You have been so helpful in helping me to build on your system. Just want to look for your opinion on the following revision:
handling dialgoues in the input such that a boundary is inserted between the part spoken by party a and the part spoken by party b
the output summary is expected to be a spreadsheet with headers as key or a passage with section heading
For 1, as I see you come up with your original way to changing the way Bert deal with the sentence boundary level and achieving result, I think your input would be very valuable. I just worry if I do it any way I want would just screw up the whole thing and make Bert no longer work.
For 2, I have some experience with adding starting tags and ending for entities in neural based ner not based on Bert. But again I start to worry about what would happen when using Bert as it does not have such tags in the pretrainig phrase. What would be your suggestion to adding such tags in a way that makes sense and with the least chance to screw up?
You have been so helpful in helping me to build on your system. Just want to look for your opinion on the following revision:
handling dialgoues in the input such that a boundary is inserted between the part spoken by party a and the part spoken by party b
the output summary is expected to be a spreadsheet with headers as key or a passage with section heading
For 1, as I see you come up with your original way to changing the way Bert deal with the sentence boundary level and achieving result, I think your input would be very valuable. I just worry if I do it any way I want would just screw up the whole thing and make Bert no longer work.
For 2, I have some experience with adding starting tags and ending for entities in neural based ner not based on Bert. But again I start to worry about what would happen when using Bert as it does not have such tags in the pretrainig phrase. What would be your suggestion to adding such tags in a way that makes sense and with the least chance to screw up?
Many thanks.