This is an attempt to create a cross-linguistic phonetic alphabet, realized as a dialect of IPA, for cross-linguistic approaches to language comparison.
The basic idea is to provide a fixed set of symbols for phonetic representation along with a full description regarding their pronunciation following the tradition of IPA. This list is essentially expandable, when new language data arises, and can be linked to alternative datasets, like Mielke's (2008) P-Base, and PHOIBLE.
In addition to the mere description of symbols, we provide also a range of scripts that can be used in order to test how well a dataset reflects our cross-linguistic standard, and to which degree it diverges from it. In this way, linguists who want to publish their data in phonetic transcriptions that follow a strict standard, they can use our tools and map their data to CLPA. By conforming to our whitelist (and informing us in cases where we miss important sounds that are essential for the description of a dataset so that we can expand the CLPA), the community can make sure that we have a maximal degree of comparability across lexical datasets.
Our initial dataset (file clpa/clpa-data/clpa.main.json) currently consists of 1192 symbols, including consonants, vowels, diphtongs, tones, and three markers (for word and morpheme boundaries). The original data is inspired by the IPA description used in the P-Base project, and we mostly follow their symbol conventions, but we added tone letters and symbols which were missing in their inventory and also re-arranged their descripting features into more classes which are now differently defined for the main classes of sounds.
Additionally, the dataset contains sets of instructions for conversion of symbols which do not occur in our whitelist. Here, we distinguish between:
- explicit mappings (clpa/clpa-data/explicit.tsv) of input segments to output segments. As an example, consider [ʔʲ] which we map to [ʔj], or [uu], which we map to [uː].
- alias symbols (clpa/clpa-data/alias.tsv), which are one-to-more mappings of symbols of length 1 in unicode, and are regularly applied to a symbol if we can't find it in our whitelist. As an example, consider [ʦ] which we map to [ts].
- symbols to be ignored (clpa/clpa-data/delete.tsv), which are symbols of length 1 which we ignore from the input data and then check whether we can find a mapping. As a an example, compare the combinging mark in the symbols [t͡s], which we delete in order to map to our [ts].
- symbols to be converted as patterns (clpa/patterns.tsv): these are potentially risky operations which we try to minimize as well as possible, but there are situations in which it is useful to apply changes on a pattern basis, as for example, in datasets in which "aspiration" is not marked by a superscript letter, where we would then turn every instance of plosive + h into plosive + ʰ
In order to test the current conversion procedure, run
$ clpa report FILENAME
in the shell. Your inputfile should be a tab-separated file in LingPy-Wordlist format, with your phonetic sequences being represented as space-segmented values in a column "TOKENS" of the input file. This is a mere proof-of-concept at the moment, and the script will be further enhanced.
If you specify further:
$ clpa report FILENAME outfile=NEWFILENAME
the data will also be written to file.
Furthermore, choose between two attributes for the format, namely "csv" and "md", for example:
$ clpa report FILENAME format=csv
will write data in CSV format.
To get your wordlist annotated with new columns CLPA_TOKENS
and CLPA_IDS
run
$ clpa annotate FILENAME
To convert a single string and see how well it converts, just type:
$ clpa check "a b ä/ x/y /"
a b ä/ x/y /
a b ä y �
Here, it is important to set your string in quotes, since it needs to be passed as one argument in space-separated form to the interpreter. As a result, an alignment will be returned in which a star indicates that the character is recognized in CLPA and a question mark indicates it is missing.
We should not take this feature set too literally, but we try to define each segment in CLPA by providing features which are largely inspired by the IPA. In this we follow the idea of the Fonetikode.
Currently, we distinguish the following feature sets:
- Basic types: consonant, diphtong, marker, tone, vowel. A marker are symbols that we use for extended annotation, like morpheme boundaries, word boundaries.
- Each type has a different feature set. New features sets can only be added to the data if they create a unique feature vector that distinguishes the glyphs of a given class from all other glyphs. Testing will be done via a test suite (not yet implemented).
- For each identifier, further metadata can be provided, be it mappings to other datasets, like Fonetikode, Phoible, etc., or to frequently occurring aliases, etc., also information in terms of "notes" is something that would be possible.