search

Blubberli / ConllSharedTask

0 stars 0 forks source link

MST #3

Open Blubberli opened 7 years ago

Blubberli commented 7 years ago

Change the MST so that it can use the output information of the network

Blubberli commented 7 years ago

TODO: After creating MST and evaluate it against the goldstandard. If there are errors change weights and run the NN again. Let the number of corrections be conditioned by sentence length.

--> how can we incorporate the information, if the MST cannot create any tree with the weights, into our learning process? add something loss.

TobiasElssner commented 7 years ago

Lecture about RNNs/ LSTMs:https://www.youtube.com/watch?v=iX5V1WpxxkY