Simple Online Phoible Interface

Detailed information

Sophoi is built using the Phoible database, which is a collection of cross-linguistic phoneme inventory data, covering over 2000 languages (and as some languages are coded more than once for different reasons, the number of coded inventories is over 3000). Phoible is not per se a collection of language descriptions, but rather a collation of ten different sub-databases (e.g. UPSID). All of the sub-databases follow slightly different transcription conventions that are not entirely normalized in the Phoible super-database. Therefore some extra processing of the Phoible data has been done for Sophoi in order to facilitate cross-linguistic comparisons. The alterations made to the Phoible data are as follows:

• Phoible transcribes post-alveolar affricates as /t̠ʃ/, but these have been simplified to /tʃ/ in Sophoi. This is just a notational convention for convenience and readability. Similarly, /ɲ̥ ŋ̥/ have been changed to have the voiceless symbol on top to not interfere with the descender: /ɲ̊ ŋ̊/.
• The UPSID sub-database occasionally indicates uncertainty in transcriptions: the Phoible symbol /t̪|t/ represents a stop that is dental or alveolar (and similarly for other manners of articulation). This is a good effort at maintaining precision when language descriptions are phonetically imprecise, but only one of ten sub-databases does this. The other sub-databases appear to mostly default to the unmarked alveolar symbols, so the underspecified UPSID symbols have been merged into alveolars.
• The breathy voiced stops of Indo-Aryan languages is represented in Phoible as /bʱ/, /b̤/, or /b̤ʱ/ seemingly at random, often differing between different descriptions of the same language. There seems to be no phonetic or phonological reason to distinguish these transcriptions, so they have been merged to /bʱ/ for Sophoi.
• Phoible mainly represents prenasalized stops as /mb nd ŋɡ/, but sometimes uses /ⁿb ⁿd ⁿɡ/. The latter have been merged into the former.
• Phoible uses two idiosyncratic fortis/lenis diacritics: fortis C͈ with a double vertical line diacritic, and lenis C͉ with a left angle diacritic. This is not a problem per se, but different sub-databases make use of this to very different extents. The ER database of Australian languages, in particular, uses a fortis/lenis transcription every time there are two stop series. This is suitable for comparisons within the ER database, but unsuitable for comparisons between languages in ER and languages transcribed in a more traditional manner, as the rest of the Phoible dataset typically is. Therefore the ER fortis/lenis distinction has been retranscribed as a voicing distinction, matching most primary sources. Be aware that this may not always represent phonetic reality (for Australian languages, but also elsewhere).
• There are several ways of transcribing a (laminal, so not retroflex) articulation intermediate between alveolar and palatal: /t̠/, /t̠ʲ/, /ȶ/, and /c̟/. All four occur in Phoible but never contrast, so for Sophoi they have been merged into the nonstandard but useful symbol /ȶ/, and similarly for other manners of articulation.

Finally it should be mentioned that Phoible provides a distinction between full and marginal phonemes. Sophoi currently only processes full phonemes, completely ignoring marginal phonemes. This may change in the future.

Sophoi was created by David Hultman, building upon the Phoible database edited by Steven Moran and Daniel McCloy. David Hultman is not involved in the Phoible project, and the editors of Phoible are not involved with Sophoi. If you have questions, comments or concerns about Sophoi, feel free to contact David at astianthus@gmail.com. Like Phoible, Sophoi is released under a CC-BY-SA license.