I want to use a pretrained summarization model on new data.
I downloaded the model textsum_epoch7_14.69_release, cpu version.
So far, I could successfully run the commands in the tutorial for preprocessing and training the model. I did not fully train the model since first I want to use the trained model and see how it performs on new data.
Also, Can I use it to get multiple sentences in output?
Do you encounter any specific issues when using the pretrained model?
The model outputs what it was trained on, which are single sentences in this case.
I can use the pretrained model with the Giga/input.txt as described. I wanted to know
- If I use new data for predictions, do I have to run through any preprocessing or it could be just a text document or do I have to run tokenizer script?
2)So both the input and output have to be single sentences. Is the implementation used similar to NAMAS model?
3)Is there any possibility that the toolkit may have any support/pretrained model for document summarization where the input or output can be more than one sentence, similar to Google model https://github.com/tensorflow/models/tree/master/textsum model