Description

The 34A forms are available at: https://forms.iebc.or.ke/. Although they are not text-readable, they contain a list of the party agents that were present at each polling station. This could be used to examine how a party's strategy of distributing party agents differs from its campaign strategy.

Description

User story: As a PI, I want to filter out "non-speech" videos so that I have space to download all videos.

Right now, there are way too many video IDs being saved to disk, such that we would require much more disk space in order to store them all. So we need to do some filtering before saving videos to disk.

filtering strategy: supervised learning. While it might seem like overkill to train a classifier to distinguish "campaign speech" from "non-speech" videos, this might end up being more efficient than generating a bunch of rules manually. To be safe, we can use a low threshold for classifying videos as "campaign speeches" (in order to keep Type I errors to a minimum).

Description

User story: As a PI, I want to compress downloaded videos so that I use disk space efficiently.

Revise video_download.py in order to zip downloaded videos before saving them to file.

Note: it may be the case that this would not reduce file size much anyways, so the better strategy will be to reduce the number of videos we need to download (see #3).

Description

Scrape 2017 election results from https://public.rts.iebc.or.ke.

Tasks

• ...