def calculate_carbon(pub):
''' Calculate the carbon emissions from travel
:publication (article) from database
'''
author_count = pub['author count']
pa_print.tprint('\nCalculating carbon footprint...')
for author in range(author_count):
if pub['author location info'][author] != 'N/A':
distance = geodesic(pub['author location info'][author][2],
pub['conference location info'][0][2]).km
pub['author distances'].append(distance)
# * Calculate C02 emissions, more details here: https://github.com/milankl/CarbonFootprintAGU
carbon = 0.0 # kgCO2e
if distance < 400: # bus / train / car at 60gCO2e / km / person
carbon = distance * 2 * 0.06
elif distance < 1500: # short flight at 200gCO2e / km / person
carbon = distance * 2 * 0.2
elif distance < 8000: # long flight at 250gCO2e / km / person
carbon = distance * 2 * 0.25
else: # super long flight at 300gCO2e / km / person
carbon = distance * 2 * 0.3
pub['author footprints'].append(carbon / 1000)
pa_print.tprint(
f'✓ - CO2 emissions for author {int(author + 1)}: {(carbon / 1000):.3f} tCO2e'
)
else:
pub['author distances'].append('N/A')
pub['author footprints'].append('N/A')
def doc_check(doc, pub, type):
''' Check for common decoding errors (does not catch all) # ! more intelligent method?
:document from text extraction (miner) or xml extraction (grobid)
:publication (article) from database
:type of doc (either 'text' or 'grobid')
'''
errored = False
alphas = re.compile('[^a-zA-Z]')
doc_alphas = alphas.sub('', doc)
if len(doc) > 2 * len(doc_alphas): # more symbols than 2x letters
pub[f'{type} non alpha'] = 'X'
pa_print.tprint('\nFile was not decoded well - non-alpha')
errored = True
cids = re.compile(r'\(cid:[0-9]+\)')
doc_cidless = cids.sub(
'', doc,
re.M) # when font cannot be decoded, (cid:#) is returned, remove these
if len(doc) > 2 * len(
doc_cidless): # if most of content was undecodable, skip
pub[f'{type} poor decoding'] = 'X'
pa_print.tprint('\nFile was not decoded well - cid: present')
errored = True
return errored
def request_location(author_info, args, pub):
''' Extracts location from author blocks or universities and queries OpenCageGeocode
:publication from bibtex file
'''
author_count = pub['author count']
# Conference location lookup
cnf_query = pub['address']
query_type = 'conference'
query_location(cnf_query, query_type,
pub) # *** creates unneeded columns ***
# Author location lookup
for author in range(author_count): # length of usable locations
query_type = 'author'
# Assign one query (in order of priority)
# 1) If there is a university address from grobid
if pub['grobid author unis'][author] != 'N/A': # uni address
location_query = ', '.join(
pub['grobid author unis'][author]) # (uni name, country)
query_origin = 'grobid uni'
# 2) If grobid was used to add address (while 'location' is api derived)
elif pub['grobid addresses'][author] != 'N/A':
location_query = pub['grobid addresses'][author]
query_origin = 'grobid address'
# 3) If theres a uni address from text block
elif pub['text author unis'][author] != 'N/A':
location_query = ', '.join(
pub['text author unis'][author]) # (uni name, country)
query_origin = 'text uni'
# 4) Else, scrape from raw author block (which may or may not have email)
elif author < len(author_info) and author_info[
author] != 'N/A': # check if author_info contains author 'i' and is non-empty
auth_block = author_info[author]
cut_line = -1 if '@' in auth_block else 0 # one line above if email present
info_lines = auth_block.split('\n')
location_query = ' '.join(info_lines[cut_line - 1:cut_line])
if len([line for line in location_query if line.isdigit()
]) > 8: # look for tele #
location_query = ' '.join(
info_lines[cut_line - 2:cut_line -
1]) # take line higher if telephone
query_origin = 'raw author block'
else:
location_query = 'N/A'
query_origin = 'No query'
pa_print.tprint("\nCouldn't find a location to use!")
pa_print.tprint(f'\nLooking for: {location_query}')
pub['author loc queries'].append(location_query)
pub['author query origins'].append(query_origin)
query_location(location_query, query_type, pub)
def download_xml(xml_path, pub):
pa_print.tprint('\nLocal PubPub XML not found - downloading...')
url = pub['url']
r = requests.get(url, allow_redirects=True)
url = re.search(r"jats","url":"(.*?.xml)",
r.text).group(1)
r = requests.get(url, allow_redirects=True)
open(jats_src, 'wb').write(r.content)
def load_unidomains(path):
''' Loads unidomain file from json or downloads if not found
:path of unisomains.json file
'''
if not os.path.isfile(path): # if not, download
pa_print.tprint('\nDownloading unidomains database...')
r = requests.get(unidomains_url, allow_redirects=True)
open(path, 'wb').write(r.content)
with open(path, 'rb') as fp:
unidomains = orjson.loads(fp.read())
return unidomains
def load_bibtex(path):
''' Loads BibTeX file into object or downloads if not found
:path of BibTeX file
'''
if not os.path.isfile(path): # if not, download
pa_print.tprint('\nDownloading bibtex database...')
r = requests.get(bibtex_url, allow_redirects=True)
open(path, 'wb').write(r.content)
with open(path) as bib_file:
parser = bibtexparser.bparser.BibTexParser()
parser.customization = bibtexparser.customization.convert_to_unicode
bib_db = bibtexparser.load(bib_file, parser=parser)
bib_db = bib_db.entries
return bib_db
def import_config(filepath):
''' Imports a custom configuration for filter words and years
:filepath the file path
'''
user_config = pd.read_csv(filepath, header=0, delimiter=',')
user_config = user_config.fillna('')
keywords = []
ignore_words = []
merge_words = []
selected_years = []
for config_tuple in user_config.itertuples(index=False):
if config_tuple[0] == 'keywords': # single list
for i in config_tuple[1:]:
keywords.append(i)
elif config_tuple[0] == 'ignore': # single list
for i in config_tuple[1:]:
ignore_words.append(i)
elif config_tuple[0] == 'merge': # list of lists
merge_group = list(filter(None, config_tuple[1:]))
merge_words.append(merge_group)
elif config_tuple[0] == 'years': # single list
year_num = [i for i in config_tuple if i != '']
if len(year_num) == 2:
selected_years.append(str(int(config_tuple[1])))
else:
year_span = int(config_tuple[2]) - int(config_tuple[1])
for i in range(year_span + 1):
selected_years.append(str(int(config_tuple[1]) + i))
keywords = list(filter(None, keywords))
ignore_words = list(filter(None, ignore_words))
pa_print.tprint('\nParameters from custom.csv:')
if selected_years:
pa_print.tprint(f'Selected years: {selected_years}')
if keywords:
pa_print.tprint(f'Search words: {keywords}')
if ignore_words:
pa_print.tprint(f'Ignored words: {ignore_words}')
if merge_words:
pa_print.tprint(f'Merged words: {merge_words}')
return (keywords, ignore_words, merge_words, selected_years)
def trim_headfoot(doc, pub=None):
''' Trim the header and footer from extracted text (unused and inferior to Grobid service)
:document from text extraction (miner) or xml extraction (grobid)
'''
# Function for trimming header and footer
# Remove until abstract or introduction
pdf_trimmed = abst_regex.split(doc, 1)
if len(pdf_trimmed) == 1:
pdf_trimmed = intro_regex.split(
pdf_trimmed[0], 1) # if no abstract, use 'introduction'
if len(pdf_trimmed) == 1:
pdf_trimmed = pdf_trimmed[0]
if pub is not None: pub['header fail'] = 'X'
pa_print.tprint('Could not split header during parsing!')
else:
pdf_trimmed = pdf_trimmed[1]
# pa_print.tprint('Split header at intro')
else:
pdf_trimmed = pdf_trimmed[1]
# pa_print.tprint('Split header at abstract')
# return pdf_trimmed
# Remove after references or acknowledgements
pdf_slimmed = ackn_regex.split(pdf_trimmed, 1)
if len(pdf_slimmed) == 1:
pdf_slimmed = ref_regex.split(pdf_slimmed[0], 1)
if len(pdf_slimmed) == 1:
if pub is not None: pub['footer fail'] = 'X'
pa_print.tprint('Could not split footer during parsing!')
else:
pdf_slimmed = pdf_slimmed[0]
# pa_print.tprint('Split footer at references')
else:
pdf_slimmed = pdf_slimmed[0]
# pa_print.tprint('Split footer at acknowledgements')
return pdf_slimmed
def extract_text(pub):
'''Extracts text content from pdf using pdfminer.six, downloads pdf if non-existant
:publication (article) from database
'''
pdf_fn = pub['url'].split('/')[-1]
pdf_path = pdf_src + pdf_fn
# Allows for override of corrupted pdfs
if os.path.isfile(pdf_path):
pass
else: # doesnt exist - download
download_pdf(pdf_path, pub)
# Page count for those without
if pub['page count'] == 'N/A':
pdf = open(pdf_path, 'rb')
check = False
while True: # try once
try:
parser = PDFParser(pdf)
document = PDFDocument(parser)
except Exception as e:
if check is True:
raise PSSyntaxError(
f'{pdf_path} appears to be malformed and qpdf cannot repair it.'
)
pa_print.tprint(str(e))
pa_print.tprint(f'Attempting to repair {pdf_path}')
pike = pikepdf.Pdf.open(pdf_path, allow_overwriting_input=True)
pike.save(pdf_path)
check = True
continue
break
pub['page count'] = resolve1(document.catalog['Pages'])['Count']
fn = pdf_fn.split('.')[0]
miner_text_file = f'{text_src}miner/miner_{fn}.txt'
# Read miner text if exists
if os.path.isfile(miner_text_file):
with open(miner_text_file, 'r') as f:
doc = f.read()
return doc
else: # if not, make them
pa_print.tprint(f'\nExtracting: {pdf_fn}')
laparams = LAParams()
setattr(laparams, 'all_texts', True)
doc = extract_pdf(pdf_path, laparams=laparams)
with open(miner_text_file, 'w') as f:
f.write(doc)
return doc
def request_scholar(pub, args):
''' Queries citations from Semantic Scholar
:publication from bibtex file
'''
try:
with open('./cache/json/scholar_cache.json', 'rb') as fp:
scholar_cache = orjson.loads(fp.read())
except FileNotFoundError:
pa_print.tprint('\nCreating new Semantic Scholar cache!')
scholar_cache = {}
semantic_scholar_data = {
"queryString": [],
"page": 1,
"pageSize": 1,
"sort": "relevance",
"authors": [],
"coAuthors": [],
"venues": [],
"yearFilter": None,
"requireViewablePdf": False,
"publicationTypes": [],
"externalContentTypes": []
}
# Fix names for searching
regextitle = re.compile(r'[^a-zA-Z0-9 ]')
regexname = re.compile(r'[^a-zA-Z- ]')
author_last_list = []
for _, (_, last) in enumerate(pub['author names']):
last = last.split('-')[-1]
author_last_list.append(last)
title = unidecode.unidecode(pub['title'])
if args.nime:
if title == 'Now': # title is too short, this return other paper, trying to filter it out by forcing full author name
author_last_list[0] = 'GarthPaine'
pub['citation count'] = 'N/A'
pub['key citation count'] = 'N/A'
# Make query title, name and year lists
query_title = list(
dict.fromkeys([
title,
regextitle.sub('', title),
' '.join([w for w in title.split() if len(w) > 1])
]))
if len(author_last_list) > 1:
query_name = [' '.join(author_last_list), author_last_list[0], '']
else:
query_name = [author_last_list[0], '']
query_year = ['', pub['year']]
# Save query to be used for cache
full_query = f"{title} {' '.join(author_last_list)} {pub['year']}"
pub['scholar query'] = full_query
if full_query not in scholar_cache or args.citations:
pa_print.tprint(f'\nQuerying Semantic Scholar...')
for temp in list(itertools.product(query_title, query_name,
query_year)):
# Generate new query from combination
temp_title, temp_author, temp_year = temp[0], temp[1], temp[2]
scholar_query = f'{temp_title} {temp_author} {temp_year}'
semantic_scholar_data['queryString'] = scholar_query
# Try query
pa_print.tprint(f"Trying query: '{scholar_query}'")
try:
query_result = scholar_api(semantic_scholar_data)
except Exception as e:
query_result = {'results': {}}
err_info = 'x - While querying Semantic Scholar an exception of type {0} occurred.\nArguments:\n{1!r}.'
err_msg = err_info.format(type(e).__name__, e.args)
pa_print.tprint(err_msg)
if not 'error' in query_result.keys():
if bool(query_result['results']) and \
bool(query_result['results'][0]['scorecardStats']) and \
len(query_result['results'][0]['authors']) <= (len(author_last_list) + 1):
result_author = ' '.join([
t[0]['name']
for t in query_result['results'][0]['authors']
])
result_author = regexname.sub(
'', unidecode.unidecode(result_author)).lower()
query_author = regexname.sub(
'', author_last_list[0].lower().split(' ')[-1])
if result_author.find(query_author) != -1:
pub['scholar query'] = scholar_query
pub['citation count'] = query_result['results'][0][
'scorecardStats'][0]['citationCount']
pub['key citation count'] = query_result['results'][0][
'scorecardStats'][0]['keyCitationCount']
scholar_cache[full_query] = query_result['results'][0][
'scorecardStats']
pa_print.tprint(
f"✓ - Paper has been cited {pub['citation count']} times"
)
break
if pub['citation count'] == 'N/A':
pa_print.tprint(
'x - Cannot find citations for paper in Semantic Scholar')
scholar_cache[full_query] = 'N/A'
with open('./cache/json/scholar_cache.json', 'wb') as fp:
fp.write(orjson.dumps(scholar_cache))
else:
if scholar_cache[full_query] != 'N/A':
pub['citation count'] = scholar_cache[full_query][0][
'citationCount']
pub['key citation count'] = scholar_cache[full_query][0][
'keyCitationCount']
else:
pub['citation count'] = 'N/A'
pub['key citation count'] = 'N/A'
pa_print.tprint(
f"\no - Retrieved from cache: {pub['citation count']} citations")
# Average citations per year of age
if pub['citation count'] != 'N/A':
pub['yearly citations'] = int(pub['citation count']) / pub['age']
else:
pub['yearly citations'] = 'N/A'
def request_uni(unidomains, author_info, args, pub):
''' Extract university from email handle
:publication from bibtex file
'''
pub_matches = 0
grob_matches = 0
text_matches = 0
author_count = pub['author count']
# Internal functions for lookup in unidomains.json
def lookup_uni(handle, email_type, pub):
nonlocal pub_matches
for uni in unidomains:
if handle in uni['domains']:
pub[f'{email_type} author unis'].append(
(uni['name'], uni['country']))
pub_matches += 1
uni_match = True
break
def handle_check(email, email_type, pub):
handle = email.split("@")[-1].strip()
# Look for handle in json, split once by dot and retry if not found
uni_match = False
lookup_uni(handle, email_type, pub)
while uni_match == False and handle.count('.') > 1:
handle = handle.split('.', 1)[-1]
lookup_uni(handle, email_type, pub)
# 1) Using grobid derived emails to choose handle
email_type = 'grobid'
for author in range(author_count):
email = pub['grobid emails'][author]
if email != 'N/A': # check for valid email
handle_check(email, email_type, pub)
grob_matches = pub_matches
# 2) Using scraped author info block from header if not enough emails
if len(author_info) > 0 and (grob_matches < author_count):
email_type = 'text'
for author in author_info: # ! could be more authors than exit
info_emails = email_regex.findall(
author) # look for '@handle.tld' in block
for _, email in enumerate(
info_emails
): # case: multiple emails are within an author block #! (will overwrite)
if email != 'N/A':
handle_check(email, email_type, pub)
# Fill in missing unis with 'N/A' # ! author block not linked in order with authors
for type, author in [(type, author) for type in ['grobid', 'text']
for author in range(author_count)]:
try:
pub[f'{type} author unis'][author]
except IndexError:
pub[f'{type} author unis'].append('N/A')
text_matches = pub_matches - grob_matches
pub_matches = max(text_matches, grob_matches)
pa_print.tprint(f'o - Found {pub_matches} uni\'s from email handles\n')
def query_location(location_query, query_type,
pub): # 'query_type is now only used to print status
# Load cache
try:
with open('./cache/json/location_cache.json', 'rb') as fp:
location_cache = orjson.loads(fp.read())
except FileNotFoundError:
pa_print.tprint('\nCreating new location cache!')
location_cache = {'N/A': 'N/A'}
# Not cached
if location_query not in location_cache:
try:
# location = geolocator.geocode(location_query, language="en") # Nominatim fallback
# OpenCageGeocode: 2,500 req/day, 1 req/s - https://github.com/OpenCageData/python-opencage-geocoder
location = geocoder.geocode(location_query,
language='en',
limit=1,
no_annotations=1,
no_record=1)[0]
# Format result
geometry = location['geometry'] # lat/long
components = location['components'] # fine loc info
location_info = (location['formatted'], (components['country'],
components['continent']),
(geometry['lat'],
geometry['lng']), location['confidence']
) # 1 (>25km) to 10 (<0.25km)
location_cache[location_query] = location_info
pub[f'{query_type} location info'].append(
location_info[:3]) # add all location into one column
pub[f'{query_type} location confidence'].append(
location_info[3]) # confidence in separate column
pa_print.tprint(
f'✓ - Parsed {query_type} location: {location_info[0]}')
time.sleep(1 + random.random())
except: # API fails
location_cache[location_query] = 'N/A'
pub[f'{query_type} location info'].append('N/A')
pub[f'{query_type} location confidence'].append('N/A')
pa_print.tprint(
f'x - Could not parse {query_type} location: {location_query}')
# Save changes to cache
with open('./cache/json/location_cache.json', 'wb') as fp:
fp.write(orjson.dumps(location_cache))
# Cached
else:
if location_cache[location_query] != 'N/A' and not (location_query
== 'N/A'):
location_info = location_cache[location_query]
pub[f'{query_type} location info'].append(location_info[:3])
pub[f'{query_type} location confidence'].append(location_info[3])
pa_print.tprint(
f'o - Cached {query_type} location: {location_info[0]}')
else:
location_info = 'N/A'
pub[f'{query_type} location info'].append('N/A')
pub[f'{query_type} location confidence'].append('N/A')
pa_print.tprint(f'o - Null {query_type} location: {location_info}')
def extract_author_info(doc, pub):
''' Searches through pdf text for author block using regex (no Grobid needed)
:document from text extraction (miner) or xml extraction (grobid)
:publication (article) from database
'''
pa_print.tprint('\nExtracting authors from paper...')
author_info = []
author_count = pub['author count']
# * Method 1 - Look for block with email tail (bibtex not needed, more robust)
author_info = auth_regex.findall(
doc)[:author_count] # grab only up to total authors
if len(author_info) != 0:
pa_print.tprint(f'✓ - Found by block')
# * Method 2 - Look for block starting with author name (bibtex needed)
else:
for author in range(author_count): # only look up to i authors
author_first = pub['author names'][author][0]
author_last = pub['author names'][author][1]
pa_print.tprint(f'\nLooking for: {author_first} {author_last}')
author_first = author_first.replace('\\',
'') # fixes issues with regex
author_last = author_last.replace('\\', '')
name_regex = r'(?:^.*' + author_first + r'.+' + author_last + r'.*$)(?:\s^[\S |].+$)*'
author_search = re.search(name_regex, doc, re.M)
try:
author_info.append(author_search.group(0))
pa_print.tprint('✓ - Found by name')
except:
pa_print.tprint('x - No match by name')
pa_print.tprint(
f'\n✓ - Found {len(author_info)} author(s) in paper of {author_count} total'
)
# If there were a different number of authors from text block
if len(author_info) < author_count:
pub['author block mismatch'] = 'Too few'
elif len(author_info) > author_count:
pub['author block mismatch'] = 'Too many'
# Add 'N/A' for missing authors # ! Note: Author block will not correspond in order to authors
authors_missed = author_count - len(author_info)
pub['author block missed'] = authors_missed
for author in range(authors_missed):
author_info.append('N/A')
# Add for visibility with csv - # ! but may not be the best idea if processing afterwards
pub['author infos'] = '\n\n'.join(author_info)
return author_info
def extract_grobid(pub, bib_db, iterator):
'''Parse xml files output from Grobid service (3rd party utility needed to generate files)
:publication (article) from database
'''
def elem_text(elem, fill='N/A'): # to get element text w/o error
if elem:
return elem.getText(separator=' ', strip=True)
else:
return fill
if 'pubpub' in pub['url']:
xml_name = f"nime{pub['year']}_{pub['article-number']}.xml"
else:
xml_name = pub['url'].split('/')[-1].split('.')[0] + '.tei.xml'
xml_path = xml_src + xml_name
if os.path.exists(xml_path):
with open(xml_path, 'r') as tei:
soup = BeautifulSoup(tei, 'lxml')
if soup.analytic is None:
pa_print.tprint(f'\n{xml_name} is empty!')
return
pa_print.tprint(f'\nParsing through grobid XML of {xml_name}')
grob_names, grob_emails, grob_orgs, grob_addrs = [], [], [], []
# Begin with parsing author info
authors = soup.analytic.find_all('author')
for author in authors:
persname = author.persname
if persname:
firstname = elem_text(persname.find("forename", type="first"),
'')
middlename = elem_text(
persname.find("forename", type="middle"), '')
surname = elem_text(persname.surname,
'') # *** should this be find? ***
name = (firstname, middlename, surname)
grob_names.append(name)
grob_emails.append(elem_text(author.email))
# There's an issue where affils can be within an <author> alongside an author or independently
# authors = [author for author in authors if not author.affiliation]
affils = [author for author in authors if author.affiliation]
for affil in affils:
grob_orgs.append(elem_text(affil.orgname))
grob_addrs.append(elem_text(affil.address))
grob_info = [grob_names, grob_emails, grob_orgs, grob_addrs]
# Fill in missing data with 'N/A'
author_count = pub['author count']
for author in range(author_count):
for info in grob_info:
try:
info[author]
except IndexError:
info.append('N/A')
# Add info to df - merge everything!
pub['grobid author names'].extend(
grob_names) # to check who appeared in grobid info
pub['grobid emails'].extend(grob_emails)
pub['grobid organisations'].extend(grob_orgs)
pub['grobid addresses'].extend(grob_addrs)
# Extract meaningful text using grobid tags (within p tags) and save to txt
grob_text_file = f"{text_src}grobid/grob_{xml_name.split('.')[0]}.txt"
if os.path.isfile(grob_text_file): # check if txt already exists
with open(grob_text_file, 'r') as f:
grob_text = f.read()
else:
# ! This needs to be a little more sophisticated
# PubPub tei's have expansive body
# /n and spaces need to be addressed
grob_text = []
grob_body = soup.body.find_all('p')
for p in grob_body:
p = re.sub(r'\s+', ' ', elem_text(p)).strip()
grob_text.append(p)
grob_text = str(grob_text)
with open(grob_text_file, 'w') as f:
f.write(grob_text)
return grob_text
else: # No XML - populate
pa_print.tprint('\nGrobid XML does not exist for paper!')
if 'pubpub' in pub['url']:
check_xml(bib_db, jats=True)
else:
check_xml(bib_db)
iterator.clear()
iterator.refresh()
def download_pdf(pdf_path, pub):
pa_print.tprint('\nLocal PDF not found - downloading...')
url = pub['url']
r = requests.get(url, allow_redirects=True)
open(pdf_path, 'wb').write(r.content)
# * Load database for email handle to uni matching
unidomains = load_unidomains(unidomains_path)
# * Load and extract BibTeX
bib_db = load_bibtex(bibtex_path)
bib_db = extract_bibtex(bib_db, args)
# * Loop here for Grobid/PDF population
if args.grobid:
check_xml(bib_db, True)
# * Parse data through pdfs
print('\nExtracting and parsing publication data...')
iterator = tqdm(bib_db)
for _, pub in enumerate(iterator):
pa_print.tprint(f"\n--- Now on: {pub['title']} ---")
# Extract text from pdf if not PubPub
if not 'pubpub' in pub['url']:
doc = extract_text(pub)
errored = doc_quality(doc, pub, 'text') # check for errors
# Only extract header meta-data if not errored
if not errored:
author_info = extract_author_info(doc, pub)
else:
author_info = []
else:
author_info = []
# Extract doc from Grobid
