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
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
|
#!/usr/bin/env python3
"""
NB: adapted to work as a library for PDF extraction. Will probably be
re-written eventually to be correct, complete, and robust; this is just a
first iteration.
This script tries to extract everything from a GROBID TEI XML fulltext dump:
- header metadata
- affiliations
- references (with context)
- abstract
- fulltext
- tables, figures, equations
A flag can be specified to disable copyright encumbered bits (--no-emcumbered):
- abstract
- fulltext
- tables, figures, equations
Prints JSON to stdout, errors to stderr
This file copied from the sandcrawler repository.
"""
import io
import json
import argparse
import xml.etree.ElementTree as ET
from typing import List, Any, Dict, AnyStr, Optional
xml_ns = "http://www.w3.org/XML/1998/namespace"
ns = "http://www.tei-c.org/ns/1.0"
def all_authors(elem: Optional[ET.Element], ns: str = ns) -> List[Dict[str, Any]]:
if not elem:
return []
names = []
for author in elem.findall(".//{%s}author" % ns):
pn = author.find("./{%s}persName" % ns)
if not pn:
continue
given_name = pn.findtext("./{%s}forename" % ns) or None
surname = pn.findtext("./{%s}surname" % ns) or None
full_name = " ".join(pn.itertext())
obj: Dict[str, Any] = dict(name=full_name)
if given_name:
obj["given_name"] = given_name
if surname:
obj["surname"] = surname
ae = author.find("./{%s}affiliation" % ns)
if ae:
affiliation: Dict[str, Any] = dict()
for on in ae.findall("./{%s}orgName" % ns):
on_type = on.get("type")
if on_type:
affiliation[on_type] = on.text
addr_e = ae.find("./{%s}address" % ns)
if addr_e:
address = dict()
for t in addr_e.getchildren():
address[t.tag.split("}")[-1]] = t.text
if address:
affiliation["address"] = address
# affiliation['address'] = {
# 'post_code': addr.findtext('./{%s}postCode' % ns) or None,
# 'settlement': addr.findtext('./{%s}settlement' % ns) or None,
# 'country': addr.findtext('./{%s}country' % ns) or None,
# }
obj["affiliation"] = affiliation
names.append(obj)
return names
def journal_info(elem: ET.Element) -> Dict[str, Any]:
journal = dict()
journal["name"] = elem.findtext(".//{%s}monogr/{%s}title" % (ns, ns))
journal["publisher"] = elem.findtext(
".//{%s}publicationStmt/{%s}publisher" % (ns, ns)
)
if journal["publisher"] == "":
journal["publisher"] = None
journal["issn"] = elem.findtext('.//{%s}idno[@type="ISSN"]' % ns)
journal["eissn"] = elem.findtext('.//{%s}idno[@type="eISSN"]' % ns)
journal["volume"] = elem.findtext('.//{%s}biblScope[@unit="volume"]' % ns)
journal["issue"] = elem.findtext('.//{%s}biblScope[@unit="issue"]' % ns)
keys = list(journal.keys())
# remove empty/null keys
for k in keys:
if not journal[k]:
journal.pop(k)
return journal
def biblio_info(elem: ET.Element, ns: str = ns) -> Dict[str, Any]:
ref: Dict[str, Any] = dict()
ref["id"] = elem.attrib.get("{http://www.w3.org/XML/1998/namespace}id")
ref["unstructured"] = elem.findtext('.//{%s}note[@type="raw_reference"]' % ns)
# Title stuff is messy in references...
ref["title"] = elem.findtext(".//{%s}analytic/{%s}title" % (ns, ns))
other_title = elem.findtext(".//{%s}monogr/{%s}title" % (ns, ns))
if other_title:
if ref["title"]:
ref["journal"] = other_title
else:
ref["journal"] = None
ref["title"] = other_title
ref["authors"] = all_authors(elem, ns=ns)
ref["publisher"] = elem.findtext(".//{%s}publicationStmt/{%s}publisher" % (ns, ns))
if not ref["publisher"]:
ref["publisher"] = elem.findtext(".//{%s}imprint/{%s}publisher" % (ns, ns))
if ref["publisher"] == "":
ref["publisher"] = None
date = elem.find('.//{%s}date[@type="published"]' % ns)
ref["date"] = (date is not None) and date.attrib.get("when")
ref["volume"] = elem.findtext('.//{%s}biblScope[@unit="volume"]' % ns)
ref["issue"] = elem.findtext('.//{%s}biblScope[@unit="issue"]' % ns)
ref["doi"] = elem.findtext('.//{%s}idno[@type="DOI"]' % ns)
ref["arxiv_id"] = elem.findtext('.//{%s}idno[@type="arXiv"]' % ns)
if ref["arxiv_id"] and ref["arxiv_id"].startswith("arXiv:"):
ref["arxiv_id"] = ref["arxiv_id"][6:]
ref["pmcid"] = elem.findtext('.//{%s}idno[@type="PMCID"]' % ns)
ref["pmid"] = elem.findtext('.//{%s}idno[@type="PMID"]' % ns)
el = elem.find('.//{%s}biblScope[@unit="page"]' % ns)
if el is not None:
if el.attrib.get("from") and el.attrib.get("to"):
ref["pages"] = "{}-{}".format(el.attrib["from"], el.attrib["to"])
else:
ref["pages"] = el.text
el = elem.find(".//{%s}ptr[@target]" % ns)
if el is not None:
ref["url"] = el.attrib["target"]
# Hand correction
if ref["url"].endswith(".Lastaccessed"):
ref["url"] = ref["url"].replace(".Lastaccessed", "")
if ref["url"].startswith("<"):
ref["url"] = ref["url"][1:]
if ">" in ref["url"]:
ref["url"] = ref["url"].split(">")[0]
else:
ref["url"] = None
return ref
def teixml2json(content: AnyStr, encumbered: bool = True) -> Dict[str, Any]:
if isinstance(content, str):
tree = ET.parse(io.StringIO(content))
elif isinstance(content, bytes):
tree = ET.parse(io.BytesIO(content))
info: Dict[str, Any] = dict()
# print(content)
# print(content.getvalue())
tei = tree.getroot()
header = tei.find(".//{%s}teiHeader" % ns)
if header is None:
raise ValueError("XML does not look like TEI format")
application_tag = header.findall(".//{%s}appInfo/{%s}application" % (ns, ns))[0]
info["grobid_version"] = application_tag.attrib["version"].strip()
info["grobid_timestamp"] = application_tag.attrib["when"].strip()
info["title"] = header.findtext(".//{%s}analytic/{%s}title" % (ns, ns))
info["authors"] = all_authors(
header.find(".//{%s}sourceDesc/{%s}biblStruct" % (ns, ns))
)
info["journal"] = journal_info(header)
date = header.find('.//{%s}date[@type="published"]' % ns)
info["date"] = (date is not None) and date.attrib.get("when")
info["fatcat_release"] = header.findtext('.//{%s}idno[@type="fatcat"]' % ns)
info["doi"] = header.findtext('.//{%s}idno[@type="DOI"]' % ns)
if info["doi"]:
info["doi"] = info["doi"].lower()
refs = []
for (i, bs) in enumerate(tei.findall(".//{%s}listBibl/{%s}biblStruct" % (ns, ns))):
ref = biblio_info(bs)
ref["index"] = i
refs.append(ref)
info["citations"] = refs
text = tei.find(".//{%s}text" % (ns))
# print(text.attrib)
if text and text.attrib.get("{%s}lang" % xml_ns):
info["language_code"] = text.attrib["{%s}lang" % xml_ns] # xml:lang
if encumbered:
el = tei.find(".//{%s}profileDesc/{%s}abstract" % (ns, ns))
info["abstract"] = (el or None) and " ".join(el.itertext()).strip()
el = tei.find(".//{%s}text/{%s}body" % (ns, ns))
info["body"] = (el or None) and " ".join(el.itertext()).strip()
el = tei.find('.//{%s}back/{%s}div[@type="acknowledgement"]' % (ns, ns))
info["acknowledgement"] = (el or None) and " ".join(el.itertext()).strip()
el = tei.find('.//{%s}back/{%s}div[@type="annex"]' % (ns, ns))
info["annex"] = (el or None) and " ".join(el.itertext()).strip()
# remove empty/null keys
keys = list(info.keys())
for k in keys:
if not info[k]:
info.pop(k)
return info
def main() -> None: # pragma no cover
parser = argparse.ArgumentParser(
formatter_class=argparse.ArgumentDefaultsHelpFormatter,
description="GROBID TEI XML to JSON",
usage="%(prog)s [options] <teifile>...",
)
parser.add_argument(
"--no-encumbered",
action="store_true",
help="don't include ambiguously copyright encumbered fields (eg, abstract, body)",
)
parser.add_argument("teifiles", nargs="+")
args = parser.parse_args()
for filename in args.teifiles:
content = open(filename, "r").read()
print(
json.dumps(
teixml2json(content, encumbered=(not args.no_encumbered)),
sort_keys=True,
)
)
if __name__ == "__main__": # pragma no cover
main()
|
