diff options
Diffstat (limited to 'fatcat_scholar')
| -rwxr-xr-x | fatcat_scholar/grobid2json.py | 201 | ||||
| -rw-r--r-- | fatcat_scholar/schema.py (renamed from fatcat_scholar/es_transform.py) | 30 | ||||
| -rw-r--r-- | fatcat_scholar/sim_pipeline.py | 12 | ||||
| -rw-r--r-- | fatcat_scholar/transform.py | 306 | ||||
| -rw-r--r-- | fatcat_scholar/work_pipeline.py | 17 |
5 files changed, 540 insertions, 26 deletions
diff --git a/fatcat_scholar/grobid2json.py b/fatcat_scholar/grobid2json.py new file mode 100755 index 0000000..9c2ffad --- /dev/null +++ b/fatcat_scholar/grobid2json.py @@ -0,0 +1,201 @@ +#!/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 + +xml_ns = "http://www.w3.org/XML/1998/namespace" +ns = "http://www.tei-c.org/ns/1.0" + +def all_authors(elem): + 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(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() + for on in ae.findall('./{%s}orgName' % ns): + affiliation[on.get('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): + 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): + ref = dict() + ref['id'] = elem.attrib.get('{http://www.w3.org/XML/1998/namespace}id') + # 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) + ref['publisher'] = elem.findtext('.//{%s}publicationStmt/{%s}publisher' % (ns, ns)) + if ref['publisher'] == '': + ref['publisher'] = None + date = elem.find('.//{%s}date[@type="published"]' % ns) + ref['date'] = (date != None) and date.attrib.get('when') + ref['volume'] = elem.findtext('.//{%s}biblScope[@unit="volume"]' % ns) + ref['issue'] = elem.findtext('.//{%s}biblScope[@unit="issue"]' % ns) + 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", "") + else: + ref['url'] = None + return ref + + +def teixml2json(content, encumbered=True): + + if type(content) == str: + content = io.StringIO(content) + elif type(content) == bytes: + content = io.BytesIO(content) + + info = dict() + + #print(content) + #print(content.getvalue()) + tree = ET.parse(content) + 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 != 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.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(): # 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') + print(json.dumps( + teixml2json(content, + encumbered=(not args.no_encumbered)), + sort_keys=True)) + +if __name__=='__main__': # pragma no cover + main() diff --git a/fatcat_scholar/es_transform.py b/fatcat_scholar/schema.py index 464b082..aa4ed52 100644 --- a/fatcat_scholar/es_transform.py +++ b/fatcat_scholar/schema.py @@ -6,23 +6,40 @@ auto-conversion of datetime objects. """ import ftfy -import typing import datetime from enum import Enum -from typing import Optional, List +from typing import Optional, List, Any from xml.etree import cElementTree as ET from pydantic import BaseModel + from fatcat_openapi_client import ReleaseEntity, ReleaseContrib +from fatcat_scholar.api_entities import entity_to_dict class DocType(str, Enum): work = "work" sim_page = "sim_page" +class IntermediateBundle(BaseModel): + doc_type: DocType + releases: List[ReleaseEntity] + biblio_release_ident: Optional[str] + grobid_fulltext: Optional[Any] + pdftotext_fulltext: Optional[Any] + sim_fulltext: Optional[Any] + + class Config: + arbitrary_types_allowed = True + json_encoders = { + ReleaseEntity: lambda re: entity_to_dict(re), + } + + class AccessType(str, Enum): ia_sim = "ia_sim" ia_file = "ia_file" wayback = "wayback" + web = "web" repository = "repository" paywall = "paywall" loginwall = "loginwall" @@ -30,7 +47,7 @@ class AccessType(str, Enum): class ScholarBiblio(BaseModel): release_ident: Optional[str] - title: str + title: Optional[str] subtitle: Optional[str] original_title: Optional[str] release_date: Optional[datetime.date] @@ -127,7 +144,7 @@ class ScholarAbstract(BaseModel): class ScholarAccess(BaseModel): access_type: AccessType access_url: str - mimetype: str + mimetype: Optional[str] file_ident: Optional[str] release_ident: Optional[str] @@ -139,8 +156,8 @@ class ScholarDoc(BaseModel): tags: List[str] = [] biblio: ScholarBiblio - fulltext: ScholarFulltext - ia_sim: ScholarSim + fulltext: Optional[ScholarFulltext] + ia_sim: Optional[ScholarSim] abstracts: List[ScholarAbstract] releases: List[ScholarRelease] access: List[ScholarAccess] @@ -187,6 +204,7 @@ def contrib_name(contrib: ReleaseContrib) -> str: return contrib.given_name def contrib_affiliation(contrib: ReleaseContrib) -> Optional[str]: + # TODO return None def es_abstracts_from_release(release: ReleaseEntity) -> List[ScholarAbstract]: diff --git a/fatcat_scholar/sim_pipeline.py b/fatcat_scholar/sim_pipeline.py index 1dd6476..4315e70 100644 --- a/fatcat_scholar/sim_pipeline.py +++ b/fatcat_scholar/sim_pipeline.py @@ -13,8 +13,7 @@ from fatcat_scholar.api_entities import * from fatcat_scholar.djvu import djvu_extract_leaf_texts from fatcat_scholar.sandcrawler import SandcrawlerPostgrestClient, SandcrawlerMinioClient from fatcat_scholar.issue_db import IssueDB, SimIssueRow -from fatcat_scholar.es_transform import es_biblio_from_release, es_release_from_release, DocType -from fatcat_scholar.work_pipeline import IntermediateBundle +from fatcat_scholar.schema import es_biblio_from_release, es_release_from_release, DocType, IntermediateBundle def truncate_pub_meta(full: Dict[str, Any]) -> Dict[str, Any]: @@ -52,8 +51,9 @@ class SimPipeline(): issue_item pages: str page_texts: list - page_number raw_text + page_num + leaf_num release_ident: Optional[str] pub_item_metadata issue_item_metadata @@ -107,6 +107,10 @@ class SimPipeline(): self.issue_db.db.row_factory = sqlite3.Row cur = self.issue_db.db.cursor() for row in cur.execute('SELECT * FROM sim_issue LEFT JOIN sim_pub ON sim_issue.sim_pubid = sim_pub.sim_pubid WHERE sim_issue.release_count < 3'): + # filter out "contents" and "index" items + # TODO: more filters; also redundant with IssueDB code? + if row['issue_item'].endswith('_contents') or row['issue_item'].endswith('_index'): + continue full_issue = self.fetch_sim_issue(row) if not full_issue: continue @@ -120,7 +124,7 @@ class SimPipeline(): sim_fulltext=dict( issue_item=full_issue['issue_item'], pages=str(leaf['page_num']), - page_texts=[leaf['raw_text']], + page_texts=[leaf], release_ident=None, pub_item_metadata=full_issue['pub_item_metadata'], issue_item_metadata=full_issue['issue_item_metadata'], diff --git a/fatcat_scholar/transform.py b/fatcat_scholar/transform.py new file mode 100644 index 0000000..54d3f71 --- /dev/null +++ b/fatcat_scholar/transform.py @@ -0,0 +1,306 @@ + +import os +import io +import sys +import argparse +from pydantic import BaseModel, validator +from typing import List, Dict, Tuple, Optional, Any, Sequence +from fatcat_openapi_client import ReleaseEntity, FileEntity +import internetarchive + +from fatcat_scholar.api_entities import * +from fatcat_scholar.djvu import djvu_extract_leaf_texts +from fatcat_scholar.sandcrawler import SandcrawlerPostgrestClient, SandcrawlerMinioClient +from fatcat_scholar.issue_db import IssueDB, SimIssueRow +from fatcat_scholar.schema import * +from fatcat_scholar.grobid2json import teixml2json + + +def es_fulltext_from_sim(sim: Dict[str, Any]) -> Optional[ScholarFulltext]: + if not sim['page_texts']: + return None + first_page = sim['page_texts'][0]['page_num'] + issue_item = sim['issue_item'] + return ScholarFulltext( + lang_code=None, # TODO: pub/issue metadata? or langdetect? + body="\n".join([p['raw_text'] for p in sim['page_texts']]), + #acknowledgement=None, + #annex=None, + release_ident=sim.get('release_ident'), + #file_ident=None, + #file_sha1=None, + #file_mimetype=None, + thumbnail_url=f"https://archive.org/serve/{issue_item}/__ia_thumb.jpg", + access_url=f"https://archive.org/details/{issue_item}/page/{first_page}", + access_type=AccessType.ia_sim, + ) + +def es_sim_from_sim(sim: Dict[str, Any]) -> ScholarSim: + first_page = None + if sim['page_texts']: + first_page = sim['page_texts'][0]['page_num'] + return ScholarSim( + issue_item=sim['issue_item'], + pub_collection=sim['pub_item_metadata']['metadata']['identifier'], + sim_pubid=sim['issue_item_metadata']['metadata']['sim_pubid'], + first_page=first_page, + ) + +SIM_RELEASE_TYPE_MAP = { + 'Scholarly Journals': 'article-journal', + # TODO: +} +SIM_LANG_MAP = { + 'English': 'en', + # TODO: +} +SIM_COUNTRY_MAP = { + 'Netherlands': 'nl', + # TODO: +} + +def es_biblio_from_sim(sim: Dict[str, Any]) -> ScholarBiblio: + + issue_meta = sim['issue_item_metadata']['metadata'] + pub_meta = sim['pub_item_metadata']['metadata'] + + first_page = None + if sim['page_texts']: + first_page = sim['page_texts'][0]['page_num'] + container_name = sim['pub_item_metadata']['metadata']['title'] + last_word = container_name.split()[-1] + if len(last_word) == 9 and last_word[4] == '-': + container_name = container_name[:-10] + + issns = [] + raw_issn = issue_meta.get('issn') + if raw_issn and len(raw_issn) == 9: + issns.append(raw_issn) + + volume = issue_meta.get('volume') + volume_int = None + if volume and volume.isdigit(): + volume_int = int(volume) + issue = issue_meta.get('issue') + issue_int = None + if issue and issue.isdigit(): + issue_int = int(issue) + + date = issue_meta.get('date') + release_year = None + if date and len(date) > 4 and date[:4].isdigit(): + release_year = int(date[:4]) + + release_date = None + if len(date) == len("2000-01-01"): + release_date = date + + return ScholarBiblio( + #release_ident=release.ident, + title=None, + #subtitle=None, + #original_title=release.original_title, + release_date=release_date, + release_year=release_year, + release_type=SIM_RELEASE_TYPE_MAP.get(pub_meta.get('pub_type')), + release_stage="published", # as a default + #withdrawn_status=release.withdrawn_status, + lang_code=SIM_LANG_MAP.get(pub_meta.get('language')), + country_code=SIM_COUNTRY_MAP.get(pub_meta.get('country')), + volume=volume, + volume_int=volume_int, + issue=issue, + issue_int=issue_int, + pages=sim.get('pages'), + first_page=first_page, + first_page_int=None, + #number=None, + + # no external identifiers + + #license_slug=release.license_slug, + publisher=issue_meta.get('publisher'), + container_name=container_name, + container_original_name=None, # TODO pass-through + container_ident=None, # TODO: pass-through + container_type=None, # TODO + container_issnl=None, # TODO: pass-through + issns=issns, + + # no contrib/affiliation info + contrib_names=[], + affiliations=[], + ) + +def _add_file_release_meta(fulltext: ScholarFulltext, re: ReleaseEntity, fe: FileEntity) -> ScholarFulltext: + best_url = None + best_url_type = None + for url in fe.urls: + best_url = url.url + best_url_type = AccessType.web + if '//archive.org/' in url.url: + best_url_type = AccessType.ia_file + break + elif '//web.archive.org/' in url.url: + best_url_type = AccessType.wayback + break + if url.rel == "repository": + best_url_type = AccessType.repository + # TODO: more file-to-access logic + + fulltext.release_ident = re.ident + fulltext.file_ident = fe.ident + fulltext.file_sha1 = fe.sha1 + fulltext.file_mimetype = fe.mimetype + fulltext.access_url = best_url + fulltext.access_type = best_url_type + return fulltext + + +def es_fulltext_from_grobid(tei_xml: str, re: ReleaseEntity, fe: FileEntity) -> Optional[ScholarFulltext]: + obj = teixml2json(tei_xml) + ret = ScholarFulltext( + lang_code=obj.get('lang'), + body=obj.get('body'), + acknowledgement=obj.get('acknowledgement'), + annex=obj.get('annex'), + thumbnail_url=None, # TODO: sandcrawler thumbnails + ) + return _add_file_release_meta(ret, re, fe) + +def es_fulltext_from_pdftotext(pdftotext: Any, re: ReleaseEntity, fe: FileEntity) -> Optional[ScholarFulltext]: + + ret = ScholarFulltext( + lang_code=re.language, + body=pdftotext['raw_text'], + acknowledgement=None, + annex=None, + thumbnail_url=None, # TODO: sandcrawler thumbnails + ) + return _add_file_release_meta(ret, re, fe) + +def transform_heavy(heavy: IntermediateBundle) -> Optional[ScholarDoc]: + + tags: List[str] = [] + work_ident: Optional[str] = None + abstracts: List[ScholarAbstract] = [] + fulltext: Optional[ScholarFulltext] + + ia_sim: Optional[ScholarSim] = None + if heavy.sim_fulltext is not None: + ia_sim = es_sim_from_sim(heavy.sim_fulltext) + + if heavy.doc_type == DocType.sim_page: + assert ia_sim is not None + key = f"page_{ia_sim.issue_item}_{ia_sim.first_page}" + biblio = es_biblio_from_sim(heavy.sim_fulltext) + fulltext = es_fulltext_from_sim(heavy.sim_fulltext) + elif heavy.doc_type == DocType.work: + work_ident = heavy.releases[0].work_id + key = f"work_{work_ident}" + assert heavy.biblio_release_ident + primary_release = [r for r in heavy.releases if r.ident == heavy.biblio_release_ident][0] + biblio = es_biblio_from_release(primary_release) + + # TODO: abstracts from releases also? abstracts_dict? + abstracts = es_abstracts_from_release(primary_release) + else: + raise NotImplementedError(f"doc_type: {heavy.doc_type}") + + if heavy.grobid_fulltext: + + fulltext_release = [r for r in heavy.releases if r.ident == heavy.grobid_fulltext['release_ident']][0] + fulltext_file = [f for f in fulltext_release.files if f.ident == heavy.grobid_fulltext['file_ident']][0] + + fulltext = es_fulltext_from_grobid(heavy.grobid_fulltext['tei_xml'], fulltext_release, fulltext_file) + + # hack to pull through thumbnail from local pdftotext + if fulltext and not fulltext.thumbnail_url and heavy.pdftotext_fulltext: + fulltext.thumbnail_url = f"https://covid19.fatcat.wiki/sha1/{fulltext_file.sha1}" # XXX + + if not fulltext and heavy.pdftotext_fulltext: + + fulltext_release = [r for r in heavy.releases if r.ident == heavy.grobid_fulltext['release_ident']][0] + fulltext_file = [f for f in fulltext_release.files if f.ident == heavy.grobid_fulltext['file_ident']][0] + fulltext = es_fulltext_from_pdftotext(heavy.pdftotext_fulltext, fulltext_release, fulltext_file) + + # TODO: additional access list + access_dict = dict() + if fulltext and fulltext.access_type: + access_dict[fulltext.access_type] = ScholarAccess( + access_type=fulltext.access_type, + access_url=fulltext.access_url, + mimetype=fulltext.file_mimetype, + file_ident=fulltext.file_ident, + release_ident=fulltext.release_ident, + ) + if ia_sim and not AccessType.ia_sim in access_dict: + access_dict[AccessType.ia_sim] = ScholarAccess( + access_type=AccessType.ia_sim, + access_url=f"https://archive.org/details/{ia_sim.issue_item}/page/{ia_sim.first_page}", + ) + + # TODO: additional abstracts + + return ScholarDoc( + key=key, + doc_type=heavy.doc_type.value, + doc_index_ts=datetime.datetime.utcnow(), + work_ident=work_ident, + tags=tags, + + biblio=biblio, + fulltext=fulltext, + ia_sim=ia_sim, + abstracts=abstracts, + releases=[es_release_from_release(r) for r in heavy.releases], + access=list(access_dict.values()), + ) + +def run_transform(infile): + for line in infile: + obj = json.loads(line) + + heavy = IntermediateBundle( + doc_type=DocType(obj['doc_type']), + releases=[entity_from_json(json.dumps(re), ReleaseEntity) for re in obj['releases']], + biblio_release_ident=obj.get('biblio_release_ident'), + grobid_fulltext=obj.get('grobid_fulltext'), + pdftotext_fulltext=obj.get('pdftotext_fulltext'), + sim_fulltext=obj.get('sim_fulltext'), + ) + es_doc = transform_heavy(heavy) + if not es_doc: + continue + print(es_doc.json()) + +def main(): + """ + Run this command like: + + python -m fatcat_scholar.transform + """ + + parser = argparse.ArgumentParser( + formatter_class=argparse.ArgumentDefaultsHelpFormatter) + subparsers = parser.add_subparsers() + + sub = subparsers.add_parser('run_transform', + help="iterates through 'heavy' intermediate") + sub.set_defaults(func='run_transform') + sub.add_argument("json_file", + help="intermediate globs as JSON-lines", + nargs='?', default=sys.stdin, type=argparse.FileType('r')) + + args = parser.parse_args() + if not args.__dict__.get("func"): + print("tell me what to do! (try --help)") + sys.exit(-1) + + if args.func == 'run_transform': + run_transform(infile=args.json_file) + else: + raise NotImplementedError(args.func) + +if __name__=="__main__": + main() diff --git a/fatcat_scholar/work_pipeline.py b/fatcat_scholar/work_pipeline.py index b9dcbe8..081878c 100644 --- a/fatcat_scholar/work_pipeline.py +++ b/fatcat_scholar/work_pipeline.py @@ -12,7 +12,7 @@ from fatcat_scholar.api_entities import * from fatcat_scholar.djvu import djvu_extract_leaf_texts from fatcat_scholar.sandcrawler import SandcrawlerPostgrestClient, SandcrawlerMinioClient from fatcat_scholar.issue_db import IssueDB, SimIssueRow -from fatcat_scholar.es_transform import es_biblio_from_release, es_release_from_release, DocType +from fatcat_scholar.schema import es_biblio_from_release, es_release_from_release, DocType, IntermediateBundle def parse_pages(raw: str) -> Tuple[Optional[int], Optional[int]]: @@ -44,21 +44,6 @@ def test_parse_pages(): assert parse_pages("iiv") == (None, None) -class IntermediateBundle(BaseModel): - doc_type: DocType - releases: List[ReleaseEntity] - biblio_release_ident: Optional[str] - grobid_fulltext: Optional[Any] - pdftotext_fulltext: Optional[Any] - sim_fulltext: Optional[Any] - - class Config: - arbitrary_types_allowed = True - json_encoders = { - ReleaseEntity: lambda re: entity_to_dict(re), - } - - def fulltext_pref_list(releases: List[ReleaseEntity]) -> List[str]: """ Returns a list of release idents in preference order (best first) to |