Bugfix: use UTF8 file encoding everywhere

For better compatibility with Windows.

Bugfix: not able to preprocess lists of URLs

Fixes a small bug in the preprocessing data, preventing passing the expected list of URLs for insertion.

Experimental like/retweet support

Adds a new preprocess type twitter_v2_likes_retweets for handling likes and retweets from the V2 Twitter API. This will allow coordinated liking to be examined using the co-retweet network type. Currently this support is very experimental and should be used with care.

An example of creating a like network from the output of the likes for a specific person:

compute_networks likes.db preprocess --format twitter_v2_likes_retweets likes_1.json likes_2.json

The only useful command for this is the "co_retweet" function, which is overloaded to match on accounts liking the same tweet. The time-window parameter is related to the logical ordering of likes (for liking-users and retweeted-by), or not used at all (for liked-tweets), as there is no precise timestamp associated with when the like happened. The logical ordering of the time window means that behaviours are considered coordinated if they are within time_window likes of each other in the stream.

compute_networks likes.db compute co_retweet --time_window 60 --output_file likes_within_60.graphml --output_format graphml

Support Python 3.6 again

Fixes a bug that prevented the toolkit working on Python 3.6.

Bugfix: actually output directed graphs.

1.4.2

Bugfix: actually writed directed graph output via graphml

Release bugfix for co link on resolved URLs

1.4.1

Fix reference to wrong column in resolved url index construction

Asymmetric Calculation of Coordination Strength

Coordination strengths are now calculated in a direct manor, by counting the number of unique messages from a user A that satisfy the coordination criteria with messages from user B. By consider the directionality and counting messages, this means that it is possible to identify where significant differences in user posting volume might lead to distortions in calculated quantification scores.

Parallel computation of networks

1.3.0

Merge pull request #17 from QUT-Digital-Observatory/parallelise_more_…

Ignore mentions for text similarity and co-tweet evaluation

Changes the default text processing to remove @mentions from text before calculating co-similarity and co-tweet networks.

Add support for directly ingesting Twitter V2 API format JSON data

Merge pull request #14 from QUT-Digital-Observatory/twitter_v2

Support ingest directly from Twitter V2 API format