add OCR annotation functionality
Kai
4 years ago
parent
a834da494a
commit
c3acd74e9f
@ -0,0 +1 @@
|
||||
__import__('pkg_resources').declare_namespace(__name__)
|
||||
@ -1,393 +0,0 @@
|
||||
import re
|
||||
import click
|
||||
import pandas as pd
|
||||
from io import StringIO
|
||||
import os
|
||||
import xml.etree.ElementTree as ET
|
||||
import requests
|
||||
import unicodedata
|
||||
import json
|
||||
|
||||
|
||||
@click.command()
|
||||
@click.argument('tsv-file', type=click.Path(exists=True), required=True, nargs=1)
|
||||
@click.argument('url-file', type=click.Path(exists=False), required=True, nargs=1)
|
||||
def extract_document_links(tsv_file, url_file):
|
||||
|
||||
parts = extract_doc_links(tsv_file)
|
||||
|
||||
urls = [part['url'] for part in parts]
|
||||
|
||||
urls = pd.DataFrame(urls, columns=['url'])
|
||||
|
||||
urls.to_csv(url_file, sep="\t", quoting=3, index=False)
|
||||
|
||||
|
||||
@click.command()
|
||||
@click.argument('tsv-file', type=click.Path(exists=True), required=True, nargs=1)
|
||||
@click.argument('annotated-tsv-file', type=click.Path(exists=False), required=True, nargs=1)
|
||||
def annotate_tsv(tsv_file, annotated_tsv_file):
|
||||
|
||||
parts = extract_doc_links(tsv_file)
|
||||
|
||||
annotated_parts = []
|
||||
|
||||
for part in parts:
|
||||
|
||||
part_data = StringIO(part['header'] + part['text'])
|
||||
|
||||
df = pd.read_csv(part_data, sep="\t", comment='#', quoting=3)
|
||||
|
||||
df['url_id'] = len(annotated_parts)
|
||||
|
||||
annotated_parts.append(df)
|
||||
|
||||
df = pd.concat(annotated_parts)
|
||||
|
||||
df.to_csv(annotated_tsv_file, sep="\t", quoting=3, index=False)
|
||||
|
||||
|
||||
def extract_doc_links(tsv_file):
|
||||
|
||||
parts = []
|
||||
|
||||
header = None
|
||||
|
||||
with open(tsv_file, 'r') as f:
|
||||
|
||||
text = []
|
||||
url = None
|
||||
|
||||
for line in f:
|
||||
|
||||
if header is None:
|
||||
header = "\t".join(line.split()) + '\n'
|
||||
continue
|
||||
|
||||
urls = [url for url in
|
||||
re.findall(r'http[s]?://(?:[a-zA-Z]|[0-9]|[$-_@.&+]|[!*\(\),]|(?:%[0-9a-fA-F][0-9a-fA-F]))+', line)]
|
||||
|
||||
if len(urls) > 0:
|
||||
if url is not None:
|
||||
parts.append({"url": url, 'header': header, 'text': "".join(text)})
|
||||
text = []
|
||||
|
||||
url = urls[-1]
|
||||
else:
|
||||
if url is None:
|
||||
continue
|
||||
|
||||
line = '\t'.join(line.split())
|
||||
|
||||
if line.count('\t') == 2:
|
||||
|
||||
line = "\t" + line
|
||||
|
||||
if line.count('\t') >= 3:
|
||||
|
||||
text.append(line + '\n')
|
||||
|
||||
continue
|
||||
|
||||
if line.startswith('#'):
|
||||
continue
|
||||
|
||||
if len(line) == 0:
|
||||
continue
|
||||
|
||||
print('Line error: |', line, '|Number of Tabs: ', line.count('\t'))
|
||||
|
||||
if url is not None:
|
||||
parts.append({"url": url, 'header': header, 'text': "".join(text)})
|
||||
|
||||
return parts
|
||||
|
||||
|
||||
def ner(tsv, ner_rest_endpoint):
|
||||
|
||||
resp = requests.post(url=ner_rest_endpoint, json={'text': " ".join(tsv.TOKEN.astype(str).tolist())})
|
||||
|
||||
resp.raise_for_status()
|
||||
|
||||
def iterate_ner_results(result_sentences):
|
||||
|
||||
for sen in result_sentences:
|
||||
|
||||
for token in sen:
|
||||
|
||||
yield unicodedata.normalize('NFC', token['word']), token['prediction'], False
|
||||
|
||||
yield '', '', True
|
||||
|
||||
ner_result = json.loads(resp.content)
|
||||
|
||||
result_sequence = iterate_ner_results(ner_result)
|
||||
|
||||
tsv_result = []
|
||||
for idx, row in tsv.iterrows():
|
||||
|
||||
row_token = unicodedata.normalize('NFC', str(row.TOKEN).replace(' ', ''))
|
||||
|
||||
ner_token_concat = ''
|
||||
while row_token != ner_token_concat:
|
||||
|
||||
ner_token, ner_tag, sentence_break = next(result_sequence)
|
||||
ner_token_concat += ner_token
|
||||
|
||||
assert len(row_token) >= len(ner_token_concat)
|
||||
|
||||
if sentence_break:
|
||||
tsv_result.append((0, '', 'O', 'O', '-', row.url_id, row.left, row.right, row.top, row.bottom))
|
||||
else:
|
||||
tsv_result.append((0, ner_token, ner_tag, 'O', '-', row.url_id, row.left, row.right, row.top,
|
||||
row.bottom))
|
||||
|
||||
return pd.DataFrame(tsv_result, columns=['No.', 'TOKEN', 'NE-TAG', 'NE-EMB', 'ID', 'url_id',
|
||||
'left', 'right', 'top', 'bottom']), ner_result
|
||||
|
||||
|
||||
def ned(tsv, ner_result, ned_rest_endpoint, json_file=None, threshold=None):
|
||||
|
||||
if json_file is not None and os.path.exists(json_file):
|
||||
|
||||
print('Loading {}'.format(json_file))
|
||||
|
||||
with open(json_file, "r") as fp:
|
||||
ned_result = json.load(fp)
|
||||
|
||||
else:
|
||||
|
||||
resp = requests.post(url=ned_rest_endpoint + '/parse', json=ner_result)
|
||||
|
||||
resp.raise_for_status()
|
||||
|
||||
ner_parsed = json.loads(resp.content)
|
||||
|
||||
ned_rest_endpoint = ned_rest_endpoint + '/ned?return_full=' + str(json_file is not None).lower()
|
||||
|
||||
resp = requests.post(url=ned_rest_endpoint, json=ner_parsed, timeout=3600000)
|
||||
|
||||
resp.raise_for_status()
|
||||
|
||||
ned_result = json.loads(resp.content)
|
||||
|
||||
rids = []
|
||||
entity = ""
|
||||
entity_type = None
|
||||
tsv['ID'] = '-'
|
||||
|
||||
def check_entity(tag):
|
||||
nonlocal entity, entity_type, rids
|
||||
|
||||
if (entity != "") and ((tag == 'O') or tag.startswith('B-') or (tag[2:] != entity_type)):
|
||||
|
||||
eid = entity + "-" + entity_type
|
||||
|
||||
if eid in ned_result:
|
||||
if 'ranking' in ned_result[eid]:
|
||||
ranking = ned_result[eid]['ranking']
|
||||
|
||||
#tsv.loc[rids, 'ID'] = ranking[0][1]['wikidata'] if threshold is None or ranking[0][1]['proba_1'] >= threshold else ''
|
||||
|
||||
tmp = "|".join([ranking[i][1]['wikidata'] for i in range(len(ranking)) if threshold is None or ranking[i][1]['proba_1'] >= threshold])
|
||||
tsv.loc[rids, 'ID'] = tmp if len(tmp) > 0 else '-'
|
||||
|
||||
rids = []
|
||||
entity = ""
|
||||
entity_type = None
|
||||
|
||||
ner_tmp = tsv.copy()
|
||||
ner_tmp.loc[~ner_tmp['NE-TAG'].isin(['O', 'B-PER', 'B-LOC','B-ORG', 'I-PER', 'I-LOC', 'I-ORG']), 'NE-TAG'] = 'O'
|
||||
|
||||
for rid, row in ner_tmp.iterrows():
|
||||
|
||||
check_entity(row['NE-TAG'])
|
||||
|
||||
if row['NE-TAG'] != 'O':
|
||||
|
||||
entity_type = row['NE-TAG'][2:]
|
||||
|
||||
entity += " " if entity != "" else ""
|
||||
|
||||
entity += str(row['TOKEN'])
|
||||
|
||||
rids.append(rid)
|
||||
|
||||
check_entity('O')
|
||||
|
||||
return tsv, ned_result
|
||||
|
||||
|
||||
@click.command()
|
||||
@click.argument('page-xml-file', type=click.Path(exists=True), required=True, nargs=1)
|
||||
@click.argument('tsv-out-file', type=click.Path(), required=True, nargs=1)
|
||||
@click.option('--image-url', type=str, default='http://empty')
|
||||
@click.option('--ner-rest-endpoint', type=str, default=None,
|
||||
help="REST endpoint of sbb_ner service. See https://github.com/qurator-spk/sbb_ner for details.")
|
||||
@click.option('--ned-rest-endpoint', type=str, default=None,
|
||||
help="REST endpoint of sbb_ned service. See https://github.com/qurator-spk/sbb_ned for details.")
|
||||
@click.option('--noproxy', type=bool, is_flag=True, help='disable proxy. default: enabled.')
|
||||
@click.option('--scale-factor', type=float, default=0.5685, help='default: 0.5685')
|
||||
@click.option('--ned-threshold', type=float, default=None)
|
||||
def page2tsv(page_xml_file, tsv_out_file, image_url, ner_rest_endpoint, ned_rest_endpoint, noproxy, scale_factor,
|
||||
ned_threshold):
|
||||
|
||||
out_columns = ['No.', 'TOKEN', 'NE-TAG', 'NE-EMB', 'ID', 'url_id', 'left', 'right', 'top', 'bottom']
|
||||
|
||||
if noproxy:
|
||||
os.environ['no_proxy'] = '*'
|
||||
|
||||
tree = ET.parse(page_xml_file)
|
||||
xmlns = tree.getroot().tag.split('}')[0].strip('{')
|
||||
|
||||
urls = []
|
||||
if os.path.exists(tsv_out_file):
|
||||
parts = extract_doc_links(tsv_out_file)
|
||||
|
||||
urls = [part['url'] for part in parts]
|
||||
else:
|
||||
pd.DataFrame([], columns=out_columns). to_csv(tsv_out_file, sep="\t", quoting=3, index=False)
|
||||
|
||||
tsv = []
|
||||
line_number = 0
|
||||
rgn_number = 0
|
||||
for region in tree.findall('.//{%s}TextRegion' % xmlns):
|
||||
rgn_number += 1
|
||||
for text_line in region.findall('.//{%s}TextLine' % xmlns):
|
||||
line_number += 1
|
||||
|
||||
for words in text_line.findall('./{%s}Word' % xmlns):
|
||||
|
||||
for word in words.findall('./{%s}TextEquiv/{%s}Unicode' % (xmlns, xmlns)):
|
||||
text = word.text
|
||||
for coords in words.findall('./{%s}Coords' % xmlns):
|
||||
|
||||
# transform OCR coordinates using `scale_factor` to derive
|
||||
# correct coordinates for the web presentation image
|
||||
points = [int(scale_factor * float(pos))
|
||||
for p in coords.attrib['points'].split(' ') for pos in p.split(',')]
|
||||
|
||||
x_points = [points[i] for i in range(0, len(points), 2)]
|
||||
y_points = [points[i] for i in range(1, len(points), 2)]
|
||||
|
||||
left = min(x_points)
|
||||
right = max(x_points)
|
||||
top = min(y_points)
|
||||
bottom = max(y_points)
|
||||
|
||||
tsv.append((rgn_number, line_number, left + (right-left)/2.0,
|
||||
0, text, 'O', 'O', '-', len(urls), left, right, top, bottom))
|
||||
|
||||
with open(tsv_out_file, 'a') as f:
|
||||
|
||||
f.write('# ' + image_url + '\n')
|
||||
|
||||
tsv = pd.DataFrame(tsv, columns=['rid', 'line', 'hcenter'] + out_columns)
|
||||
|
||||
if len(tsv)==0:
|
||||
return
|
||||
|
||||
vlinecenter = pd.DataFrame(tsv[['line', 'top']].groupby('line', sort=False).mean().top +
|
||||
(tsv[['line', 'bottom']].groupby('line', sort=False).mean().bottom -
|
||||
tsv[['line', 'top']].groupby('line', sort=False).mean().top) / 2,
|
||||
columns=['vlinecenter'])
|
||||
|
||||
tsv = tsv.merge(vlinecenter, left_on='line', right_index=True)
|
||||
|
||||
regions = [region.sort_values(['vlinecenter', 'hcenter']) for rid, region in tsv.groupby('rid', sort=False)]
|
||||
|
||||
tsv = pd.concat(regions)
|
||||
|
||||
tsv = tsv[out_columns].reset_index(drop=True)
|
||||
|
||||
try:
|
||||
if ner_rest_endpoint is not None:
|
||||
|
||||
tsv, ner_result = ner(tsv, ner_rest_endpoint)
|
||||
|
||||
if ned_rest_endpoint is not None:
|
||||
|
||||
tsv, _ = ned(tsv, ner_result, ned_rest_endpoint, threshold=ned_threshold)
|
||||
|
||||
tsv.to_csv(tsv_out_file, sep="\t", quoting=3, index=False, mode='a', header=False)
|
||||
except requests.HTTPError as e:
|
||||
print(e)
|
||||
|
||||
|
||||
@click.command()
|
||||
@click.argument('tsv-file', type=click.Path(exists=True), required=True, nargs=1)
|
||||
@click.argument('tsv-out-file', type=click.Path(), required=True, nargs=1)
|
||||
@click.option('--ner-rest-endpoint', type=str, default=None,
|
||||
help="REST endpoint of sbb_ner service. See https://github.com/qurator-spk/sbb_ner for details.")
|
||||
@click.option('--ned-rest-endpoint', type=str, default=None,
|
||||
help="REST endpoint of sbb_ned service. See https://github.com/qurator-spk/sbb_ned for details.")
|
||||
@click.option('--ned-json-file', type=str, default=None)
|
||||
@click.option('--noproxy', type=bool, is_flag=True, help='disable proxy. default: proxy is enabled.')
|
||||
@click.option('--ned-threshold', type=float, default=None)
|
||||
def find_entities(tsv_file, tsv_out_file, ner_rest_endpoint, ned_rest_endpoint, ned_json_file, noproxy, ned_threshold):
|
||||
|
||||
out_columns = ['No.', 'TOKEN', 'NE-TAG', 'NE-EMB', 'ID', 'url_id', 'left', 'right', 'top', 'bottom']
|
||||
|
||||
if noproxy:
|
||||
os.environ['no_proxy'] = '*'
|
||||
|
||||
tsv = pd.read_csv(tsv_file, sep='\t', comment='#', quoting=3).rename(columns={'GND-ID': 'ID'})
|
||||
|
||||
parts = extract_doc_links(tsv_file)
|
||||
|
||||
urls = [part['url'] for part in parts]
|
||||
|
||||
try:
|
||||
if ner_rest_endpoint is not None:
|
||||
|
||||
tsv, ner_result = ner(tsv, ner_rest_endpoint)
|
||||
|
||||
elif os.path.exists(tsv_file):
|
||||
|
||||
print('Using NER information that is already contained in file: {}'.format(tsv_file))
|
||||
|
||||
tmp = tsv.copy()
|
||||
tmp['sen'] = (tmp['No.'] == 0).cumsum()
|
||||
tmp.loc[~tmp['NE-TAG'].isin(['O', 'B-PER', 'B-LOC','B-ORG', 'I-PER', 'I-LOC', 'I-ORG']), 'NE-TAG'] = 'O'
|
||||
|
||||
ner_result = [[{'word': str(row.TOKEN), 'prediction': row['NE-TAG']} for _, row in sen.iterrows()]
|
||||
for _, sen in tmp.groupby('sen')]
|
||||
else:
|
||||
raise RuntimeError("Either NER rest endpoint or NER-TAG information within tsv_file required.")
|
||||
|
||||
if ned_rest_endpoint is not None:
|
||||
|
||||
tsv, ned_result = ned(tsv, ner_result, ned_rest_endpoint, json_file=ned_json_file, threshold=ned_threshold)
|
||||
|
||||
if ned_json_file is not None and not os.path.exists(ned_json_file):
|
||||
|
||||
with open(ned_json_file, "w") as fp_json:
|
||||
json.dump(ned_result, fp_json, indent=2, separators=(',', ': '))
|
||||
|
||||
if len(urls) == 0:
|
||||
print('Writing to {}...'.format(tsv_out_file))
|
||||
|
||||
tsv.to_csv(tsv_out_file, sep="\t", quoting=3, index=False)
|
||||
else:
|
||||
pd.DataFrame([], columns=out_columns). to_csv(tsv_out_file, sep="\t", quoting=3, index=False)
|
||||
|
||||
for url_id, part in tsv.groupby('url_id'):
|
||||
with open(tsv_out_file, 'a') as f:
|
||||
f.write('# ' + urls[url_id] + '\n')
|
||||
|
||||
part.to_csv(tsv_out_file, sep="\t", quoting=3, index=False, mode='a', header=False)
|
||||
|
||||
except requests.HTTPError as e:
|
||||
print(e)
|
||||
|
||||
|
||||
@click.command()
|
||||
@click.argument('xls-file', type=click.Path(exists=True), required=True, nargs=1)
|
||||
def make_page2tsv_commands(xls_file):
|
||||
|
||||
df = pd.read_excel(xls_file)
|
||||
|
||||
for _, row in df.iterrows():
|
||||
print('page2tsv $(OPTIONS) {}.xml {}.tsv --image-url={} --scale-factor={}'.
|
||||
format(row.Filename, row.Filename, row.iiif_url.replace('/full/full', '/left,top,width,height/full'),
|
||||
row.scale_factor))
|
||||
@ -0,0 +1 @@
|
||||
__import__('pkg_resources').declare_namespace(__name__)
|
||||
@ -0,0 +1,248 @@
|
||||
import click
|
||||
import pandas as pd
|
||||
from io import StringIO
|
||||
import os
|
||||
import xml.etree.ElementTree as ET
|
||||
import requests
|
||||
import json
|
||||
|
||||
from .ned import ned
|
||||
from .ner import ner
|
||||
from .tsv import read_tsv, write_tsv, extract_doc_links
|
||||
|
||||
|
||||
@click.command()
|
||||
@click.argument('tsv-file', type=click.Path(exists=True), required=True, nargs=1)
|
||||
@click.argument('url-file', type=click.Path(exists=False), required=True, nargs=1)
|
||||
def extract_document_links(tsv_file, url_file):
|
||||
|
||||
parts = extract_doc_links(tsv_file)
|
||||
|
||||
urls = [part['url'] for part in parts]
|
||||
|
||||
urls = pd.DataFrame(urls, columns=['url'])
|
||||
|
||||
urls.to_csv(url_file, sep="\t", quoting=3, index=False)
|
||||
|
||||
|
||||
@click.command()
|
||||
@click.argument('tsv-file', type=click.Path(exists=True), required=True, nargs=1)
|
||||
@click.argument('annotated-tsv-file', type=click.Path(exists=False), required=True, nargs=1)
|
||||
def annotate_tsv(tsv_file, annotated_tsv_file):
|
||||
|
||||
parts = extract_doc_links(tsv_file)
|
||||
|
||||
annotated_parts = []
|
||||
|
||||
for part in parts:
|
||||
|
||||
part_data = StringIO(part['header'] + part['text'])
|
||||
|
||||
df = pd.read_csv(part_data, sep="\t", comment='#', quoting=3)
|
||||
|
||||
df['url_id'] = len(annotated_parts)
|
||||
|
||||
annotated_parts.append(df)
|
||||
|
||||
df = pd.concat(annotated_parts)
|
||||
|
||||
df.to_csv(annotated_tsv_file, sep="\t", quoting=3, index=False)
|
||||
|
||||
|
||||
@click.command()
|
||||
@click.argument('page-xml-file', type=click.Path(exists=True), required=True, nargs=1)
|
||||
@click.argument('tsv-out-file', type=click.Path(), required=True, nargs=1)
|
||||
@click.option('--purpose', type=click.Choice(['NERD', 'OCR'], case_sensitive=False), default="NERD",
|
||||
help="Purpose of output tsv file. "
|
||||
"\n\nNERD: NER/NED application/ground-truth creation. "
|
||||
"\n\nOCR: OCR application/ground-truth creation. "
|
||||
"\n\ndefault: NERD.")
|
||||
@click.option('--image-url', type=str, default='http://empty')
|
||||
@click.option('--ner-rest-endpoint', type=str, default=None,
|
||||
help="REST endpoint of sbb_ner service. See https://github.com/qurator-spk/sbb_ner for details. "
|
||||
"Only applicable in case of NERD.")
|
||||
@click.option('--ned-rest-endpoint', type=str, default=None,
|
||||
help="REST endpoint of sbb_ned service. See https://github.com/qurator-spk/sbb_ned for details. "
|
||||
"Only applicable in case of NERD.")
|
||||
@click.option('--noproxy', type=bool, is_flag=True, help='disable proxy. default: enabled.')
|
||||
@click.option('--scale-factor', type=float, default=0.5685, help='default: 0.5685')
|
||||
@click.option('--ned-threshold', type=float, default=None)
|
||||
def page2tsv(page_xml_file, tsv_out_file, purpose, image_url, ner_rest_endpoint, ned_rest_endpoint, noproxy, scale_factor,
|
||||
ned_threshold):
|
||||
|
||||
if purpose == "NERD":
|
||||
out_columns = ['No.', 'TOKEN', 'NE-TAG', 'NE-EMB', 'ID', 'url_id', 'left', 'right', 'top', 'bottom']
|
||||
elif purpose == "OCR":
|
||||
out_columns = ['TEXT', 'url_id', 'left', 'right', 'top', 'bottom']
|
||||
else:
|
||||
raise RuntimeError("Unknown purpose.")
|
||||
|
||||
if noproxy:
|
||||
os.environ['no_proxy'] = '*'
|
||||
|
||||
tree = ET.parse(page_xml_file)
|
||||
xmlns = tree.getroot().tag.split('}')[0].strip('{')
|
||||
|
||||
urls = []
|
||||
if os.path.exists(tsv_out_file):
|
||||
parts = extract_doc_links(tsv_out_file)
|
||||
|
||||
urls = [part['url'] for part in parts]
|
||||
else:
|
||||
pd.DataFrame([], columns=out_columns). to_csv(tsv_out_file, sep="\t", quoting=3, index=False)
|
||||
|
||||
tsv = []
|
||||
line_info = []
|
||||
line_number = 0
|
||||
rgn_number = 0
|
||||
for region in tree.findall('.//{%s}TextRegion' % xmlns):
|
||||
rgn_number += 1
|
||||
for text_line in region.findall('.//{%s}TextLine' % xmlns):
|
||||
line_number += 1
|
||||
|
||||
points = [int(scale_factor * float(pos)) for coords in text_line.findall('./{%s}Coords' % xmlns) for p in
|
||||
coords.attrib['points'].split(' ') for pos in p.split(',')]
|
||||
|
||||
x_points, y_points = points[0::2], points[1::2]
|
||||
|
||||
left, right, top, bottom = min(x_points), max(x_points), min(y_points), max(y_points)
|
||||
|
||||
line_info.append((line_number, len(urls), left, right, top, bottom))
|
||||
|
||||
for word in text_line.findall('./{%s}Word' % xmlns):
|
||||
|
||||
for text_equiv in word.findall('./{%s}TextEquiv/{%s}Unicode' % (xmlns, xmlns)):
|
||||
text = text_equiv.text
|
||||
|
||||
for coords in word.findall('./{%s}Coords' % xmlns):
|
||||
|
||||
# transform OCR coordinates using `scale_factor` to derive
|
||||
# correct coordinates for the web presentation image
|
||||
points = [int(scale_factor * float(pos))
|
||||
for p in coords.attrib['points'].split(' ') for pos in p.split(',')]
|
||||
|
||||
x_points, y_points = points[0::2], points[1::2]
|
||||
|
||||
left, right, top, bottom = min(x_points), max(x_points), min(y_points), max(y_points)
|
||||
|
||||
tsv.append((rgn_number, line_number, left + (right - left) / 2.0, text,
|
||||
len(urls), left, right, top, bottom))
|
||||
|
||||
line_info = pd.DataFrame(line_info, columns=['line', 'url_id', 'left', 'right', 'top', 'bottom'])
|
||||
|
||||
tsv = pd.DataFrame(tsv, columns=['rid', 'line', 'hcenter'] + ['TEXT', 'url_id', 'left', 'right', 'top', 'bottom'])
|
||||
|
||||
if len(tsv) == 0:
|
||||
return
|
||||
|
||||
with open(tsv_out_file, 'a') as f:
|
||||
|
||||
f.write('# ' + image_url + '\n')
|
||||
|
||||
vlinecenter = pd.DataFrame(tsv[['line', 'top']].groupby('line', sort=False).mean().top +
|
||||
(tsv[['line', 'bottom']].groupby('line', sort=False).mean().bottom -
|
||||
tsv[['line', 'top']].groupby('line', sort=False).mean().top) / 2,
|
||||
columns=['vlinecenter'])
|
||||
|
||||
tsv = tsv.merge(vlinecenter, left_on='line', right_index=True)
|
||||
|
||||
regions = [region.sort_values(['vlinecenter', 'hcenter']) for rid, region in tsv.groupby('rid', sort=False)]
|
||||
|
||||
tsv = pd.concat(regions)
|
||||
|
||||
if purpose == 'NERD':
|
||||
|
||||
tsv['No.'] = 0
|
||||
tsv['NE-TAG'] = 'O'
|
||||
tsv['NE-EMB'] = 'O'
|
||||
tsv['ID'] = '-'
|
||||
|
||||
tsv = tsv.rename(columns={'TEXT': 'TOKEN'})
|
||||
elif purpose == 'OCR':
|
||||
|
||||
tsv = pd.DataFrame([(line, " ".join(part.TEXT.to_list())) for line, part in tsv.groupby('line')],
|
||||
columns=['line', 'TEXT'])
|
||||
|
||||
tsv = tsv.merge(line_info, left_on='line', right_on='line')
|
||||
|
||||
tsv = tsv[out_columns].reset_index(drop=True)
|
||||
|
||||
try:
|
||||
if purpose == 'NERD' and ner_rest_endpoint is not None:
|
||||
|
||||
tsv, ner_result = ner(tsv, ner_rest_endpoint)
|
||||
|
||||
if ned_rest_endpoint is not None:
|
||||
|
||||
tsv, _ = ned(tsv, ner_result, ned_rest_endpoint, threshold=ned_threshold)
|
||||
|
||||
tsv.to_csv(tsv_out_file, sep="\t", quoting=3, index=False, mode='a', header=False)
|
||||
except requests.HTTPError as e:
|
||||
print(e)
|
||||
|
||||
|
||||
@click.command()
|
||||
@click.argument('tsv-file', type=click.Path(exists=True), required=True, nargs=1)
|
||||
@click.argument('tsv-out-file', type=click.Path(), required=True, nargs=1)
|
||||
@click.option('--ner-rest-endpoint', type=str, default=None,
|
||||
help="REST endpoint of sbb_ner service. See https://github.com/qurator-spk/sbb_ner for details.")
|
||||
@click.option('--ned-rest-endpoint', type=str, default=None,
|
||||
help="REST endpoint of sbb_ned service. See https://github.com/qurator-spk/sbb_ned for details.")
|
||||
@click.option('--ned-json-file', type=str, default=None)
|
||||
@click.option('--noproxy', type=bool, is_flag=True, help='disable proxy. default: proxy is enabled.')
|
||||
@click.option('--ned-threshold', type=float, default=None)
|
||||
def find_entities(tsv_file, tsv_out_file, ner_rest_endpoint, ned_rest_endpoint, ned_json_file, noproxy, ned_threshold):
|
||||
|
||||
if noproxy:
|
||||
os.environ['no_proxy'] = '*'
|
||||
|
||||
tsv, urls = read_tsv(tsv_file)
|
||||
|
||||
try:
|
||||
if ner_rest_endpoint is not None:
|
||||
|
||||
tsv, ner_result = ner(tsv, ner_rest_endpoint)
|
||||
|
||||
elif os.path.exists(tsv_file):
|
||||
|
||||
print('Using NER information that is already contained in file: {}'.format(tsv_file))
|
||||
|
||||
tmp = tsv.copy()
|
||||
tmp['sen'] = (tmp['No.'] == 0).cumsum()
|
||||
tmp.loc[~tmp['NE-TAG'].isin(['O', 'B-PER', 'B-LOC', 'B-ORG', 'I-PER', 'I-LOC', 'I-ORG']), 'NE-TAG'] = 'O'
|
||||
|
||||
ner_result = [[{'word': str(row.TOKEN), 'prediction': row['NE-TAG']} for _, row in sen.iterrows()]
|
||||
for _, sen in tmp.groupby('sen')]
|
||||
else:
|
||||
raise RuntimeError("Either NER rest endpoint or NER-TAG information within tsv_file required.")
|
||||
|
||||
if ned_rest_endpoint is not None:
|
||||
|
||||
tsv, ned_result = ned(tsv, ner_result, ned_rest_endpoint, json_file=ned_json_file, threshold=ned_threshold)
|
||||
|
||||
if ned_json_file is not None and not os.path.exists(ned_json_file):
|
||||
|
||||
with open(ned_json_file, "w") as fp_json:
|
||||
json.dump(ned_result, fp_json, indent=2, separators=(',', ': '))
|
||||
|
||||
write_tsv(tsv, urls, tsv_out_file)
|
||||
|
||||
except requests.HTTPError as e:
|
||||
print(e)
|
||||
|
||||
|
||||
@click.command()
|
||||
@click.argument('xls-file', type=click.Path(exists=True), required=True, nargs=1)
|
||||
@click.option('--purpose', type=click.Choice(['NERD', 'OCR'], case_sensitive=False), default="NERD",
|
||||
help="Purpose of output tsv file. "
|
||||
"\n\nNERD: NER/NED application/ground-truth creation. "
|
||||
"\n\nOCR: OCR application/ground-truth creation. "
|
||||
"\n\ndefault: NERD.")
|
||||
def make_page2tsv_commands(xls_file, purpose):
|
||||
|
||||
df = pd.read_excel(xls_file)
|
||||
|
||||
for _, row in df.iterrows():
|
||||
print('page2tsv $(OPTIONS) {}.xml {}.tsv --image-url={} --scale-factor={} --purpose={}'.
|
||||
format(row.Filename, row.Filename, row.iiif_url.replace('/full/full', '/left,top,width,height/full'),
|
||||
row.scale_factor, purpose))
|
||||
@ -0,0 +1,77 @@
|
||||
import os
|
||||
import requests
|
||||
import json
|
||||
|
||||
|
||||
def ned(tsv, ner_result, ned_rest_endpoint, json_file=None, threshold=None):
|
||||
|
||||
if json_file is not None and os.path.exists(json_file):
|
||||
|
||||
print('Loading {}'.format(json_file))
|
||||
|
||||
with open(json_file, "r") as fp:
|
||||
ned_result = json.load(fp)
|
||||
|
||||
else:
|
||||
|
||||
resp = requests.post(url=ned_rest_endpoint + '/parse', json=ner_result)
|
||||
|
||||
resp.raise_for_status()
|
||||
|
||||
ner_parsed = json.loads(resp.content)
|
||||
|
||||
ned_rest_endpoint = ned_rest_endpoint + '/ned?return_full=' + str(json_file is not None).lower()
|
||||
|
||||
resp = requests.post(url=ned_rest_endpoint, json=ner_parsed, timeout=3600000)
|
||||
|
||||
resp.raise_for_status()
|
||||
|
||||
ned_result = json.loads(resp.content)
|
||||
|
||||
rids = []
|
||||
entity = ""
|
||||
entity_type = None
|
||||
tsv['ID'] = '-'
|
||||
|
||||
def check_entity(tag):
|
||||
nonlocal entity, entity_type, rids
|
||||
|
||||
if (entity != "") and ((tag == 'O') or tag.startswith('B-') or (tag[2:] != entity_type)):
|
||||
|
||||
eid = entity + "-" + entity_type
|
||||
|
||||
if eid in ned_result:
|
||||
if 'ranking' in ned_result[eid]:
|
||||
ranking = ned_result[eid]['ranking']
|
||||
|
||||
#tsv.loc[rids, 'ID'] = ranking[0][1]['wikidata'] if threshold is None or ranking[0][1]['proba_1'] >= threshold else ''
|
||||
|
||||
tmp = "|".join([ranking[i][1]['wikidata']
|
||||
for i in range(len(ranking))
|
||||
if threshold is None or ranking[i][1]['proba_1'] >= threshold])
|
||||
tsv.loc[rids, 'ID'] = tmp if len(tmp) > 0 else '-'
|
||||
|
||||
rids = []
|
||||
entity = ""
|
||||
entity_type = None
|
||||
|
||||
ner_tmp = tsv.copy()
|
||||
ner_tmp.loc[~ner_tmp['NE-TAG'].isin(['O', 'B-PER', 'B-LOC','B-ORG', 'I-PER', 'I-LOC', 'I-ORG']), 'NE-TAG'] = 'O'
|
||||
|
||||
for rid, row in ner_tmp.iterrows():
|
||||
|
||||
check_entity(row['NE-TAG'])
|
||||
|
||||
if row['NE-TAG'] != 'O':
|
||||
|
||||
entity_type = row['NE-TAG'][2:]
|
||||
|
||||
entity += " " if entity != "" else ""
|
||||
|
||||
entity += str(row['TOKEN'])
|
||||
|
||||
rids.append(rid)
|
||||
|
||||
check_entity('O')
|
||||
|
||||
return tsv, ned_result
|
||||
@ -0,0 +1,49 @@
|
||||
import pandas as pd
|
||||
import requests
|
||||
import unicodedata
|
||||
import json
|
||||
|
||||
|
||||
def ner(tsv, ner_rest_endpoint):
|
||||
|
||||
resp = requests.post(url=ner_rest_endpoint, json={'text': " ".join(tsv.TOKEN.astype(str).tolist())})
|
||||
|
||||
resp.raise_for_status()
|
||||
|
||||
def iterate_ner_results(result_sentences):
|
||||
|
||||
for sen in result_sentences:
|
||||
|
||||
for token in sen:
|
||||
|
||||
yield unicodedata.normalize('NFC', token['word']), token['prediction'], False
|
||||
|
||||
yield '', '', True
|
||||
|
||||
ner_result = json.loads(resp.content)
|
||||
|
||||
result_sequence = iterate_ner_results(ner_result)
|
||||
|
||||
tsv_result = []
|
||||
for idx, row in tsv.iterrows():
|
||||
|
||||
row_token = unicodedata.normalize('NFC', str(row.TOKEN).replace(' ', ''))
|
||||
|
||||
ner_token_concat = ''
|
||||
while row_token != ner_token_concat:
|
||||
|
||||
ner_token, ner_tag, sentence_break = next(result_sequence)
|
||||
ner_token_concat += ner_token
|
||||
|
||||
assert len(row_token) >= len(ner_token_concat)
|
||||
|
||||
if sentence_break:
|
||||
tsv_result.append((0, '', 'O', 'O', '-', row.url_id, row.left, row.right, row.top, row.bottom))
|
||||
else:
|
||||
tsv_result.append((0, ner_token, ner_tag, 'O', '-', row.url_id, row.left, row.right, row.top,
|
||||
row.bottom))
|
||||
|
||||
return pd.DataFrame(tsv_result, columns=['No.', 'TOKEN', 'NE-TAG', 'NE-EMB', 'ID', 'url_id',
|
||||
'left', 'right', 'top', 'bottom']), ner_result
|
||||
|
||||
|
||||
@ -0,0 +1,84 @@
|
||||
import pandas as pd
|
||||
import re
|
||||
|
||||
|
||||
def read_tsv(tsv_file):
|
||||
|
||||
tsv = pd.read_csv(tsv_file, sep='\t', comment='#', quoting=3).rename(columns={'GND-ID': 'ID'})
|
||||
|
||||
parts = extract_doc_links(tsv_file)
|
||||
|
||||
urls = [part['url'] for part in parts]
|
||||
|
||||
return tsv, urls
|
||||
|
||||
|
||||
def write_tsv(tsv, urls, tsv_out_file):
|
||||
|
||||
out_columns = ['No.', 'TOKEN', 'NE-TAG', 'NE-EMB', 'ID', 'url_id', 'left', 'right', 'top', 'bottom']
|
||||
|
||||
if len(urls) == 0:
|
||||
print('Writing to {}...'.format(tsv_out_file))
|
||||
|
||||
tsv.to_csv(tsv_out_file, sep="\t", quoting=3, index=False)
|
||||
else:
|
||||
pd.DataFrame([], columns=out_columns).to_csv(tsv_out_file, sep="\t", quoting=3, index=False)
|
||||
|
||||
for url_id, part in tsv.groupby('url_id'):
|
||||
with open(tsv_out_file, 'a') as f:
|
||||
f.write('# ' + urls[url_id] + '\n')
|
||||
|
||||
part.to_csv(tsv_out_file, sep="\t", quoting=3, index=False, mode='a', header=False)
|
||||
|
||||
|
||||
def extract_doc_links(tsv_file):
|
||||
parts = []
|
||||
|
||||
header = None
|
||||
|
||||
with open(tsv_file, 'r') as f:
|
||||
|
||||
text = []
|
||||
url = None
|
||||
|
||||
for line in f:
|
||||
|
||||
if header is None:
|
||||
header = "\t".join(line.split()) + '\n'
|
||||
continue
|
||||
|
||||
urls = [url for url in
|
||||
re.findall(r'http[s]?://(?:[a-zA-Z]|[0-9]|[$-_@.&+]|[!*\(\),]|(?:%[0-9a-fA-F][0-9a-fA-F]))+', line)]
|
||||
|
||||
if len(urls) > 0:
|
||||
if url is not None:
|
||||
parts.append({"url": url, 'header': header, 'text': "".join(text)})
|
||||
text = []
|
||||
|
||||
url = urls[-1]
|
||||
else:
|
||||
if url is None:
|
||||
continue
|
||||
|
||||
line = '\t'.join(line.split())
|
||||
|
||||
if line.count('\t') == 2:
|
||||
line = "\t" + line
|
||||
|
||||
if line.count('\t') >= 3:
|
||||
text.append(line + '\n')
|
||||
|
||||
continue
|
||||
|
||||
if line.startswith('#'):
|
||||
continue
|
||||
|
||||
if len(line) == 0:
|
||||
continue
|
||||
|
||||
print('Line error: |', line, '|Number of Tabs: ', line.count('\t'))
|
||||
|
||||
if url is not None:
|
||||
parts.append({"url": url, 'header': header, 'text': "".join(text)})
|
||||
|
||||
return parts
|
||||
Loading…
Reference in New Issue