The following functions both require a pretrained model,
which can be generated using
framenet_tools train as explained previously.
- Stages:The System is split into 4 distinct pipeline stages, namely:
- 1 Frameevoking element identification
- 2 Frame identification
- 3 Spanidentification (WIP)
- 4 Role identification (WIP)
Each stage can individually be trained by calling it
--frameid. Also combinations of mutliple stages are possible.
This can be done for every option. NOTE: A usage of
predict requires a previous training of the same stage level!
framenet_tools predict --path [path]annotates the given raw text file located at
--pathand prints the result. Optionally
--out_pathcan be used to write the results directly to a file. Also a prediction can be limited to a certain stage by specifying it (e.g.
--feeid). NOTE: As the stages build on the previous ones, this option represents a upper bound.
framenet_tools evaluateevaluates the F1-Score of the model on the evaluation files. Here, evaluation can be exclusively limited to a certain stage.
Training automatically logs the loss and accuracy of the train- and devset in TensorBoard format.
tensorboard --logdir=runscan beused to run TensorBoard and visualize the data.