File size: 6,043 Bytes
aeb6dbc 8bc2fc9 aeb6dbc 82adb1c 3d5f9ea aeb6dbc 6e567cd aeb6dbc 6e567cd aeb6dbc 6e567cd 8bc2fc9 6e567cd aeb6dbc 6e567cd aeb6dbc 6e567cd aeb6dbc 82adb1c aeb6dbc 0404a52 aeb6dbc 0404a52 aeb6dbc 0404a52 aeb6dbc |
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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 |
# 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.
#
import logging
from tika import parser
import re
from io import BytesIO
from deepdoc.parser.utils import get_text
from rag.nlp import bullets_category, is_english,remove_contents_table, \
hierarchical_merge, make_colon_as_title, naive_merge, random_choices, tokenize_table, \
tokenize_chunks
from rag.nlp import rag_tokenizer
from deepdoc.parser import PdfParser, DocxParser, PlainParser, HtmlParser
class Pdf(PdfParser):
def __call__(self, filename, binary=None, from_page=0,
to_page=100000, zoomin=3, callback=None):
from timeit import default_timer as timer
start = timer()
callback(msg="OCR started")
self.__images__(
filename if not binary else binary,
zoomin,
from_page,
to_page,
callback)
callback(msg="OCR finished ({:.2f}s)".format(timer() - start))
start = timer()
self._layouts_rec(zoomin)
callback(0.67, "Layout analysis ({:.2f}s)".format(timer() - start))
logging.debug("layouts: {}".format(timer() - start))
start = timer()
self._table_transformer_job(zoomin)
callback(0.68, "Table analysis ({:.2f}s)".format(timer() - start))
start = timer()
self._text_merge()
tbls = self._extract_table_figure(True, zoomin, True, True)
self._naive_vertical_merge()
self._filter_forpages()
self._merge_with_same_bullet()
callback(0.8, "Text extraction ({:.2f}s)".format(timer() - start))
return [(b["text"] + self._line_tag(b, zoomin), b.get("layoutno", ""))
for b in self.boxes], tbls
def chunk(filename, binary=None, from_page=0, to_page=100000,
lang="Chinese", callback=None, **kwargs):
"""
Supported file formats are docx, pdf, txt.
Since a book is long and not all the parts are useful, if it's a PDF,
please setup the page ranges for every book in order eliminate negative effects and save elapsed computing time.
"""
doc = {
"docnm_kwd": filename,
"title_tks": rag_tokenizer.tokenize(re.sub(r"\.[a-zA-Z]+$", "", filename))
}
doc["title_sm_tks"] = rag_tokenizer.fine_grained_tokenize(doc["title_tks"])
pdf_parser = None
sections, tbls = [], []
if re.search(r"\.docx$", filename, re.IGNORECASE):
callback(0.1, "Start to parse.")
doc_parser = DocxParser()
# TODO: table of contents need to be removed
sections, tbls = doc_parser(
binary if binary else filename, from_page=from_page, to_page=to_page)
remove_contents_table(sections, eng=is_english(
random_choices([t for t, _ in sections], k=200)))
tbls = [((None, lns), None) for lns in tbls]
callback(0.8, "Finish parsing.")
elif re.search(r"\.pdf$", filename, re.IGNORECASE):
pdf_parser = Pdf() if kwargs.get(
"parser_config", {}).get(
"layout_recognize", True) else PlainParser()
sections, tbls = pdf_parser(filename if not binary else binary,
from_page=from_page, to_page=to_page, callback=callback)
elif re.search(r"\.txt$", filename, re.IGNORECASE):
callback(0.1, "Start to parse.")
txt = get_text(filename, binary)
sections = txt.split("\n")
sections = [(line, "") for line in sections if line]
remove_contents_table(sections, eng=is_english(
random_choices([t for t, _ in sections], k=200)))
callback(0.8, "Finish parsing.")
elif re.search(r"\.(htm|html)$", filename, re.IGNORECASE):
callback(0.1, "Start to parse.")
sections = HtmlParser()(filename, binary)
sections = [(line, "") for line in sections if line]
remove_contents_table(sections, eng=is_english(
random_choices([t for t, _ in sections], k=200)))
callback(0.8, "Finish parsing.")
elif re.search(r"\.doc$", filename, re.IGNORECASE):
callback(0.1, "Start to parse.")
binary = BytesIO(binary)
doc_parsed = parser.from_buffer(binary)
sections = doc_parsed['content'].split('\n')
sections = [(line, "") for line in sections if line]
remove_contents_table(sections, eng=is_english(
random_choices([t for t, _ in sections], k=200)))
callback(0.8, "Finish parsing.")
else:
raise NotImplementedError(
"file type not supported yet(doc, docx, pdf, txt supported)")
make_colon_as_title(sections)
bull = bullets_category(
[t for t in random_choices([t for t, _ in sections], k=100)])
if bull >= 0:
chunks = ["\n".join(ck)
for ck in hierarchical_merge(bull, sections, 5)]
else:
sections = [s.split("@") for s, _ in sections]
sections = [(pr[0], "@" + pr[1]) if len(pr) == 2 else (pr[0], '') for pr in sections ]
chunks = naive_merge(
sections, kwargs.get(
"chunk_token_num", 256), kwargs.get(
"delimer", "\n。；！？"))
# is it English
# is_english(random_choices([t for t, _ in sections], k=218))
eng = lang.lower() == "english"
res = tokenize_table(tbls, doc, eng)
res.extend(tokenize_chunks(chunks, doc, eng, pdf_parser))
return res
if __name__ == "__main__":
import sys
def dummy(prog=None, msg=""):
pass
chunk(sys.argv[1], from_page=1, to_page=10, callback=dummy)
|