stanford-oval / genie-toolkit / 1085 / 5

56%

master: 56%

push

travis-ci-com

Add option for per-device evaluation

The evaluation functionality now takes an optional `--split-by-device`
flag. If true, the evaluation statistics are calculated for each device
encountered in the evaluation set, and reported separately.

Also fixed the getFunctions (and as a consequence, getDevices)
functions in requoting.js. These were incorrectly  including the dialogue
policy as a function / device. Now they don't anymore. Also added unit
tests for these two functions.

2753 of 11066 branches covered (24.88%)

Branch coverage included in aggregate %.

7439 of 21826 relevant lines covered (34.08%)

23611.92 hits per line