def download_info(self):
self.articles = arxivpy.query(search_query=['cond-mat.quant-gas'],
start_index=0,
max_index=30,
sort_by='submittedDate')
p_ = 'interesting_authors.csv'
url = "https://www.dropbox.com/s/yismcsi2ti35qse/interesting_authors.csv?dl=1"
u = urllib.request.urlopen(url)
data = u.read()
u.close()
with open(p_, "wb") as f:
f.write(data)
with open(p_, 'r') as my_file:
reader = csv.reader(my_file, delimiter=',')
self.interesting_authors = list(reader)[0]
p_ = 'interesting_keywords.csv'
url = "https://www.dropbox.com/s/u9pqzmomoa0jgmm/interesting_keywords.csv?dl=1"
u = urllib.request.urlopen(url)
data = u.read()
u.close()
with open(p_, "wb") as f:
f.write(data)
with open(p_, 'r') as my_file:
reader = csv.reader(my_file, delimiter=',')
self.interesting_title_keywords = list(reader)[0]
def scrape_arxiv():
articles = arxivpy.query(
search_query=['cs.CV', 'cs.LG', 'cs.CL', 'cs.NE', 'stat.ML'],
start_index=0,
max_index=200,
results_per_iteration=100,
wait_time=5.0,
sort_by='lastUpdatedDate') # grab 200 articles
arxivpy.download(articles, path='arxiv_pdf')
def crawl_machine_learning(start_index: int, sort_order: str):
conn = connect_database()
machine_learning_categories = [
'cs.CV', 'cs.CL', 'cs.LG', 'cs.AI', 'cs.NE', 'stat.ML'
]
STEP = 100
articles_per_minute = STEP * 2
article_len = articles_per_minute
if start_index == -1:
start_index = Article.get_n_articles(conn) - STEP
logging.info('crawling start')
logging.info('start index : ' + str(start_index))
logging.info('sort_order : ' + sort_order)
while article_len == articles_per_minute:
# query 100 results per iteration
# wait 30 seconds per query
try:
articles = arxivpy.query(search_query=machine_learning_categories,
start_index=start_index,
max_index=start_index +
articles_per_minute,
results_per_iteration=STEP,
wait_time=30,
sort_by='lastUpdatedDate',
sort_order=sort_order)
# crawling log
logging.info('last: ' + articles[-1]['published'])
logging.info(str(start_index + STEP * 2) + ' articles crawled')
# save articles
for article in articles:
Article(article, conn).save()
# compute start_index
start_index += STEP * 2
# compute article_len
article_len = len(articles)
# sleep 5 minute
time.sleep(MINUTE * 5)
except Exception as e:
logging.error(e)
time.sleep(60 * 30)
conn.close()
def fetch_recent_cv_papers(filename, num=65536):
papers = arxivpy.query(search_query=['cs.CV'],
start_index=0,
max_index=num - 1,
results_per_iteration=128,
wait_time=2.0,
sort_by='submittedDate')
# Normalise articles
for paper in papers:
# Dates as strings
paper['publish_date'] = paper['publish_date'].isoformat()
paper['update_date'] = paper['update_date'].isoformat()
with open(filename, 'w') as f:
f.write(json.dumps(papers))
import urllib.request, json
from phraseg import *
import matplotlib.pyplot as plt
from wordcloud import WordCloud
import matplotlib as mpl
import csv
import arxivpy
mpl.rcParams['figure.dpi'] = 300
articles = arxivpy.query(search_query=['cs.CL'],
start_index=0,
max_index=500,
results_per_iteration=100,
wait_time=1.0,
sort_by='lastUpdatedDate')
datas = ""
for a in articles:
datas += a['title'] + "\n"
datas += a['abstract'] + "\n"
print("Finish fetching")
phraseg = Phraseg(datas, idf_chunk=300)
result = phraseg.extract(result_word_minlen=1, merge_overlap=True)
wordcloud = WordCloud(font_path='wordcloud/NotoSansCJKtc-Medium.otf',
width=1800,
height=1000,
margin=1,
background_color="white").fit_words(result)
def crawl_category(term='cs.LG'):
index_iteration = 500
logging.info("Crawling category : %s", term)
for index in range(start_index, end_index, index_iteration):
logging.info("\nBatch : %d-%d" % (index, index + index_iteration))
articles = arxivpy.query(search_query=[term],
start_index=index,
max_index=index + index_iteration,
results_per_iteration=index_iteration,
wait_time=0.2,
sort_by='lastUpdatedDate')
article_batch_count = len(articles)
if article_batch_count == 0:
logging.warning('Article not found in batch %d - %d' %
(index, index + index_iteration))
for idx, article in tqdm(enumerate(articles),
total=article_batch_count):
arvixID = article['id'].split('v')[0]
query = Paper.select().where(Paper.arvixID == arvixID)
if query.exists():
paper = Paper.get(Paper.arvixID == arvixID)
categories = paper.category
if term not in categories:
categories.append(term)
Paper.update(category=categories).where(
Paper.arvixID == arvixID).execute()
continue
success, article_meta = get_arvixpaper_semantic_scholar(arvixID)
if success is False:
logging.debug(
"Paper not exists in semantic scholar, arvixID : %s" %
arvixID)
continue
authorIDList = [
int(author['authorId'])
if author['authorId'] is not None else -1
for author in article_meta['authors']
]
authorNames = [article['main_author']]
authorCount = len(article_meta['authors'])
if authorCount > 1:
other_author = [
name.strip() for name in article['authors'].split(',')
if len(name) > 1 and name != article['main_author']
]
authorNames += other_author
paper_category = [article['term']]
if article['term'] != term:
paper_category.append(term)
try:
paper = Paper.create(
indexID=idx + index,
arvixID=arvixID,
paperId=article_meta['paperId'],
doiID=str(article_meta['doi']),
title=article['title'],
summary=article['abstract'],
category=paper_category,
comments=article['comment'],
journal_ref=article['journal_ref'],
url=article['url'],
authorID=authorIDList,
authorName=authorNames,
authorCount=authorCount,
publishedDate=article['publish_date'],
citationVelocity=article_meta['citationVelocity'],
referencesCount=len(article_meta['references']),
topics=article_meta['topics'],
venue=str(article_meta['venue']),
year=article_meta['year'],
influentialCitationCount=article_meta[
'influentialCitationCount'],
citationCount=len(article_meta['citations']),
citations=article_meta['citations'],
)
try:
for meta in ['page', 'figure', 'table']:
if meta in article['comment']:
comment = article['comment'].replace(';', ',')
for segment in comment.split(','):
if meta in segment:
page_prefix = segment.split(meta)[0]
if meta == 'page':
paper.pages = int(page_prefix.strip())
elif meta == 'figure':
paper.figures = int(
page_prefix.strip())
elif meta == 'table':
paper.table = int(page_prefix.strip())
break
except:
logging.debug("Error in parsing meta data")
paper.save()
except BaseException as e:
logging.warning("Error in arvix id %s, error: %s" %
(arvixID, str(e)))
time.sleep(0.3)
if args.random:
args.start = int(np.random.uniform(low=0, high=10000))
args.number = int(np.random.uniform(low=1, high=3))
print("random mode, start: {}, number: {}".format(
args.start, args.number))
if args.field == 'cv':
search_query = ['cs.CV']
else:
search_query = args.field
print('Searching for {}'.format(search_query))
articles = arxivpy.query(search_query=search_query,
start_index=args.start,
max_index=args.start + args.number,
results_per_iteration=100,
wait_time=5.0,
sort_by='lastUpdatedDate') # grab 200 articles
print("Available Keys: ", articles[0].keys())
# print(articles[1])
paperlist_file = open("paperlist.txt", "w")
items = []
for idx, article in enumerate(articles):
items.append(
"============================================ Paper {} ===========================================\n"
.format(idx + 1))
items.append("Title: \n {}\n".format(article['title']))
items.append("Author: \n {}\n".format(article['authors']))
def crawl_machine_learning(start_index: int, sort_order: str):
# conn = connect_database()
DBU = RDS_utils()
check_axv = arxiv_id_check(DBU)
machine_learning_categories = [
'cs.CV', 'cs.CL', 'cs.LG', 'cs.AI', 'cs.NE', 'stat.ML', 'cs.MA'
]
STEP = 100
articles_per_minute = STEP * 2
article_len = articles_per_minute
# if start_index == -1:
# start_index = Article.get_n_articles(conn) - STEP
logging.info('crawling start')
logging.info('start index : ' + str(start_index))
logging.info('sort_order : ' + sort_order)
update_paper, insert_paper, insertfail, updatafail = 0, 0, 0, 0
while article_len == articles_per_minute:
# query 100 results per iteration
# wait 30 seconds per query
try:
start = time.time()
articles = arxivpy.query(search_query=machine_learning_categories,
start_index=start_index,
max_index=start_index +
articles_per_minute,
results_per_iteration=STEP,
wait_time=5,
sort_by='lastUpdatedDate',
sort_order=sort_order)
# crawling log
# logging.info('last: ' + articles[-1].get('published', ''))
logging.info(str(start_index + STEP * 2) + ' articles crawled')
# save articles
for article in articles:
data = Article(article, None).tolist()
print("'{}' cralwed / arxiv_id : {}".format(data[1], data[0]))
axvid, pubyear = data[0], data[6]
data[1], qt = get_qtitle(data[1]) # title
if pubyear:
qt = qt.strip() + str(pubyear)
pid = None
if type(check_axv) == dict:
if check_axv.get(axvid):
pid = check_axv[axvid]
else:
pid = DBU.get_pid_from_arXiv_id(axvid)
if not pid:
pid = Duplication_check.check(qt)
if pid:
ori = DBU.get_paper_by_p_id(pid)
data = Update_aXv_paper(ori, data)
if DBU.update_axv(pid, data):
update_paper += 1
else:
updatafail += 1
else:
if DBU.insert_axv(data):
pid = DBU.get_pid_from_arXiv_id(axvid)
Duplication_check.insert_title_year(qt, pid)
insert_paper += 1
else:
insertfail += 1
# compute start_index
start_index += STEP * 2
# compute article_len
article_len = len(articles)
e = int(time.tim() - start)
print('took {:02d}:{:02d}:{:02d} to crawl {} paper'.format(
e // 3600, (e % 3600 // 60), e % 60, article_len))
# sleep 1 minute, no 30 seconds
time.sleep(MINUTE / 2)
except Exception as e:
logging.error(e)
print("insert fail : {}, update fail : {}".format(
insertfail, updatafail))
DBU.DB.conn.close()
return start_index, insert_paper, update_paper
print("insert fail : {}, update fail : {}".format(insertfail, updatafail))
DBU.DB.conn.close()
return start_index, insert_paper, update_paper
