def test_syllabify():
seq1 = "t i a o ¹ b u ² d a o"
seq2 = "jabloko"
seq3 = "jabəlko"
seq4 = "j a b u - k o"
assert_raises(ValueError, syllabify, seq1, output="test")
assert syllabify(seq1, output="flat").count(rc('morpheme_separator')) == 2
assert syllabify(seq2, output="breakpoints")[0] == (0,2)
assert syllabify(seq3, output="nested")[1] == list("bəl")
assert syllabify(seq4, output="nested")[1] == list("bu-")
def test_syllabify(self):
seq1 = "t i a o ¹ b u ² d a o"
seq2 = "jabloko"
seq3 = "jabəlko"
seq4 = "j a b u - k o"
seq5 = "ma⁵io"
assert_raises(ValueError, syllabify, seq1, output="test")
assert syllabify(seq1,
output="flat").count(rc('morpheme_separator')) == 2
assert syllabify(seq2, output="breakpoints")[0] == (0, 2)
assert syllabify(seq3, output="nested")[1] == list("bəl")
assert syllabify(seq4, output="nested")[1] == list("bu-")
assert ''.join(syllabify(seq5, sep="+")).split('+')[-1] == 'io'
def prepare(ds):
rawdata = csv2list(ds.raw('Burling1967.csv'), strip_lines=False)
concepts = []
header = rawdata[0]
crect = {}
D = {
0: [
'doculect', 'concept', 'doculect_is', 'concept_number_is',
'value_is', 'value', 'notes', 'notes_mm', 'cogid', 'segments'
]
}
idx = 1
cogid = 1
for i, line in enumerate(rawdata[1:]):
tmp = dict(zip(header, line))
concepts += [(tmp['number'], tmp['meaning'])]
for lid, lang in sorted(ds.lid2lang.items()):
cell = tmp[lid]
if cell.strip() != '-' and cell.strip():
ccell = crect.get(cell, cell)
syls = ' + '.join(
[' '.join(x) for x in syllabify(ccell, output='nested')])
D[idx] = [
lang, tmp['meaning'], lid, tmp['number'], cell, ccell,
tmp['notes'], tmp['MM notes'], cogid, syls
]
idx += 1
cogid += 1
with open(ds.raw('concepts.tsv'), 'w') as f:
f.write('NUMBER\tENGLISH\n')
for a, b in concepts:
f.write(a + '\t' + b + '\n')
ds.write_wordlist(Wordlist(D))
def lexeme2syllables(lexeme):
# Process the lexeme; the syllabification algorithm is stricter,
# so we cannot have the option of keeping slashes, marks, etc
phonemes = [
phoneme.split("/")[1] if "/" in phoneme else phoneme
for phoneme in lexeme.split()
]
phonemes = [phoneme for phoneme in phonemes if phoneme not in ["+", "_"]]
clean_lexeme = " ".join(phonemes)
# Get the syllables
syllables = syllabify(clean_lexeme, output="nested")
return syllables
def prepare(ds):
rawdata = csv2list(ds.raw('Burling1967.csv'), strip_lines=False)
concepts = []
header = rawdata[0]
crect = {}
D = {0: [
'doculect',
'concept',
'doculect_is',
'concept_number_is',
'value_is',
'value',
'notes',
'notes_mm',
'cogid',
'segments'
]}
idx = 1
cogid = 1
for i, line in enumerate(rawdata[1:]):
tmp = dict(zip(header, line))
concepts += [(tmp['number'], tmp['meaning'])]
for lid, lang in sorted(ds.lid2lang.items()):
cell = tmp[lid]
if cell.strip() != '-' and cell.strip():
ccell = crect.get(cell, cell)
syls = ' + '.join([' '.join(x) for x in syllabify(
ccell, output='nested')])
D[idx] = [
lang, tmp['meaning'], lid, tmp['number'], cell,
ccell, tmp['notes'], tmp['MM notes'],
cogid, syls]
idx += 1
cogid += 1
with open(ds.raw('concepts.tsv'), 'w') as f:
f.write('NUMBER\tENGLISH\n')
for a, b in concepts:
f.write(a+'\t'+b+'\n')
ds.write_wordlist(Wordlist(D))
'Burmese (Written)' : 'Written_Burmese',
'Burmese (Rangoon)' : 'Rangoon',
'Achang (Xiandao)' : 'Xiandao',
'Langsu (Luxi)' : 'Maru'
}
D = {}
idx = 1
for line in tbl[1:]:
ridx, reflex, gloss = line[:3]
if line[-1].strip():
tbl = str(int(line[-1].split('.')[0]))
lang = line[6].strip()
if lang in doculects and tbl in new_concepts:
doc = doculects[lang]
tokens = syllabify(ipa2tokens(reflex.replace(' ', '_'),
merge_vowels=False, expand_nasals=True,
semi_diacritics='shɕʑʃʒʐʂ'))
ipa = ''.join(tokens)
alm = []
for t in tokens:
alm += [sounds.get((t, doc), '_'+t if t in '()_' else t)]
alm = ' '.join(alm)
tokens = ' '.join(tokens)
concept = new_concepts[tbl][0]
if reflex.strip() == '*':
pass
else:
D[idx] = [ridx,
reflex,
gloss,
ipa,
for i, w in enumerate(wordlists):
wl = Wordlist(w)
colors = {}
tmp = defaultdict(int)
sylen = []
clen = []
for k in wl:
dolgos = tokens2class(wl[k, 'tokens'], 'dolgo')
for idx, t in zip(dolgos, wl[k, 'tokens']):
if idx not in '+':
tmp[idx] += 1
colors[idx] = token2class(t, color)
all_cols += [(k, colors[idx])]
all_sounds[idx] += 1
all_colors[idx] = colors[idx]
sylen += [len(syllabify(' '.join(wl[k, 'tokens']), output='nested'))]
clen += [len([x for x in dolgos if x not in '1V'])]
print(w, sum(sylen) / len(sylen), sum(clen) / len(clen))
ax = plt.subplot(gs[i])
labels = [x for x, y in sorted(tmp.items(), key=lambda x: x[0])]
ax.pie([y for x, y in sorted(tmp.items(), key=lambda x: x[0])],
colors=[y for x, y in sorted(colors.items(), key=lambda x: x[0])],
radius = 0.95, frame=True, shadow=True)
ax.set_autoscale_on(False)
plt.ylim(-1, 1)
plt.xlim(-1, 1)
plt.title(w.split('_')[2].split('-')[0])
plt.axis('off')
ax.set_aspect('equal')
print('plotting')
'Burmese (Written)' : 'Written_Burmese',
'Burmese (Rangoon)' : 'Rangoon',
'Achang (Xiandao)' : 'Xiandao',
'Langsu (Luxi)' : 'Maru'
}
D = {}
idx = 1
for line in tbl[1:]:
ridx, reflex, gloss = line[:3]
if line[-1].strip():
tbl = str(int(line[-1].split('.')[0]))
lang = line[6].strip()
if lang in doculects and tbl in new_concepts:
doc = doculects[lang]
tokens = syllabify(ipa2tokens(reflex.replace(' ', '_'),
merge_vowels=False, expand_nasals=True,
semi_diacritics='shɕʑʃʒʐʂ'))
ipa = ''.join(tokens)
alm = []
for t in tokens:
alm += [sounds.get((t, doc), '†'+t if t in '()_' else t)]
alm = ' '.join(alm)
tokens = ' '.join(tokens)
concept = new_concepts[tbl][0]
if reflex.strip() == '*':
pass
else:
D[idx] = [ridx,
reflex,
gloss,
ipa,
def cmd_makecldf(self, args):
with self.cldf_writer(args) as writer:
# load data as wordlists, as we need to bring the already segmented
# entries in line with clts
wl = lingpy.Wordlist(
self.raw_dir.joinpath("words.tsv").as_posix(),
conf=self.raw_dir.joinpath("wordlist.rc").as_posix(),
)
wl.add_entries(
"new_segments",
"segments",
lambda x: syllabify(self.tokenizer(
{}, "^" + "".join(x) + "$", column="IPA"),
cldf=True),
)
writer.add_sources()
# note: no way to easily replace this with the direct call to `add_concepts`
# as we add the Chinese gloss via concept.attributes
concept_lookup = {}
for concept in self.conceptlists[0].concepts.values():
idx = concept.id.split("-")[-1] + "_" + slug(concept.gloss)
writer.add_concept(
ID=idx,
Name=concept.gloss,
Chinese_Gloss=concept.attributes["chinese"],
Number=concept.number,
Concepticon_ID=concept.concepticon_id,
Concepticon_Gloss=concept.concepticon_gloss,
)
concept_lookup[concept.number] = idx
language_lookup = writer.add_languages(lookup_factory="Name")
for k in pylexibank.progressbar(wl,
desc="wl-to-cldf",
total=len(wl)):
if wl[k, "value"]:
form = self.form_spec.clean(form=wl[k, "value"], item=None)
writer.add_form_with_segments(
Language_ID=language_lookup[wl[k, "doculect"]],
Parameter_ID=concept_lookup[wl[k, "beida_id"]],
Value=wl[k, "value"],
Form=form,
Segments=wl[k, "new_segments"],
Source="Cihui",
Benzi=wl[k, "benzi"],
)
# We explicitly remove the ISO code column since the languages in
# this datasets do not have an ISO code.
writer.cldf["LanguageTable"].tableSchema.columns = [
col for col in writer.cldf["LanguageTable"].tableSchema.columns
if col.name != "ISO639P3code"
]
language_table = writer.cldf["LanguageTable"]
with self.cldf_writer(args, cldf_spec="structure",
clean=False) as writer:
writer.cldf.add_component(language_table)
writer.objects["LanguageTable"] = self.languages
inventories = self.raw_dir.read_csv("inventories.tsv",
normalize="NFC",
delimiter="\t",
dicts=True)
writer.cldf.add_columns(
"ParameterTable",
{
"name": "CLTS_BIPA",
"datatype": "string"
},
{
"name": "CLTS_Name",
"datatype": "string"
},
{
"name": "Lexibank_BIPA",
"datatype": "string"
},
{
"name": "Prosody",
"datatype": "string"
},
)
writer.cldf.add_columns("ValueTable", {
"name": "Context",
"datatype": "string"
})
clts = CLTS(args.clts.dir)
bipa = clts.transcriptionsystem_dict["bipa"]
td = clts.transcriptiondata_dict["beidasinitic"]
pids, visited = {}, set()
for row in pylexibank.progressbar(inventories, desc="inventories"):
if not row["Value"].startswith("(") and row["Value"] != "Ø":
for s1, s2, p in zip(row["Value"].split(),
row["Lexibank"].split(),
row["Prosody"].split()):
s1 = normalize("NFD", s1)
pidx = "-".join(
[str(hex(ord(s)))[2:].rjust(4, "0")
for s in s1]) + "_" + p
if not s1 in td.grapheme_map:
args.log.warn("missing sound {0} / {1}".format(
s1, " ".join([str(hex(ord(x))) for x in s1])))
else:
sound = bipa[td.grapheme_map[s1]]
sound_name = (sound.name if sound.type
not in ["unknown", "marker"] else "")
if not pidx in visited:
visited.add(pidx)
writer.objects["ParameterTable"].append({
"ID":
pidx,
"Name":
s1,
"Description":
sound_name,
"CLTS_BIPA":
td.grapheme_map[s1],
"CLTS_Name":
sound_name,
"Lexibank_BIPA":
s2,
"Prosody":
p,
})
writer.objects["ValueTable"].append({
"ID":
row["Language_ID"] + "_" + pidx,
"Language_ID":
row["Language_ID"],
"Parameter_ID":
pidx,
"Value":
s1,
"Context":
p,
"Source": ["Cihui"],
})