Full unofficial data set of Reuters composed of 8,551,441 news titles, links and timestamps (Jan 2007 - Aug 2016).
NB: To generate it from scrach (from 2007 up to today), please scroll down.
git clone https://github.com/philipperemy/Reuters-full-data-set.git
cd Reuters-full-data-set
python read.py
ts = 20070228 11:46 AM EST, t = European stocks hit 7-week low amid new sell-off, h= http://www.reuters.com/article/companyNewsAndPR/idUSWEB277620070228
ts = 20070228 11:46 AM EST, t = Schering-Plough announces Ismail Kola as VP and Chief Scientific Officer, h= http://www.reuters.com/article/inPlayBriefing/idUSIN20070228164651SGP20070228
ts = 20070228 11:46 AM EST, t = O'Reilly Automotive forecasts 2007 earnings growth, h= http://www.reuters.com/article/marketsNews/idUSN2845320220070228
ts = 20070228 11:42 AM EST, t = Market Wrap, h= http://www.reuters.com/article/inPlayBriefing/idUSIN20070228164235WRAPX20070228
ts = 20070228 11:42 AM EST, t = Chile's CMPC net profit falls 13 pct in 2006, h= http://www.reuters.com/article/tnBasicIndustries-SP/idUSN2844077020070228
ts = 20070228 11:42 AM EST, t = Toyota Venezuela to halt March ops on currency woes, h= http://www.reuters.com/article/tnBasicIndustries-SP/idUSN2827887820070228
Each pickle file in data
represents a day (e.g. 20160102.pkl
is for Jan, 2 2016).
One day is composed of several news, gathered in a list
.
Each news is a dict
of the form:
ts: timestamp of the form 20070228 11:46 AM EST
title: title of the news
href: link to the article to get the full content
Nothing could be easier. Just run those commands to generate pickle and CSV files.
I get the data from http://www.reuters.com/resources/archive/us
.
git clone https://github.com/philipperemy/Reuters-full-data-set.git
cd Reuters-full-data-set
python3 generate.py
python3 dump_to_csv.py DATA_DIR # where DATA_DIR is the directory contained your pickle files from generate.py