stanford-oval / thingtalk / 398 / 1

78%

master: 78%

push

travis-ci

ToNN: add the ability to allocate the entities sequentially

When passing the context to the neural network, we want a representation
that does not depend on the input sentence, in particular in which order
entities where represented in the sentence. Instead, we want
a normalized, sequential order.

This adds an option to toNN that uses a different entity assignment
strategy, instead of fetching from a precomputed bag.

3184 of 4065 branches covered (78.33%)

Branch coverage included in aggregate %.

5621 of 6672 relevant lines covered (84.25%)

123.28 hits per line