def __call__(self,
data,
positions=True,
chars=True,
keeporiginal=True,
removestops=True,
tokenize=True,
start_pos=0,
start_char=0,
mode="",
**kwargs):
if kwargs.get("docix", None) == self._docix and self._cache:
yield from self.cache
else:
t = CylleneusToken(positions,
chars,
removestops=removestops,
mode=mode,
**kwargs)
if t.mode == "query":
t.original = data
t.text = data.translate(jvmap)
yield t
else:
self._cache = []
self._docix = kwargs.get("docix", None)
if tokenize:
punc = str.maketrans("", "", string.punctuation)
tags = data["meta"].split("-")
meta = {"meta": data["meta"].lower()}
meta.update({tag: "-" for tag in tags})
divs = ["div1", "div2", "div3", "div4", "div5"]
sect_sent = 0
sect_pos = 0
sent_id = 0
pos = 0
for el in data["text"].find("text").find("body").iter():
if el.tag in divs:
current_div_ix = divs.index(el.tag)
meta[tags[current_div_ix]] = el.get("n", "-")
sect_sent = 0
sect_pos = 0
elif el.tag in ["head", "p", "l"]:
sent_id += 1
sect_sent += 1
if el.text:
text = el.text
else:
text = "".join([
subel.text + subel.tail
for subel in el.iter()
if subel.tag != el.tag
])
subs = [
r"<note>(.*?)</note>",
r'<sic corr="(\w+?)">\w+?</sic>',
r'<reg orig="\w+?">(\w+?)</reg>',
]
for sub in subs:
text = re.sub(sub, "\1", text)
tokens = word_tokenizer.word_tokenize(text)
for i, token in enumerate(tokens):
pos += 1
sect_pos += 1
t.text = (token.translate(
punc).lower().translate(jvmap))
if not t.text or t.text in string.whitespace:
start_char += 1
continue
t.boost = 1.0
meta["sent_id"] = sent_id
meta["sent_pos"] = i
meta["sect_sent"] = sect_sent
meta["sect_pos"] = sect_pos
t.meta = copy.copy(meta)
if keeporiginal:
t.original = token
t.stopped = False
if positions:
t.pos = start_pos + pos
is_enclitic = False
for enclitic in enclitics:
if token.endswith(enclitic):
if enclitic == "n":
t.text = (token[:-len(enclitic)] +
"s")
t.startchar = start_char
t.endchar = (start_char +
len(token) -
len(enclitic))
if mode == "index":
self._cache.append(
copy.deepcopy(t))
yield t
t.text = "ne"
t.startchar = start_char + len(
token[:-len(enclitic)])
t.endchar = (
start_char +
len(token[:-len(enclitic)]) +
len(enclitic))
if mode == "index":
self._cache.append(
copy.deepcopy(t))
yield t
elif enclitic == "ne":
t.text = token[:-len(enclitic)]
t.startchar = start_char
t.endchar = start_char + (
len(token) - len(enclitic))
if mode == "index":
self._cache.append(
copy.deepcopy(t))
yield t
t.text = "ne"
t.startchar = start_char + len(
token[:-len(enclitic)])
t.endchar = (
start_char +
len(token[:-len(enclitic)]) +
len(enclitic))
if mode == "index":
self._cache.append(
copy.deepcopy(t))
yield t
elif enclitic == "st":
if token.endswith("ust"):
t.text = token[:-len(enclitic)]
t.startchar = start_char
t.endchar = (start_char + len(
token[:-len(enclitic)]) -
len(enclitic))
if mode == "index":
self._cache.append(
copy.deepcopy(t))
yield t
t.text = "est"
t.startchar = start_char + len(
token[:-len(enclitic)])
t.endchar = (start_char + len(
token[:-len(enclitic)]) +
len(enclitic))
if mode == "index":
self._cache.append(
copy.deepcopy(t))
yield t
else:
t.text = token[:-len(enclitic)]
t.startchar = start_char
t.endchar = (start_char + len(
token[:-len(enclitic)]) -
len(enclitic))
if mode == "index":
self._cache.append(
copy.deepcopy(t))
yield t
t.text = "est"
t.startchar = start_char + len(
token[:-len(enclitic)])
t.endchar = (start_char + len(
token[:-len(enclitic)]) +
len(enclitic))
if mode == "index":
self._cache.append(
copy.deepcopy(t))
yield t
elif enclitic == "'s":
t.text = token + "s"
t.startchar = start_char
t.endchar = start_char + len(token)
if mode == "index":
self._cache.append(
copy.deepcopy(t))
yield t
t.text = "es"
t.startchar = (start_char +
len(token) + 1)
t.endchar = (start_char +
len(token) +
len(enclitic))
if mode == "index":
self._cache.append(
copy.deepcopy(t))
yield t
else:
t.text = token[:-len(enclitic)]
t.startchar = start_char
t.endchar = start_char + len(
token[:-len(enclitic)])
if mode == "index":
self._cache.append(
copy.deepcopy(t))
yield t
t.text = enclitic
t.startchar = start_char + len(
token[:-len(enclitic)])
t.endchar = (
start_char +
len(token[:-len(enclitic)]) +
len(enclitic))
if mode == "index":
self._cache.append(
copy.deepcopy(t))
yield t
is_enclitic = True
break
if not is_enclitic:
if chars:
t.startchar = start_char
original_len = len(token)
t.endchar = start_char + original_len
if self.cached:
self._cache.append(copy.copy(t))
yield t
start_char += len(token)
else:
t.original = ""
for el in data["text"].find("text").find("body").iter():
if el.tag in ["head", "p", "l"]:
if el.text:
text = el.text
else:
text = "".join([
subel.text + subel.tail
for subel in el.iter()
if subel.tag != el.tag
])
subs = [
r"<note>(.*?)</note>",
r'<sic corr="(\w+?)">\w+?</sic>',
r'<reg orig="\w+?">(\w+?)</reg>',
]
for sub in subs:
text = re.sub(sub, "\1", text)
t.original += text
t.text = t.original
t.boost = 1.0
if positions:
t.pos = start_pos
if chars:
t.startchar = start_char
t.endchar = start_char + len(t.original)
yield t
def __call__(
self,
data,
positions=True,
chars=True,
keeporiginal=True,
removestops=True,
tokenize=True,
start_pos=0,
start_char=0,
mode="",
**kwargs,
):
if kwargs.get("docix", None) == self._docix and self._cache:
yield from self.cache
else:
t = CylleneusToken(positions,
chars,
removestops=removestops,
mode=mode,
**kwargs)
if t.mode == "query":
t.original = t.text = data
yield t
else:
self._cache = []
self._docix = kwargs.get("docix", None)
if not tokenize:
t.original = ""
for token in data["text"].iter("token"):
form = token.get("form")
if not form:
continue
t.original += f"{form}"
t.text = t.original
t.boost = 1.0
if positions:
t.pos = start_pos
if chars:
t.startchar = start_char
t.endchar = start_char + len(t.original)
yield t
else:
for sentence in data["text"].iter("sentence"):
sect_pos = -1
curr_line = None
for pos, token in enumerate(sentence.iter("word")):
if token.get("artificial", False):
continue
form = token.get("form")
if not form:
continue
t.text = form
lemma = token.get("lemma")
if not lemma or lemma in (
"???",
".",
",",
";",
"·",
"punc1",
"comma1",
"PERIOD1",
):
continue
t.lemma = lemma
t.morpho = agldt2wn(token.get("postag"))
t.morphosyntax = token.get("relation", None)
t.boost = 1.0
meta = {"meta": data["meta"].lower()}
divs = data["meta"].split("-")
refs = (token.get("cite").rsplit(
":", maxsplit=1)[1].split("."))
for i, div in enumerate(divs):
meta[div] = refs[i]
meta["sent_id"] = sentence.get("id")
meta["sent_pos"] = str(int(token.get("id")))
if curr_line and refs[-1] > curr_line:
sect_pos = 0
else:
sect_pos += 1
curr_line = refs[-1]
meta["sect_pos"] = sect_pos # ref in line
t.meta = meta
if keeporiginal:
t.original = f"{form}"
t.stopped = False
if positions:
t.pos = start_pos + pos
if chars:
t.startchar = start_char
original_len = len(form)
t.endchar = start_char + original_len
if self.cached:
self._cache.append(copy.copy(t))
yield t
start_char += len(form)
def __call__(
self,
value,
positions=True,
chars=True,
keeporiginal=True,
removestops=True,
tokenize=True,
start_pos=0,
start_char=0,
mode="",
**kwargs,
):
if kwargs.get("docix") == self._docix and self._cache:
yield from self.cache
else:
t = CylleneusToken(positions,
chars,
removestops=removestops,
mode=mode,
**kwargs)
if t.mode == "query":
t.original = t.text = value.translate(jvmap)
yield t
else:
if not tokenize:
t.original = t.text = value["text"]
t.boost = 1.0
if positions:
t.pos = start_pos
if chars:
t.startchar = start_char
t.endchar = start_char + len(value["text"])
yield t
else:
self._cache = []
self._docix = kwargs.get("docix", None)
word_tokenizer = PunktLatinCharsVars()
stopchars = str.maketrans(
"",
"",
string.punctuation.replace("&", "").replace("^", "") +
"†“”—\n\ŕ",
)
divs = {
i: div.lower()
for i, div in enumerate(value["meta"].split("-"))
}
lines = iter(value["text"].split("\n"))
tpos = start_pos
xtitle = ytitle = ztitle = speaker = ""
buffer = deque()
for line in lines:
def parse_phi_line(_line):
result = []
nonlocal xtitle, ytitle, ztitle, speaker, buffer
try:
ref, text = _line.split("\t")
except ValueError:
result.append((None, None))
else:
v, w, x, y, z = ref.rstrip(".").split(".")
offset = 0
# d is a number, followed by -, t, a then possibly another number or . for a title
# d can be 'opinc' 'sedinc' 'dub', 'inc',
# c can be 'Summ'
if x == "t":
xtitle = text.translate(stopchars).strip()
if y == "t":
if z:
ytitle = text.translate(
stopchars).strip()
else:
speaker = text.translate(
stopchars).strip()
result.append((None, [text]))
elif z == "t":
ztitle = text.translate(stopchars).strip()
result.append((None, [text]))
elif " {" in text:
result.append((None, [text]))
else:
temp_tokens = word_tokenizer.word_tokenize(
text)
if temp_tokens:
if (temp_tokens[0].replace(
"j", "i").replace("v",
"u") not in
proper_names.proper_names):
temp_tokens[0] = temp_tokens[
0].lower()
if (temp_tokens[-1].endswith(".") and
temp_tokens[-1] != ". . ."):
final_word = temp_tokens[-1][:-1]
del temp_tokens[-1]
temp_tokens += [final_word, "."]
if temp_tokens[-1].endswith("-"):
buffer += list(
parse_phi_line(next(lines)))
new_ref, new_tokens = buffer.pop()
merged_word = (
"2&" + temp_tokens[-1][:-1] +
new_tokens[0])
del temp_tokens[-1]
temp_tokens += [merged_word]
del new_tokens[0]
if new_tokens:
if (new_tokens[0]
in string.punctuation):
new_token = (
f"^1{new_tokens[0]}")
del new_tokens[0]
new_tokens.insert(
0, new_token)
buffer.appendleft(
(new_ref, new_tokens))
for ix, token in enumerate(
temp_tokens):
if temp_tokens[ix] == ". . .":
temp_tokens.insert(
ix + 1, "&1")
if "&" in token:
ppp = compound.is_ppp(
re.sub(
r"[&\d]", "", token))
else:
ppp = compound.is_ppp(token)
if ppp:
if ix == len(temp_tokens) - 1:
if not buffer:
try:
buffer += list(
parse_phi_line(
next(lines)
))
except StopIteration:
continue
if "&" in buffer[0][1][0]:
copula = compound.is_copula(
buffer[0][1][0]
[2:])
else:
copula = compound.is_copula(
buffer[0][1][0])
else:
copula = compound.is_copula(
temp_tokens[ix + 1])
if (copula and ppp[1]
== copula[2]):
(
tense,
mood,
number,
i,
) = copula
if buffer:
token = f"{token} &2{compound.copula[tense][mood][number][i]}"
else:
token = f"{token} {compound.copula[tense][mood][number][i]}"
del temp_tokens[ix]
if buffer:
del buffer[0][1][0]
else:
del temp_tokens[ix]
temp_tokens.insert(
ix, token)
if (ix !=
len(temp_tokens) -
1):
if (temp_tokens[ix + 1]
in string.
punctuation):
new_token = f"^1{temp_tokens[ix + 1]} "
del temp_tokens[ix
+
1]
temp_tokens.insert(
ix + 1,
new_token,
)
if buffer:
for i in range(len(buffer)):
result.append(buffer.pop())
result.append(
((v, w, x, y, z), temp_tokens))
yield from result
result = list(parse_phi_line(line))
act = scene = None
for ref, tokens in reversed(result):
enjambed = False
if not ref and not tokens:
start_char += len(line) + 1
continue
elif not ref:
text = tokens[0].strip().strip("{}")
if re.match(
r"[IVXLDMivxldm]+\.[IVXLDMivxldm]+",
text):
act, scene = text.split(".")
act = str(roman_to_arabic(act))
scene = str(roman_to_arabic(scene))
start_char += len(line.split("\t")[1]) + 1
continue
notoken = 0
skip = False
for line_pos, token in enumerate(tokens):
if token == "{" or token == "}":
skip = not skip
start_char += len(token)
continue
if skip:
speaker = token.replace("v", "u")
start_char += len(token)
continue
offset = 0
line_pos -= notoken
meta = {}
# extra['meta'] = value['meta'].lower()
# setattr(t, 'meta', value['meta'].lower())
for i in range(len(divs)):
meta[divs[len(divs) - (i + 1)]] = ref[-(
5 - (5 - (i + 1)))].strip("t")
# setattr(t, divs[len(divs) - (i + 1)], ref[-(5 - (5 - (i + 1)))].strip('t'))
if xtitle:
if len(divs) >= 3:
meta[
f"{divs[len(divs) - 3]}_title"] = xtitle
# setattr(t, f"{divs[len(divs)-3]}_title", xtitle)
if ytitle:
if len(divs) >= 2:
meta[
f"{divs[len(divs) - 2]}_title"] = ytitle
# setattr(t, f"{divs[len(divs)-2]}_title", ytitle)
if ztitle:
if len(divs) >= 1:
meta[
f"{divs[len(divs) - 1]}_title"] = ztitle
# setattr(t, f"{divs[len(divs)-1]}_title", ztitle)
if act:
meta["act"] = act
if scene:
meta["scene"] = scene
# if speaker:
# t.speaker = speaker
t.boost = 1.0
pre = re.search(r"^\^(\d+?)", token)
if pre:
start_char -= int(pre.group(1))
token = re.sub(r"^\^\d+?", "", token)
pre = re.search(r"^&(\d+?)", token)
if pre:
start_char += int(pre.group(1))
token = re.sub(r"^&\d+?", "", token)
if keeporiginal:
t.original = token
t.stopped = False
original_length = len(token)
ltoken = token.lstrip(string.punctuation)
ldiff = original_length - len(ltoken)
if ldiff != 0:
token = ltoken
rtoken = token.rstrip(string.punctuation)
rdiff = len(token) - len(rtoken)
if rdiff != 0:
token = rtoken
ntoken = token.translate(stopchars)
ndiff = len(token) - len(ntoken)
if ndiff:
token = ntoken
if not re.match(
r"(?:[\d]&)?[\w]+\s(?:&[\d])?[\w]+",
token):
token = token.replace(" ", "")
if not token:
start_char += original_length
notoken += 1
continue
else:
if positions:
meta["line_pos"] = line_pos
t.pos = tpos
t.meta = meta
if (token not in exceptions
and token.lower() not in exceptions
and re.sub(r"\d&|&\d", "", token)
not in exceptions):
if token in replacements: # t.original
for subtoken in replacements[
token]:
t.text = subtoken.lower()
t.startchar = start_char
t.endchar = (start_char +
original_length)
if mode == "index":
if self.cached:
self._cache.append(
copy.copy(t))
yield t
start_char += original_length
tpos += 1
continue
if re.match(
r"(?:[\d]&)?[\w]+\s(?:&[\d])?[\w]+",
token,
):
ppp, copula = token.split(" ")
post = re.match(
r"([\d])&[\w]+", ppp)
if post:
offset += int(post.group(1))
ppp = re.sub(r"[\d]&", "", ppp)
original_length -= 2
enjambed = True
t.text = ppp.lower()
t.startchar = start_char
t.endchar = (start_char +
len(ppp) + offset)
if mode == "index":
if self.cached:
self._cache.append(
copy.copy(t))
yield t
pre = re.search(r"&(\d+?)", copula)
if pre:
start_char += int(pre.group(1))
copula = re.sub(
r"&\d+?", "", copula)
original_length -= 2
enjambed = True
t.text = copula.lower()
t.startchar = (start_char +
len(ppp) + 1)
t.endchar = (start_char +
len(ppp) + 1 +
len(copula))
if mode == "index":
if self.cached:
self._cache.append(
copy.copy(t))
yield t
start_char += original_length
tpos += 1
continue
else:
post = re.match(
r"([\d])&[\w]+", token)
if post:
offset += int(post.group(1))
token = re.sub(
r"[\d]&", "", token)
original_length -= 2
enjambed = True
else:
offset = 0
is_enclitic = False
for enclitic in enclitics:
if token.lower().endswith(
enclitic):
is_enclitic = True
if enclitic == "ne":
t.text = (
token[:-len(enclitic)]
).lower()
t.startchar = start_char
t.endchar = start_char + (
len(token) -
len(enclitic))
if mode == "index":
if self.cached:
self._cache.append(
copy.copy(t))
yield t
t.text = "ne"
t.startchar = (
start_char +
len(token[:-len(
enclitic)]) +
offset)
t.endchar = (
start_char +
len(token[:-len(
enclitic)]) +
len(enclitic) + offset)
if mode == "index":
if self.cached:
self._cache.append(
copy.copy(t))
yield t
elif enclitic == "n":
t.text = (
token[:-len(enclitic)]
+ "s").lower()
t.startchar = start_char
t.endchar = (
start_char +
(len(token) + 1) -
len(enclitic))
if mode == "index":
if self.cached:
self._cache.append(
copy.copy(t))
yield t
t.text = "ne"
t.startchar = (
start_char +
len(token[:-len(
enclitic)]) +
offset)
t.endchar = (
start_char +
len(token[:-len(
enclitic)]) +
len(enclitic) + offset)
if mode == "index":
if self.cached:
self._cache.append(
copy.copy(t))
yield t
elif enclitic == "st":
if token.endswith("ust"):
t.text = (
token[:-len(
enclitic) +
1]).lower()
t.startchar = (
start_char)
t.endchar = (
start_char +
len(token[:-len(
enclitic) + 1])
- len(enclitic))
if mode == "index":
if self.cached:
self._cache.append(
copy.copy(
t))
yield t
t.text = "est"
t.startchar = (
start_char +
len(token[:-len(
enclitic) + 1])
+ offset)
t.endchar = (
start_char +
len(token[:-len(
enclitic) + 1])
+ len(enclitic) +
offset)
if mode == "index":
if self.cached:
self._cache.append(
copy.copy(
t))
yield t
else:
t.text = (token[:-len(
enclitic)]
).lower()
t.startchar = (
start_char)
t.endchar = (
start_char +
len(token[:-len(
enclitic)]) -
len(enclitic))
if mode == "index":
if self.cached:
self._cache.append(
copy.copy(
t))
yield t
t.text = "est"
t.startchar = (
start_char +
len(token[:-len(
enclitic)]) +
offset)
t.endchar = (
start_char +
len(token[:-len(
enclitic)]) +
len(enclitic) +
offset)
if mode == "index":
if self.cached:
self._cache.append(
copy.copy(
t))
yield t
elif enclitic == "'s":
t.text = (token.lower() +
"s")
t.startchar = start_char
t.endchar = (start_char +
len(token))
if mode == "index":
if self.cached:
self._cache.append(
copy.copy(t))
yield t
t.text = "es"
t.startchar = (start_char +
len(token) +
1)
t.endchar = (start_char +
len(token) +
len(enclitic))
if mode == "index":
if self.cached:
self._cache.append(
copy.copy(t))
yield t
else:
t.text = (
token[:-len(enclitic)]
).lower()
t.startchar = start_char
t.endchar = (
start_char +
len(token[:-len(
enclitic)]))
if mode == "index":
if self.cached:
self._cache.append(
copy.copy(t))
yield t
t.text = enclitic
t.startchar = (
start_char +
len(token[:-len(
enclitic)]) +
offset)
t.endchar = (
start_char +
len(token[:-len(
enclitic)]) +
len(enclitic) + offset)
if mode == "index":
if self.cached:
self._cache.append(
copy.copy(t))
yield t
break
else:
is_enclitic = False
post = re.match(r"([\d])&[\w]+", token)
if post:
offset += int(post.group(1))
token = re.sub(r"[\d]&", "", token)
original_length -= 2
enjambed = True
if not is_enclitic:
t.text = token
if chars:
t.startchar = start_char + ldiff
t.endchar = (start_char +
original_length -
rdiff + offset)
if mode == "index":
if self.cached:
self._cache.append(
copy.copy(t))
yield t
tpos += 1
if enjambed:
start_char += original_length + offset
else:
start_char += original_length
start_char += 1 # \n
def __call__(
self,
data,
positions=True,
chars=True,
keeporiginal=True,
removestops=True,
tokenize=True,
start_pos=0,
start_char=0,
mode="",
**kwargs,
):
if kwargs.get("docix", None) == self._docix and self._cache:
yield from self.cache
else:
t = CylleneusToken(positions,
chars,
removestops=removestops,
mode=mode,
**kwargs)
if t.mode == "query":
t.original = data
t.text = normalize("NFKC", data)
yield t
else:
self._cache = []
self._docix = kwargs.get("docix", None)
if tokenize:
titleStmt = data.find('.//teiHeader').find(
'fileDesc').find('titleStmt')
auth_code = f"tlg{titleStmt.find('tlgAuthor').text}"
work_code = f"tlg{titleStmt.find('tlgId').text}"
body = data.find('.//text').find('body')
divs = AUTHOR_TAB[auth_code]["works"][work_code]["meta"]
meta = {"meta": divs}
divv = divs.split("-")
for k in divv:
meta[k] = None
sect_sent = 0
sect_pos = 0
current_refs = None
pos = 0
for sentence in body.iter("sentence"):
refs = sentence.get("location")
if refs != current_refs:
current_refs = refs
sect_pos = 0
sect_sent = 0
sent_id = sentence.get("id")
sect_sent += 1
for i, ref in enumerate(refs.split(".")):
meta[divv[i]] = ref
for sent_pos, word in enumerate(sentence.iter("word")):
t.boost = 1.0
sect_pos += 1
pos += 1
lemma = word.find("lemma").get("entry", None)
t.lemma = normalize("NFKC", lemma)
meta["sent_id"] = sent_id
meta["sent_pos"] = word.get("id")
meta["sect_pos"] = str(sect_pos)
meta["sect_sent"] = str(sect_sent)
t.meta = copy.copy(meta)
beta = word.get("form").upper()
form = normalize(
"NFKC",
beta2unicode(
beta +
"\n" if beta.endswith("S") else beta))
if (t.lemma.istitle()):
form = form.title()
t.text = form
if keeporiginal:
t.original = beta
t.stopped = False
if positions:
t.pos = start_pos + pos
original_len = len(form)
if chars:
t.startchar = start_char
t.endchar = start_char + original_len
start_char += original_len
POS = word.find("lemma").get("POS", None)
analyses = [
analysis.get("morph", None) for analysis in
word.find("lemma").iter("analysis")
]
morphos = []
for analysis in analyses:
morphos += diorisis2wn(POS, analysis)
t.morpho = " ".join(morphos)
if self.cached:
self._cache.append(copy.deepcopy(t))
yield t
else:
body = data.find('.//text').find('body')
tokens = []
for sentence in body.iter("sentence"):
for word in sentence.iter("word"):
form = word.get("form")
if not form:
continue
else:
tokens.append(form)
t.original = t.text = " ".join(tokens)
t.boost = 1.0
if positions:
t.pos = start_pos
if chars:
t.startchar = start_char
t.endchar = start_char + len(t.original)
yield t
def __call__(
self,
data,
positions=True,
chars=True,
keeporiginal=True,
removestops=True,
tokenize=True,
start_pos=0,
start_char=0,
mode="",
**kwargs,
):
if kwargs.get("docix", None) == self._docix and self._cache:
yield from self.cache
else:
t = CylleneusToken(
positions, chars, removestops=removestops, mode=mode, **kwargs
)
if t.mode == "query":
t.original = data
t.text = data.translate(jvmap)
yield t
else:
self._cache = []
self._docix = kwargs.get("docix", None)
if not tokenize:
t.original = ""
for el in data["text"].find("text").find("body").iter():
if el.tag in ["head", "p", "l"]:
if not el.text:
text = "".join(
[
subel.text + subel.tail
for subel in el.iter()
if subel.tag != el.tag
]
)
else:
text = el.text
subs = [
r"<note>(.*?)</note>",
r'<sic corr="(\w+?)">\w+?</sic>',
r'<reg orig="\w+?">(\w+?)</reg>',
]
for sub in subs:
text = re.sub(sub, "\1", text)
t.original += text
t.text = t.original
t.boost = 1.0
if positions:
t.pos = start_pos
if chars:
t.startchar = start_char
t.endchar = start_char + len(t.original)
yield t
else:
punc = str.maketrans("", "", string.punctuation)
tags = data["meta"].lower()
meta = {"meta": tags}
if tags != "-":
divs = data["meta"].split("-")
meta.update({div: "-" for div in divs})
sect_sent = 0
sect_pos = 0
sent_id = 0
pos = 0
for el in (
data["text"]
.find("{http://www.tei-c.org/ns/1.0}text")
.find("{http://www.tei-c.org/ns/1.0}body")
.findall(".//{http://www.tei-c.org/ns/1.0}*")
):
if el.tag == "{http://www.tei-c.org/ns/1.0}milestone":
meta[el.get("unit")] = el.get("n", "-")
sect_sent = 0
sect_pos = 0
elif (
el.tag == "{http://www.tei-c.org/ns/1.0}div"
and el.get("n")
):
meta[el.get("type")] = el.get("n", "-")
sect_sent = 0
sect_pos = 0
if not el.text:
text = el.tail if el.tail else ""
else:
text = el.text + (el.tail if el.tail else "")
subs = [
(r"<supplied>(.*?)</supplied>", "\1"),
(r'<quote type="\w+?">(.+?)</quote>', "\1"),
(r'<hi rend="\w+?">(.+?)</hi>', "\1"),
(r'<g ref="\w+?">(.+?)</g>', "\1"),
(
r'<foreign xml:lang="\w+?">(\w+?)</foreign>',
"\1",
),
(r"<del>.+?</del>", ""),
]
for old, new in subs:
text = re.sub(old, new, text)
if text:
for sentence in sent_tokenizer.tokenize(text):
sent_id += 1
sect_sent += 1
sentence = sentence.strip()
replacements = [(r"\n", ""), (r"\s+", " ")]
for old, new in replacements:
sentence = re.sub(old, new, sentence)
sent_pos = 0
tokens = word_tokenizer.word_tokenize(sentence)
for token in tokens:
token = (
token.translate(punc)
.lower()
.translate(jvmap)
.strip()
)
if not token or token in string.whitespace:
start_char += 1
continue
else:
pos += 1
sect_pos += 1
sent_pos += 1
t.text = token
t.boost = 1.0
meta["sent_id"] = sent_id
meta["sent_pos"] = sent_pos
meta["sect_sent"] = sect_sent
meta["sect_pos"] = sect_pos
t.meta = copy.copy(meta)
if keeporiginal:
t.original = token
t.stopped = False
if positions:
t.pos = start_pos + pos
is_enclitic = False
for enclitic in enclitics:
if (
token.endswith(enclitic)
and token
not in latin_exceptions
):
if enclitic == "ne":
t.text = token[
: -len(enclitic)
]
t.startchar = start_char
t.endchar = start_char + (
len(token)
- len(enclitic)
)
if mode == "index":
self._cache.append(
copy.deepcopy(t)
)
yield t
t.text = "ne"
t.startchar = (
start_char
+ len(
token[
: -len(
enclitic
)
]
)
)
t.endchar = (
start_char
+ len(
token[
: -len(
enclitic
)
]
)
+ len(enclitic)
)
if mode == "index":
self._cache.append(
copy.deepcopy(t)
)
yield t
elif enclitic == "n":
t.text = (
token[: -len(enclitic)]
+ "s"
)
t.startchar = start_char
t.endchar = (
start_char
+ len(token)
- len(enclitic)
)
if mode == "index":
self._cache.append(
copy.deepcopy(t)
)
yield t
t.text = "ne"
t.startchar = (
start_char
+ len(
token[
: -len(
enclitic
)
]
)
)
t.endchar = (
start_char
+ len(
token[
: -len(
enclitic
)
]
)
+ len(enclitic)
)
if mode == "index":
self._cache.append(
copy.deepcopy(t)
)
yield t
elif enclitic == "st":
if token.endswith("ust"):
t.text = token[
: -len(enclitic)
]
t.startchar = (
start_char
)
t.endchar = (
start_char
+ len(
token[
: -len(
enclitic
)
]
)
- len(enclitic)
)
if mode == "index":
self._cache.append(
copy.deepcopy(
t
)
)
yield t
t.text = "est"
t.startchar = (
start_char
+ len(
token[
: -len(
enclitic
)
]
)
)
t.endchar = (
start_char
+ len(
token[
: -len(
enclitic
)
]
)
+ len(enclitic)
)
if mode == "index":
self._cache.append(
copy.deepcopy(
t
)
)
yield t
else:
t.text = token[
: -len(enclitic)
]
t.startchar = (
start_char
)
t.endchar = (
start_char
+ len(
token[
: -len(
enclitic
)
]
)
- len(enclitic)
)
if mode == "index":
self._cache.append(
copy.deepcopy(
t
)
)
yield t
t.text = "est"
t.startchar = (
start_char
+ len(
token[
: -len(
enclitic
)
]
)
)
t.endchar = (
start_char
+ len(
token[
: -len(
enclitic
)
]
)
+ len(enclitic)
)
if mode == "index":
self._cache.append(
copy.deepcopy(
t
)
)
yield t
elif enclitic == "'s":
t.text = token + "s"
t.startchar = start_char
t.endchar = (
start_char + len(token)
)
if mode == "index":
self._cache.append(
copy.deepcopy(t)
)
yield t
t.text = "es"
t.startchar = (
start_char
+ len(token)
+ 1
)
t.endchar = (
start_char
+ len(token)
+ len(enclitic)
)
if mode == "index":
self._cache.append(
copy.deepcopy(t)
)
yield t
else:
t.text = token[
: -len(enclitic)
]
t.startchar = start_char
t.endchar = (
start_char
+ len(
token[
: -len(
enclitic
)
]
)
)
if mode == "index":
self._cache.append(
copy.deepcopy(t)
)
yield t
t.text = enclitic
t.startchar = (
start_char
+ len(
token[
: -len(
enclitic
)
]
)
)
t.endchar = (
start_char
+ len(
token[
: -len(
enclitic
)
]
)
+ len(enclitic)
)
if mode == "index":
self._cache.append(
copy.deepcopy(t)
)
yield t
is_enclitic = True
break
if not is_enclitic:
original_len = len(token)
if chars:
t.startchar = start_char
t.endchar = (
start_char + original_len
)
if self.cached:
self._cache.append(
copy.copy(t)
)
yield t
start_char += len(token)
def __call__(self,
value,
positions=True,
chars=True,
keeporiginal=True,
removestops=True,
tokenize=True,
start_pos=0,
start_char=0,
mode="",
**kwargs):
if (kwargs.get("docix", None) == self._docix
and self._cache is not None):
yield from self.cache
else:
t = CylleneusToken(positions,
chars,
removestops=removestops,
mode=mode,
**kwargs)
if t.mode == "query":
t.original = t.text = value
yield t
else:
if not tokenize:
text = "\n".join([el for el in flatten(value["text"])])
t.original = t.text = text
t.boost = 1.0
if positions:
t.pos = start_pos
if chars:
t.startchar = start_char
t.endchar = start_char + len(text)
yield t
else:
self._cache = []
self._docix = kwargs.get("docix", None)
tokenizer = word_tokenize
stopchars = string.punctuation
doc = value["text"]
divs = [
cref.get("n") for cref in reversed(
doc.findall(
".//{http://www.tei-c.org/ns/1.0}cRefPattern"))
]
ss = doc.iter("{http://www.tei-c.org/ns/1.0}s")
sect_sent = 0
sect_pos = 0
for n, s in enumerate(ss):
meta = {
"meta":
"-".join(divs),
"sent_id":
s.get("{http://www.w3.org/XML/1998/namespace}id"),
"sect_sent":
sect_sent,
"alignment":
s.get("n"),
}
el = s
j = 0
while el is not None:
if el.getparent() is not None:
if (el.getparent().get("type",
None) == "textpart"):
j -= 1
if (divs[j] in meta
and el.getparent().get("n") !=
meta[divs[j]]):
sect_sent = 0
sect_pos = 0
meta[divs[j]] = el.getparent().get("n")
el = el.getparent()
text = stringify(s)
sent_pos = 0
for i, token in enumerate(tokenizer(text)):
if token == " " or token in stopchars:
sect_pos += 1
sent_pos += 1
continue
t.boost = 1.0
if keeporiginal:
t.original = token
t.stopped = False
meta["sent_pos"] = sent_pos
meta["sect_pos"] = sect_pos
if positions:
t.pos = start_pos + sect_pos
length = len(token)
token = token.strip()
if not token:
start_char += length
continue
t.meta = copy.deepcopy(meta)
t.text = token
if chars:
t.startchar = start_char
t.endchar = start_char + length
if mode == "index":
self._cache.append(copy.deepcopy(t))
yield t
start_char += length
sect_pos += 1
sent_pos += 1
sect_sent += 1
start_char += 1
def __call__(
self,
data,
positions=True,
chars=True,
keeporiginal=True,
removestops=True,
tokenize=True,
start_pos=0,
start_char=0,
mode="",
**kwargs,
):
if kwargs.get("docix", None) == self._docix and self._cache:
yield from self.cache
else:
t = CylleneusToken(
positions, chars, removestops=removestops, mode=mode, **kwargs
)
if t.mode == "query":
t.original = data
t.text = data.translate(jvmap)
yield t
else:
if not tokenize:
t.original = ""
for token in data.iter("token"):
form = token.get("form")
if not form:
continue
after = token.get("presentation-after", "")
before = token.get("presentation-before", "")
t.original += f"{before}{form}{after}"
t.text = t.original
t.boost = 1.0
if positions:
t.pos = start_pos
if chars:
t.startchar = start_char
t.endchar = start_char + len(t.original)
yield t
else:
self._cache = []
self._docix = kwargs.get("docix", None)
for sentence in data["text"].iter("sentence"):
for pos, token in enumerate(sentence.iter("token")):
form = token.get("form")
if not form:
continue
else:
form = form.replace(" ", " ").replace(" ", " ")
form = re.sub(r"\.([^ ]|^$)", r". \1", form)
t.lemma = token.get("lemma")
t.morpho = proiel2wn(
token.get("part-of-speech"),
token.get("morphology"),
)
t.morphosyntax = token.get("relation", None)
t.boost = 1.0
meta = {"meta": data["meta"].lower()}
for i, div in enumerate(data["meta"].split("-")):
meta[div] = token.get("citation-part").split(
"."
)[i]
meta["sent_id"] = sentence.get("id")
meta["sent_pos"] = token.get("id")
t.meta = meta
before = token.get("presentation-before", "")
after = token.get("presentation-after", "")
if keeporiginal:
t.original = f"{before}{form}{after}"
t.stopped = False
if positions:
t.pos = start_pos + pos
original_len = len(form)
if (
form.istitle()
and pos == 0
and not t.lemma.istitle()
):
form = form.lower()
t.text = form
if chars:
t.startchar = start_char + len(before)
t.endchar = (
start_char + len(before) + original_len
)
self._cache.append(copy.deepcopy(t))
yield t
if form in editorial:
t.text = editorial[form]
self._cache.append(copy.deepcopy(t))
yield t
start_char += len(before) + len(form) + len(after)
def __call__(
self,
data,
positions=True,
chars=True,
keeporiginal=True,
removestops=True,
tokenize=True,
start_pos=0,
start_char=0,
mode="",
**kwargs,
):
if kwargs.get("docix", None) == self._docix and self._cache:
yield from self.cache
else:
t = CylleneusToken(positions,
chars,
removestops=removestops,
mode=mode,
**kwargs)
if t.mode == "query":
t.original = data
t.text = data.translate(jvmap)
yield t
else:
self._cache = []
self._docix = kwargs.get("docix", None)
if tokenize:
for sentence in data["text"].iter("sentence"):
for pos, token in enumerate(sentence.iter("word")):
if token.get("artificial", False):
continue
form = token.get("form")
if form:
form = form.replace(" ", " ").replace(" ", " ")
form = re.sub(r"\.([^ ]|^$)", r". \1", form)
else:
continue
lemma = token.get("lemma", None)
if not lemma or lemma in (
".",
",",
"punc1",
"comma1",
"PERIOD1",
):
continue
t.lemma = lemma.strip("0123456789")
t.morpho = agldt2wn(token.get("postag"))
t.morphosyntax = token.get("relation", None)
t.boost = 1.0
meta = {"meta": data["meta"].lower()}
divs = data["meta"].split("-")
for i, div in enumerate(divs):
if len(divs) <= 2 or div != "line":
meta[div] = sentence.get("subdoc").split(
".")[i]
meta["sent_id"] = sentence.get("id")
meta["sent_pos"] = token.get("id")
t.meta = meta
if keeporiginal:
t.original = f"{form}"
t.stopped = False
if positions:
t.pos = copy.copy(start_pos + pos)
original_len = len(form)
if (form.istitle() and pos == 0
and not t.lemma.istitle()):
form = form.lower()
t.text = form
if chars:
t.startchar = copy.copy(start_char)
t.endchar = copy.copy(start_char +
original_len)
if self.cached:
self._cache.append(copy.deepcopy(t))
yield t
if form in editorial:
t.text = editorial[form]
if self.cached:
self._cache.append(copy.copy(t))
yield t
start_char += len(form)
else:
t.original = ""
for token in data.iter("token"):
form = token.get("form")
if not form:
continue
t.original += f"{form}"
t.text = t.original
t.boost = 1.0
if positions:
t.pos = start_pos
if chars:
t.startchar = start_char
t.endchar = start_char + len(t.original)
yield t
def __call__(self,
value: dict,
positions=False,
chars=False,
keeporiginal=True,
removestops=True,
tokenize=True,
start_pos=0,
start_char=0,
mode="",
**kwargs):
if kwargs.get("docix", None) == self._docix and self._cache:
yield from self.cache
else:
t = CylleneusToken(positions,
chars,
removestops=removestops,
mode=mode,
**kwargs)
if t.mode == "query":
t.original = value
t.text = value.translate(jvmap)
yield t
else:
if not tokenize:
t.original = t.text = "\n".join(
[el for el in value["text"]])
t.boost = 1.0
if positions:
t.pos = start_pos
if chars:
t.startchar = start_char
t.endchar = start_char + len(value["text"])
yield t
else:
self._cache = []
self._docix = kwargs.get("docix", None)
punctuation = str.maketrans("", "", string.punctuation)
editorial = str.maketrans("", "", "[{(<>)}]")
added = re.compile(r"(\s?[<(][\w .]+[>)]\s?)")
t.boost = 1.0
t.pos = t.startchar = t.endchar = 0
sect_sent = 0 # sentence count within passage
sent_id = "0001"
sect_pos = 0 # word pos within passage
sent_pos = 0 # word pos within sentence
current_refs = tuple(["0"] * len(value["meta"]))
nflag = None
for pos, line in enumerate(value["text"]):
t.pos = pos
parsed = parse_bpn(line)
if not parsed:
continue
if int(parsed["sent_id"]) > int(sent_id):
sent_pos = 0
sent_id = parsed["sent_id"]
if (tuple([
alnum(i) for i in parsed["refs"].split(",")
]) > current_refs):
sect_sent = 1
sect_pos = 0
else:
sect_sent += 1
if keeporiginal:
if added.search(parsed["form"]):
t.original = added.sub("", parsed["form"])
else:
t.original = parsed["form"]
t.stopped = False
if parsed["form_code"] in "&+":
if parsed["lemma"] != "#":
if parsed["lemma"] == "_SVM":
t.morpho = None
t.lemma = parsed["lemma"]
t.lemma_n = parsed["lemma_n"]
t.original = added.sub("", parsed["form"])
t.text = parsed["form"].translate(
editorial)
else:
form = parsed["form"]
t.morpho = parsed["morpho"]
if " " in form:
t.original = added.sub("", form)
text = form.translate(editorial)
else:
t.original = form
text = form
t.lemma = parsed["lemma"]
t.lemma_n = parsed["lemma_n"]
if added.search(parsed["form"]):
t.original = added.sub(
"", parsed["form"])
t.text = text.translate(editorial)
nflag = False
else:
# could be a Greek form, do we index it?
t.morpho = ""
t.lemma = ""
t.lemma_n = ""
t.original = added.sub("", parsed["form"])
t.text = parsed["form"].translate(editorial)
elif parsed["form_code"] == "@": # combined forms
if parsed["lemma"] != "#":
t.lemma = parsed["lemma"]
t.lemma_n = parsed["lemma_n"]
t.text = parsed["form"].translate(editorial)
t.morpho = parsed["morpho"]
if nflag:
sect_pos -= 1
sent_pos -= 1
else:
nflag = True
else:
sent_pos += 1
sect_pos += 1
continue
elif parsed["form_code"] == "=": # que
t.text = parsed["form"].translate(editorial)
t.lemma = parsed["lemma"]
t.lemma_n = parsed["lemma_n"]
t.morpho = parsed["morpho"]
sent_pos -= 1
sect_pos -= 1
nflag = False
meta = {"meta": value["meta"].lower()}
tags = value["meta"].split("-")
divs = {i: div.lower() for i, div in enumerate(tags)}
refs = tuple([
ref.translate(punctuation)
for ref in parsed["refs"].strip().split(",")
])
for i in range(len(divs)):
meta[divs[i]] = refs[i]
current_refs = refs
t.morphosyntax = parsed["subord"]
meta["sect_sent"] = str(sect_sent)
meta["sect_pos"] = str(sect_pos)
meta["sent_id"] = parsed["sent_id"]
meta["sent_pos"] = str(sent_pos)
t.meta = meta
t.startchar = start_char
t.endchar = start_char + len(t.original)
if t.text != t.original:
tc = copy.deepcopy(t)
tc.text = t.original
yield tc
yield t
sent_pos += 1
sect_pos += 1
start_char += len(t.original) + 1
def __call__(
self,
value,
positions=True,
chars=True,
keeporiginal=True,
removestops=True,
tokenize=True,
start_pos=0,
start_char=0,
mode="",
**kwargs,
):
if kwargs.get("docix", None) == self._docix and self._cache:
yield from self.cache
else:
t = CylleneusToken(positions,
chars,
removestops=removestops,
mode=mode,
**kwargs)
if t.mode == "query":
t.original = t.text = value.translate(jvmap)
yield t
elif t.mode == "index":
if not tokenize:
t.original = t.text = "\n".join(
[el for el in flatten(value["text"])])
t.boost = 1.0
if positions:
t.pos = start_pos
if chars:
t.startchar = start_char
t.endchar = start_char + len(t.original)
yield t
else:
self._cache = []
self._docix = kwargs.get("docix", None)
tokenizer = PunktLatinCharsVars()
stopchars = str.maketrans("", "",
string.punctuation + "“”—\n")
divs = {
i: div.lower()
for i, div in enumerate(value["meta"].split("-"))
}
sect_sent = 0
prev_sect = 0
sect_pos = 0
for i, (path, text) in enumerate(
nested_dict_iter(value["text"])):
sent_id = i
if len(path) >= 2 and int(path[-2]) > prev_sect:
sect_sent = 0
sect_pos = 0
prev_sect = int(path[-2])
tokens = []
temp_tokens = tokenizer.word_tokenize(text)
if temp_tokens:
if (temp_tokens[0].replace("j",
"i").replace("v", "u")
not in proper_names.proper_names):
temp_tokens[0] = temp_tokens[0]
for ix, token in enumerate(temp_tokens):
ppp = compound.is_ppp(token)
if ppp and ix < len(temp_tokens) - 2:
copula = compound.is_copula(
temp_tokens[ix + 2]) # whitespace
if copula and ppp[1] == copula[2]:
tense, mood, number, i = copula
token = f"{token} {compound.copula[tense][mood][number][i]}"
del temp_tokens[ix + 1:ix + 3]
tokens.insert(ix, token)
else:
tokens.append(token)
else:
tokens.append(token)
pos = 0
sent_pos = 0
for token in tokens:
meta = {"meta": value["meta"].lower()}
for i in range(len(divs)):
meta[divs[i]] = str(int(path[i]) + 1)
t.boost = 1.0
if keeporiginal:
t.original = token
t.stopped = False
token = convert_diphthongs(
strip_diacritics(token)).translate(jvmap)
if (token in (" ", "\n") or token in punctuation
or token in stopchars):
pos -= 1
else:
pos += 2
if positions:
t.pos = start_pos + pos
original_length = len(token)
token = token.strip()
ltoken = token.lstrip(string.punctuation)
ldiff = original_length - len(ltoken)
if ldiff != 0:
token = ltoken
rtoken = token.rstrip(string.punctuation)
rdiff = len(token) - len(rtoken)
if rdiff != 0:
token = rtoken
ntoken = token.translate(stopchars)
ndiff = len(token) - len(ntoken)
if ndiff:
token = ntoken
if not token:
start_char += original_length
continue
meta["sect_sent"] = sect_sent
meta["sect_pos"] = sect_pos
meta["sent_id"] = sent_id
meta["sent_pos"] = sent_pos
t.meta = meta
is_enclitic = False
if token not in exceptions:
if t.original in replacements:
for subtoken in replacements[t.original]:
t.text = subtoken
t.startchar = start_char
t.endchar = (start_char +
original_length)
if mode == "index":
self._cache.append(
copy.deepcopy(t))
yield t
start_char += original_length
continue
if re.match(r"(?:\w+) (?:\w+)", token):
ppp, copula = token.split(" ")
t.text = ppp
t.startchar = start_char
t.endchar = start_char + len(ppp) + 1
if mode == "index":
self._cache.append(copy.deepcopy(t))
yield t
t.text = copula
t.startchar = start_char + len(ppp)
t.endchar = (start_char + len(ppp) +
len(copula))
if mode == "index":
self._cache.append(copy.deepcopy(t))
yield t
start_char += original_length
continue
for enclitic in enclitics:
if token.endswith(enclitic):
if enclitic == "ne":
t.text = token[:-len(enclitic)]
t.startchar = start_char
t.endchar = start_char + (
len(token) - len(enclitic))
if mode == "index":
self._cache.append(
copy.deepcopy(t))
yield t
t.text = "ne"
t.startchar = start_char + len(
token[:-len(enclitic)])
t.endchar = (
start_char +
len(token[:-len(enclitic)]) +
len(enclitic))
if mode == "index":
self._cache.append(
copy.deepcopy(t))
yield t
elif enclitic == "n":
t.text = (token[:-len(enclitic)] +
"s")
t.startchar = start_char
t.endchar = (start_char +
len(token) -
len(enclitic))
if mode == "index":
self._cache.append(
copy.deepcopy(t))
yield t
t.text = "ne"
t.startchar = start_char + len(
token[:-len(enclitic)])
t.endchar = (
start_char +
len(token[:-len(enclitic)]) +
len(enclitic))
if mode == "index":
self._cache.append(
copy.deepcopy(t))
yield t
elif enclitic == "st":
if token.endswith("ust"):
t.text = token[:-len(enclitic)]
t.startchar = start_char
t.endchar = (start_char + len(
token[:-len(enclitic)]) -
len(enclitic))
if mode == "index":
self._cache.append(
copy.deepcopy(t))
yield t
t.text = "est"
t.startchar = start_char + len(
token[:-len(enclitic)])
t.endchar = (start_char + len(
token[:-len(enclitic)]) +
len(enclitic))
if mode == "index":
self._cache.append(
copy.deepcopy(t))
yield t
else:
t.text = token[:-len(enclitic)]
t.startchar = start_char
t.endchar = (start_char + len(
token[:-len(enclitic)]) -
len(enclitic))
if mode == "index":
self._cache.append(
copy.deepcopy(t))
yield t
t.text = "est"
t.startchar = start_char + len(
token[:-len(enclitic)])
t.endchar = (start_char + len(
token[:-len(enclitic)]) +
len(enclitic))
if mode == "index":
self._cache.append(
copy.deepcopy(t))
yield t
elif enclitic == "'s":
t.text = token + "s"
t.startchar = start_char
t.endchar = start_char + len(token)
if mode == "index":
self._cache.append(
copy.deepcopy(t))
yield t
t.text = "es"
t.startchar = (start_char +
len(token) + 1)
t.endchar = (start_char +
len(token) +
len(enclitic))
if mode == "index":
self._cache.append(
copy.deepcopy(t))
yield t
else:
t.text = token[:-len(enclitic)]
t.startchar = start_char
t.endchar = start_char + len(
token[:-len(enclitic)])
if mode == "index":
self._cache.append(
copy.deepcopy(t))
yield t
t.text = enclitic
t.startchar = start_char + len(
token[:-len(enclitic)])
t.endchar = (
start_char +
len(token[:-len(enclitic)]) +
len(enclitic))
if mode == "index":
self._cache.append(
copy.deepcopy(t))
yield t
is_enclitic = True
break
if not is_enclitic:
t.text = token
if chars:
t.startchar = start_char + ldiff
t.endchar = (start_char + original_length -
rdiff) # - ndiff - rdiff
if mode == "index":
self._cache.append(copy.deepcopy(t))
yield t
start_char += original_length
sent_pos += 1
sect_pos += 1
start_char += 1
def __call__(
self,
value,
positions=False,
chars=False,
keeporiginal=True,
removestops=True,
tokenize=True,
start_pos=0,
start_char=0,
mode="",
**kwargs
):
if kwargs.get("docix", None) == self._docix and self._cache:
yield from self.cache
else:
t = CylleneusToken(
positions, chars, removestops=removestops, mode=mode, **kwargs
)
if t.mode == "query":
t.text = t.original = value
yield t
else:
if not tokenize:
lines = []
for line in value["text"]:
line = re.sub(r"\t+", "\t", line.strip())
if line and line.startswith("# text_line"):
text = line.split("# text_line: ")[1]
lines.append(text)
t.original = t.text = "\n".join([line for line in lines])
t.boost = 1.0
if positions:
t.pos = start_pos
if chars:
t.startchar = start_char
t.endchar = start_char + len(t.text)
yield t
else:
self._cache = []
self._docix = kwargs.get("docix", None)
t.boost = 1.0
t.pos = t.startchar = t.endchar = 0
meta = {
"text": None, # work title
"text_id": None,
"chapter": None, # reference
"chapter_id": None,
"text_line": None, # the text
"text_line_id": None,
"text_line_counter": None, # line number
"text_line_subcounter": None, # token number
}
sect_pos = 0
sent_pos = 0
for line in value["text"]:
line = line.strip()
if line:
if line.startswith("#"):
try:
label, value = line.split(":", maxsplit=1)
except ValueError:
continue
label = label.split(" ", maxsplit=1)[1].strip()
value = value.strip()
meta[label] = (
value
if not value.isnumeric()
else int(value)
)
if label in [
"text_line_counter",
"text_line_subcounter",
]:
sent_pos = 0
else:
try:
(
ID,
FORM,
LEMMA,
UPOS,
XPOS,
MORPHO,
_,
_,
_,
_,
LEMMA_ID,
PADA,
SEM,
) = line.split("\t")
except ValueError:
try:
(
ID,
FORM,
LEMMA,
_,
XPOS,
_,
_,
_,
_,
LEMMA_ID,
_,
_,
) = line.split("\t")
except ValueError:
try:
(
ID,
FORM,
_,
_,
_,
_,
_,
_,
_,
_,
) = line.split("\t")
except ValueError:
continue
else:
t.original = FORM
sect_pos += 1
sent_pos += 1
t.pos = sent_pos
continue
else:
if FORM == "_":
t.text = t.original
else:
sect_pos += 1
sent_pos += 1
t.text = FORM
t.original = FORM
t.pos = sent_pos
t.lemma = LEMMA
t.dcs_id = LEMMA_ID
t.morphosyntax = XPOS
t.morpho = None
t.synset = None
t.meta = {
"meta": "chapter-line",
"chapter": meta["chapter"],
"line": meta["text_line_counter"],
"sect_pos": sect_pos,
"sect_sent": meta[
"text_line_counter"
],
"sent_id": meta["text_line_id"],
"sent_pos": sent_pos,
}
t.startchar = start_char
t.endchar = start_char + len(
t.original
)
yield t
# # Emit Devanagari
# t.text = slp2deva(iast2slp(t.text))
# t.mode = "skip"
# yield t
start_char += len(t.original) + 1
else:
if FORM == "_":
t.text = t.original
else:
sect_pos += 1
sent_pos += 1
t.text = FORM
t.original = FORM
t.pos = sent_pos
t.lemma = LEMMA
t.dcs_id = LEMMA_ID
t.morphosyntax = XPOS
t.morpho = None if MORPHO == "_" or not MORPHO else parse_morpho(XPOS, MORPHO)
t.synset = None if SEM == "_" else SEM
t.meta = {
"meta": "chapter-line",
"chapter": meta["chapter"],
"line": meta["text_line_counter"],
"sect_pos": sect_pos,
"sect_sent": meta[
"text_line_counter"
],
"sent_id": meta["text_line_id"],
"sent_pos": sent_pos,
}
t.startchar = start_char
t.endchar = start_char + len(t.original)
yield t
start_char += len(t.original) + 1
def __call__(
self,
value: str,
positions=True,
chars=True,
keeporiginal=True,
removestops=True,
tokenize=True,
start_pos=0,
start_char=0,
mode="",
**kwargs,
):
if kwargs.get("docix") == self._docix and self._cache:
yield from self.cache
else:
t = CylleneusToken(positions,
chars,
removestops=removestops,
mode=mode,
**kwargs)
stopchars = '!"#$%()*+,-—./:;<=>?@[\]^_`{|}~'
punctuation = str.maketrans("", "", stopchars)
if t.mode == "query":
t.original = t.text = value.translate(jvmap)
yield t
else:
if not tokenize:
# Assume value is a list
for pos, token in enumerate(value):
t.original = t.text = token
t.boost = 1.0
if positions:
t.pos = pos
if chars:
t.startchar = start_char
t.endchar = start_char + len(token)
start_char += len(token)
yield t
else:
self._cache = []
self._docix = kwargs.get("docix", None)
work_pos = 0
for i, sentence in enumerate(
sent_tokenizer.tokenize(value)):
sent_pos = 0
for token in word_tokenizer.word_tokenize(sentence):
if token in string.whitespace:
start_char += 1
continue
t.boost = 1.0
if keeporiginal:
t.original = token
original_length = len(token)
t.stopped = False
token = convert_diphthongs(
strip_diacritics(token)).translate(jvmap)
if token in stopchars:
start_char += original_length
continue
t.text = token.translate(punctuation)
if positions:
t.pos = start_pos + work_pos
if chars:
t.startchar = start_char
t.endchar = start_char + original_length
t.meta = {"sent_id": i, "sent_pos": sent_pos}
if mode == "index" and self.cached:
self._cache.append(copy.copy(t))
yield t
work_pos += 1
sent_pos += 1
start_char += original_length
start_char += 1
def __call__(
self,
value,
positions=True,
chars=True,
keeporiginal=True,
removestops=True,
tokenize=True,
start_pos=0,
start_char=0,
mode="",
**kwargs,
):
if (kwargs.get("docix", None) == self._docix
and self._cache is not None):
yield from self.cache
else:
t = CylleneusToken(positions,
chars,
removestops=removestops,
mode=mode,
**kwargs)
if t.mode == "query":
t.original = t.text = value.translate(jvmap)
yield t
else:
if not tokenize:
text = "\n".join([el for el in flatten(value["text"])])
t.original = t.text = text
t.boost = 1.0
if positions:
t.pos = start_pos
if chars:
t.startchar = start_char
t.endchar = start_char + len(text)
yield t
else:
self._cache = []
self._docix = kwargs.get("docix", None)
tokenizer = PunktLatinCharsVars()
stopchars = str.maketrans("", "", string.punctuation)
doc = value["text"]
divs = [
cref.get("n") for cref in doc.findall(
".//{http://www.tei-c.org/ns/1.0}cRefPattern")
]
tei_base = "/tei:TEI/tei:text/tei:body/tei:div"
# Prose divisions
sentences = doc.xpath(
tei_base + ("/tei:div" * len(divs)),
namespaces={"tei": "http://www.tei-c.org/ns/1.0"},
)
# Fall back to poetry divisions
if len(sentences) == 0:
sentences = doc.xpath(
tei_base + ("/tei:div" * (len(divs) - 1)) +
"/tei:l",
namespaces={"tei": "http://www.tei-c.org/ns/1.0"},
)
# Fall back to speaker divisions (plays)
if len(sentences) == 0:
sentences = doc.xpath(
tei_base + ("/tei:div" * (len(divs) - 1)) +
"/tei:sp/tei:l",
namespaces={"tei": "http://www.tei-c.org/ns/1.0"},
)
for i, sentence in enumerate(sentences):
meta = {
"meta": "-".join(divs),
divs[-1]: sentence.get("n"),
"sent_id": i,
}
el = sentence
j = -1
while el is not None:
if el.getparent() is not None:
if (el.getparent().get("type", None)
== "textpart" or el.getparent().tag
== "{http://www.tei-c.org/ns/1.0}sp"):
if (el.getparent().tag ==
"{http://www.tei-c.org/ns/1.0}sp"):
meta["speaker"] = (el.getparent().find(
".//{http://www.tei-c.org/ns/1.0}speaker"
).text)
elif (el.getparent().get(
"type", None) == "textpart"):
j -= 1
meta[divs[j]] = el.getparent().get("n")
el = el.getparent()
text = sentence.text
# If the text is not embedded in an XML node, use the 'text' attribute
if not text:
text = stringify(sentence)
if not text:
continue
tokens = []
temp_tokens = tokenizer.word_tokenize(text.strip())
if temp_tokens:
if (temp_tokens[0].replace("j",
"i").replace("v", "u")
not in proper_names.proper_names):
temp_tokens[0] = temp_tokens[0]
for ix, token in enumerate(temp_tokens):
ppp = compound.is_ppp(token)
if ppp and ix < len(temp_tokens) - 2:
copula = compound.is_copula(
temp_tokens[ix + 2]) # whitespace
if copula and ppp[1] == copula[2]:
tense, mood, number, i = copula
token = f"{token} {compound.copula[tense][mood][number][i]}"
del temp_tokens[ix + 1:ix + 3]
tokens.insert(ix, token)
else:
tokens.append(token)
else:
tokens.append(token)
pos = 0
for token in tokens:
meta["sent_pos"] = pos
t.boost = 1.0
if keeporiginal:
t.original = token
t.stopped = False
token = convert_diphthongs(
strip_diacritics(token)).translate(jvmap)
if positions:
t.pos = start_pos + pos
if (token == " " or token in punctuation
or token in stopchars):
pos += 1
continue
original_length = len(token)
token = token.strip()
ltoken = token.lstrip(string.punctuation)
ldiff = original_length - len(ltoken)
if ldiff != 0:
token = ltoken
rtoken = token.rstrip(string.punctuation)
rdiff = len(token) - len(rtoken)
if rdiff != 0:
token = rtoken
ntoken = token.translate(stopchars)
ndiff = len(token) - len(ntoken)
if ndiff:
token = ntoken
if not token:
start_char += original_length
continue
t.meta = copy.deepcopy(meta)
is_enclitic = False
if token not in exceptions:
if t.original in replacements:
for subtoken in replacements[t.original]:
t.text = subtoken
t.startchar = start_char
t.endchar = (start_char +
original_length)
if mode == "index":
self._cache.append(
copy.deepcopy(t))
yield t
start_char += original_length
continue
if re.match(r"(?:\w+) (?:\w+)", token):
ppp, copula = token.split(" ")
t.text = ppp
t.startchar = start_char
t.endchar = start_char + len(ppp) + 1
if mode == "index":
self._cache.append(copy.deepcopy(t))
yield t
t.text = copula
t.startchar = start_char + len(ppp)
t.endchar = (start_char + len(ppp) +
len(copula))
if mode == "index":
self._cache.append(copy.deepcopy(t))
yield t
start_char += original_length
continue
for enclitic in enclitics:
if token.endswith(enclitic):
if enclitic == "ne":
t.text = token[:-len(enclitic)]
t.startchar = start_char
t.endchar = start_char + (
len(token) - len(enclitic))
if mode == "index":
self._cache.append(
copy.deepcopy(t))
yield t
t.text = "ne"
t.startchar = start_char + len(
token[:-len(enclitic)])
t.endchar = (
start_char +
len(token[:-len(enclitic)]) +
len(enclitic))
if mode == "index":
self._cache.append(
copy.deepcopy(t))
yield t
elif enclitic == "n":
t.text = (token[:-len(enclitic)] +
"s")
t.startchar = start_char
t.endchar = (start_char +
len(token) -
len(enclitic))
if mode == "index":
self._cache.append(
copy.deepcopy(t))
yield t
t.text = "ne"
t.startchar = start_char + len(
token[:-len(enclitic)])
t.endchar = (
start_char +
len(token[:-len(enclitic)]) +
len(enclitic))
if mode == "index":
self._cache.append(
copy.deepcopy(t))
yield t
elif enclitic == "st":
if token.endswith("ust"):
t.text = token[:-len(enclitic)]
t.startchar = start_char
t.endchar = (start_char + len(
token[:-len(enclitic)]) -
len(enclitic))
if mode == "index":
self._cache.append(
copy.deepcopy(t))
yield t
t.text = "est"
t.startchar = start_char + len(
token[:-len(enclitic)])
t.endchar = (start_char + len(
token[:-len(enclitic)]) +
len(enclitic))
if mode == "index":
self._cache.append(
copy.deepcopy(t))
yield t
else:
t.text = token[:-len(enclitic)]
t.startchar = start_char
t.endchar = (start_char + len(
token[:-len(enclitic)]) -
len(enclitic))
if mode == "index":
self._cache.append(
copy.deepcopy(t))
yield t
t.text = "est"
t.startchar = start_char + len(
token[:-len(enclitic)])
t.endchar = (start_char + len(
token[:-len(enclitic)]) +
len(enclitic))
if mode == "index":
self._cache.append(
copy.deepcopy(t))
yield t
elif enclitic == "'s":
t.text = token + "s"
t.startchar = start_char
t.endchar = start_char + len(token)
if mode == "index":
self._cache.append(
copy.deepcopy(t))
yield t
t.text = "es"
t.startchar = (start_char +
len(token) + 1)
t.endchar = (start_char +
len(token) +
len(enclitic))
if mode == "index":
self._cache.append(
copy.deepcopy(t))
yield t
else:
t.text = token[:-len(enclitic)]
t.startchar = start_char
t.endchar = start_char + len(
token[:-len(enclitic)])
if mode == "index":
self._cache.append(
copy.deepcopy(t))
yield t
t.text = enclitic
t.startchar = start_char + len(
token[:-len(enclitic)])
t.endchar = (
start_char +
len(token[:-len(enclitic)]) +
len(enclitic))
if mode == "index":
self._cache.append(
copy.deepcopy(t))
yield t
is_enclitic = True
break
if not is_enclitic:
t.text = token
if chars:
t.startchar = start_char + ldiff
t.endchar = (start_char + original_length -
rdiff)
if mode == "index":
self._cache.append(copy.deepcopy(t))
yield t
start_char += original_length
start_char += 1
