File size: 4,708 Bytes
9cba22c f4456af cfd6ece 3245107 f4456af 2d09c38 f4456af cfd6ece f4456af cfd6ece f4456af 51482f3 79ada0b 51482f3 79ada0b 51482f3 f4456af |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 |
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
#
from docx import Document
import re
import pandas as pd
from collections import Counter
from rag.nlp import rag_tokenizer
from io import BytesIO
class RAGFlowDocxParser:
def __extract_table_content(self, tb):
df = []
for row in tb.rows:
df.append([c.text for c in row.cells])
return self.__compose_table_content(pd.DataFrame(df))
def __compose_table_content(self, df):
def blockType(b):
patt = [
("^(20|19)[0-9]{2}[年/-][0-9]{1,2}[月/-][0-9]{1,2}日*$", "Dt"),
(r"^(20|19)[0-9]{2}年$", "Dt"),
(r"^(20|19)[0-9]{2}[年/-][0-9]{1,2}月*$", "Dt"),
("^[0-9]{1,2}[月/-][0-9]{1,2}日*$", "Dt"),
(r"^第*[一二三四1-4]季度$", "Dt"),
(r"^(20|19)[0-9]{2}年*[一二三四1-4]季度$", "Dt"),
(r"^(20|19)[0-9]{2}[ABCDE]$", "DT"),
("^[0-9.,+%/ -]+$", "Nu"),
(r"^[0-9A-Z/\._~-]+$", "Ca"),
(r"^[A-Z]*[a-z' -]+$", "En"),
(r"^[0-9.,+-]+[0-9A-Za-z/$￥%<>（）()' -]+$", "NE"),
(r"^.{1}$", "Sg")
]
for p, n in patt:
if re.search(p, b):
return n
tks = [t for t in rag_tokenizer.tokenize(b).split(" ") if len(t) > 1]
if len(tks) > 3:
if len(tks) < 12:
return "Tx"
else:
return "Lx"
if len(tks) == 1 and rag_tokenizer.tag(tks[0]) == "nr":
return "Nr"
return "Ot"
if len(df) < 2:
return []
max_type = Counter([blockType(str(df.iloc[i, j])) for i in range(
1, len(df)) for j in range(len(df.iloc[i, :]))])
max_type = max(max_type.items(), key=lambda x: x[1])[0]
colnm = len(df.iloc[0, :])
hdrows = [0] # header is not nessesarily appear in the first line
if max_type == "Nu":
for r in range(1, len(df)):
tys = Counter([blockType(str(df.iloc[r, j]))
for j in range(len(df.iloc[r, :]))])
tys = max(tys.items(), key=lambda x: x[1])[0]
if tys != max_type:
hdrows.append(r)
lines = []
for i in range(1, len(df)):
if i in hdrows:
continue
hr = [r - i for r in hdrows]
hr = [r for r in hr if r < 0]
t = len(hr) - 1
while t > 0:
if hr[t] - hr[t - 1] > 1:
hr = hr[t:]
break
t -= 1
headers = []
for j in range(len(df.iloc[i, :])):
t = []
for h in hr:
x = str(df.iloc[i + h, j]).strip()
if x in t:
continue
t.append(x)
t = ",".join(t)
if t:
t += ": "
headers.append(t)
cells = []
for j in range(len(df.iloc[i, :])):
if not str(df.iloc[i, j]):
continue
cells.append(headers[j] + str(df.iloc[i, j]))
lines.append(";".join(cells))
if colnm > 3:
return lines
return ["\n".join(lines)]
def __call__(self, fnm, from_page=0, to_page=100000):
self.doc = Document(fnm) if isinstance(
fnm, str) else Document(BytesIO(fnm))
pn = 0
secs = []
for p in self.doc.paragraphs:
if pn > to_page:
break
if from_page <= pn < to_page and p.text.strip():
secs.append((p.text, p.style.name))
for run in p.runs:
if 'lastRenderedPageBreak' in run._element.xml:
pn += 1
continue
if 'w:br' in run._element.xml and 'type="page"' in run._element.xml:
pn += 1
tbls = [self.__extract_table_content(tb) for tb in self.doc.tables]
return secs, tbls
|