Skip to content


Switch branches/tags

Latest commit


Git stats


Failed to load latest commit information.
Latest commit message
Commit time

Index Phonemica

Data format

Doculect entries are stored as Python INI files, named with the Glottocode of the language followed by a hyphen and an index number: 1 for the first entry for that Glottocode, 2 for the second, etc.

Entry files have five headers: core, source, (optionally) notes, phonemes, and allophonic_rules. An optional todo section is also permitted.


core stores two required attributes:

  • name: the name of the doculect as given in the source
  • glottocode: the Glottocode of the language

And two optional attributes:

  • dialect: the Glottocode of the specific dialect, if one is defined
  • dialect_name: the name of the specific dialect as given in the source, if a specific dialect is referenced


source stores many attributes, of which the most common ones are:

  • glottolog
  • url
  • doi
  • author
  • title
  • publisher
  • volume
  • number
  • year

Enough information should be given that the paper can be found. At the minimum, a Glottolog ID should be provided if one is available; other information can then be added automatically from Glottolog when a numbered release of the Index is built.


notes stores notes relevant to the doculect entry.


phonotactics stores information about the language's syllable structure; currently this contains only two fields, max_initial and max_final. If a doculect's source does not provide the necessary information, the special value no_info may be stored in this field to reflect this.


phonemes stores a set of phonemes, separated by newlines.

To mark a phoneme as marginal, enclose it in parentheses.

To mark a phoneme as only occurring in non-nativized loans, enclose it in curly brackets.

To mark a phoneme as marginal outside non-nativized loans, enclose it in parentheses and curly brackets.

In some cases, phonemes may be too underspecified or under-described to be easily reducible to one IPA representation, as with the Rotokas voiced series, or coronal plosives that may be either dental or alveolar. Indicate these cases by listing the candidate representations separated by vertical bars, with the canonical representation used by the source document in the first position.


allophonic_rules stores a set of allophonic rules, written in source > realization / environment format.

  • The source must be a phoneme listed in phonemes.
  • The realization must be a phoneme.
  • The environment is optional, and may be free-form text.

In cases where an entire cluster or sequence has a specific realization, such as English /nð/ > [n̪ː], join the source phonemes in the sequence with a plus sign: n+ð > n̪ː. If this rule has no conditioning factor outside the cluster itself, the / environment component may be omitted.

For cases of free variation, such as Nuosu m+ɨ >~ m̩, use the digraph >~. For cases of free variation among obligatory conditioned allophones, such as t > s ~ ts / _i in Rotokas, use > and separate the variants with ~.

Non-IPA conventions


The frication diacritic is carried over from PHOIBLE: for example, the voiced velar lateral fricative is ʟ͓.

The retroflex lateral flap is written ɺ̢.

The IPA palatal series is here interpreted as velar palatals; coronal palatals are represented by the Sinological ȶ series.

Affricates and consonants with bilabially trilled release are assumed to agree in voice unless otherwise specified.

Prenasalized consonants are written with preceding superscript n: for example, ⁿp instead of mp or m̥p. Postnasalized consonants or prestopped nasals are written as digraphs: pm̥ or pm (depending on whether the nasal element is voiced) instead of pⁿ.


Fricated or 'super-close' vowels such as Mandarin -i are written with extensions of the Sinological characters:

  • ɿ instead of
  • ɿᶾ instead of ʒ̩
  • ʅ instead of ʐ̩
  • ɿᶽ instead of ʑ̩
  • instead of

There may eventually be a ʮ series also, but we haven't needed one yet.

The retraction diacritic on vowels is used in the Tibeto-Burman manner, to represent the 'tight throat' quality or 'tense voice' that appears in Liangshan Yi and Bai. Uvularization (which appears in some Qiangic languages) is transcribed with a following ʶ.

ʵ replaces ˞ as a marker of rhoticity.


Tone is written with Chao tone letters. The super-high 66 tone of Bai is written ˥́.

Other conventions


Inventories of 'eroded' Sino-Tibetan languages are typically given as onsets, rimes, and tones. We convert these to inventories of consonants and vowels, and err on the side of segmental simplicity, although complex rimes may be represented as unit segments in certain cases where we can identify good reason to do so.


The non-syllabicity diacritic is used on diphthongs when:

  • it is used in the source
  • every diphthong given in the source is closing, with the possible exception of close-to-close diphthongs (since in these cases, it's likely that all diphthongs are falling in prominence)

If diphthongs that are not closing or close-to-close are present and the source does not use the non-syllabicity diacritic, it is not used.

For example, if a source lists a diphthong inventory of /ai au ei eu oi ou iu/, these diphthongs will be input as /ai̯ au̯ ei̯ eu̯ oi̯ ou̯ iu̯/. But if a source lists /ai au ea oa/, these will be input as /ai au ea oa/, since it isn't clear whether /ea oa/ are falling or rising in prominence.


SIL Organized Phonology Data sheets almost always list the low vowel as /ɑ/ rather than /a/. In these cases, /a/ will be input unless the low vowel is clearly described to be backed.


An example file, roto1249-1.ini, is given below.

name = Rotokas
glottocode = roto1249

author = Firchow, Irwin; Firchow, Jacqueline
title = An Abbreviated Phoneme Inventory
publisher = Anthropological Linguistics
volume = 11
number = 9
year = 1969
pages = 271-276
glottolog = 110896
url =

max_initial = 1
max_final = 0


t > s ~ ts / _i

Blank file

glottocode = REQUIRED
notes = OPTIONAL
dialect = OPTIONAL

glottolog = IDEAL
url = IDEAL
author = OPTIONAL (but REQUIRED if there's no glottolog ID)
title = OPTIONAL (but REQUIRED if there's no glottolog ID)
publisher = OPTIONAL
volume = OPTIONAL
number = OPTIONAL
year = OPTIONAL (but REQUIRED if there's no glottolog ID)
pages = OPTIONAL

max_initial = REQUIRED IF no_info IS NOT PRESENT
max_final = REQUIRED IF no_info IS NOT PRESENT




Creates a blank doculect file. Usage: >python <glottocode>. Prints the name of the created file. Notable options:

  • -h: Display a help message listing all script options (can be used without providing a glottocode)
  • -b <bibkey>: Convenience option for inputting the Glottolog bibkey of the source; this will auto-fill as many of the source fields as possible, but requires a local copy of the Glottolog database and the installation of the pyglottolog library.
  • -n <name>: Convenience option for inputting the name of the doculect as given in the source
  • --simple: Omits unfilled optional keys and default text

For example, roto1249 -b sil16:10670 -n Rotokas --simple.

Validates and git adds a provided entry in .ini format; for example, roto1249-1.

Prints statistics about the local database. Run help to list available reports.


No description, website, or topics provided.






No packages published