API

[Phaqui]

The actual language tool programs are native binaries, which also requires big datasets to work. The simplified conclusion is therefore that they cannot run in the user's browser, but must run on a server. This server will be "the API", and the initial assumption was that this project is tangential to the API. However:

• This page must call the API. We can probably just build our own fetch() based wrappers to do so.
• The existing API*, does it have all required functionality, or will we have to augment / replace it with our own?
• Preliminary testing showed cors-problems. Needs further exploration and testing.

*) Will be updated with more info

[trondtynnol]

For reference, these are currently the different pipelines used by smi.cgi:

• "generate"

echo input | hfst-lookup lang/generator-gt-norm.hfstol

• "analyze"

echo input | hfst-lookup lang/analyser-gt-desc.hfstol

• "hyphenate"

echo input | hfst-lookup lang/hyphenator-gt-desc.hfstol

• "transcribe"

echo input | hfst-lookup lang/txt2ipa.compose.hfst

• "convert"

echo input | hfst-lookup lang/oldorthography2norm.compose.hfst only for kalaallisut (greenlandic) iso-code kal

• "lat2syll"

echo input | hfst-lookup Latn-to-Cans.compose.hfs only for plains cree, iso-code crk (maybe not even in use)

• "syll2lat"

echo input | hfst-lookup Cans-to-Latn.compose.hfst only for plains cree, iso-code crk (maybe not even in use)

• "disamb"

echo input | hfst-tokenize -cg lang/tokeniser-disamb-gt-desc.pmhfst | vislcg3 -g lang/disambiguator.cg3

or for some languages (fao, sma, sme, smj, nob):

echo input | hfst-tokenize -cg lang/tokeniser-disamb-gt-desc.pmhfst | vislcg3 -g lang/korp.cg3 | vislcg3 -g lang/disambiguator.cg3

• "dependency"

echo input | hfst-tokenize -cg lang/tokeniser-disamb-gt-desc.pmhfst | vislcg3 -g lang/disambiguator.cg3 | vislcg3 -g lang/dependency.cg3

or for some languages (fao, sma, sme, smj, nob):

echo input | hfst-tokenize -cg lang/tokeniser-disamb-gt-desc.pmhfst | vislcg3 -g lang/disambiguator.cg3 | vislcg3 -g lang/korp.cg3 | vislcg3 -g lang/dependency.cg3

• "paradigm" Performs "generate" for a different inflections of the input word and presents the resulting forms to the user.

• "placenames"

echo input | hfst-lookup geo.hfst

In addition, there are the number conversion programs:

• num.cgi

echo input | hfst-lookup lang/transcriptor-numbers-digit2text.filtered.lookup.hfstol

• iclock.cgi

echo input | hfst-lookup lang/transcriptor-clock-digit2text.filtered.lookup.hfstol

• idate.cgi

echo input | hfst-lookup lang/transcriptor-date-digit2text.filtered.lookup.hfstol