes iké. Of 41 Maolin word types, 23 match the bank's spelling directly and a further 7 match once word-final doubling is collapsed. Whether the doubling is phonemic length or predictable final lengthening is unresolved.
The particle na is never glossed. All seven occurrences are blank in the source's aligned display and no meaning has been coined for them: the word, and its morpheme where the sentence is parsed, carry a form and no translation. Seven further word-gloss cells are blank for the same reason, and Tona 9's a-kakə carries one fused source gloss 1S.TOP for two morphemes without either being given an invented meaning. One gloss is inferred, not from the source: page 43 leaves the ki column of Tona 14 blank, and since every other ki in the corpus is glossed NOM and none is glossed anything else, NOM is supplied there and marked as inferred.
Case marking differs from the bank's other Rukai sources. Huteson glosses ki as NOM; ePark and the NTU Formosan Corpus gloss it OBL or GEN, never NOM. The source's own analysis is preserved.
Access Details
The XML and reproduction files are maintained in the FormosanBank repository. The source archive identifies Huteson's report. This corpus covers its Appendix B, with all 46 PDF pages accounted for in the source inventory.
Corpus Processing
The reviewed transcription preserves natural and analyzed lines separately, with stable example/page identities. The PDF's legacy embedded fonts require visual source review. The sentence tier uses the natural test-list line and the word tier the analyzed one, so the interlinear's hyphens appear at word and morpheme level but not at the sentence. Expert corrections to ɖ and the literal alternate translation in Tona 15 remain. Tona 9 retains printed a-kakə and the combined word gloss 1S.TOP; no meanings are invented for its two individual morphemes. Four translation shorthands become same-sentence alternatives, and every gloss is the source's apart from the single inferred one noted above.
The source's writing system is registered as a scheme like any other: Orthographies/Huteson/Rukai.tsv maps its letters to IPA for the original phonology, and Orthographies/ConversionTables/Rukai_Huteson_113.tsv converts it to Ortho113 for the standard tier. It is a mixed system rather than plain IPA — c is [ʦ] and y is [j] — so the conversion table carries rules only where those two systems disagree.
The corpus's CodeAndDocs/generate_xml.sh rebuilds from committed transcription inputs and current shared cleaning, standardization and phonology tools. No private repository, source download or historical Git objects are needed. Tool provenance records the actual build, without pinning later runs. Reproduction and validation instructions include the source-specific gloss findings and correction history.
Copyright
License: CC BY-NC-SA 4.0
SIL's terms of use state: "Unless stated otherwise in the file or item description, all items are available under the Creative Commons Attribution-Noncommercial-Share Alike 4.0 Unported License." Attribution to Greg Huteson and SIL International, noncommercial use and ShareAlike distribution are therefore required. The corpus is also subject to FormosanBank's Terms of Use and AI Use and Commercial Licensing.
Citation
Under the Terms of Use, cite FormosanBank and:
Huteson, Greg. 2003. Sociolinguistic Survey Report for the Tona and Maga Dialects of the Rukai Language. SIL International.
Last updated