Regarding retraining figer

[praneethgb]

Hi xiaoling,

I want to re-train Figer with the addition of some data to existing training data but training data available in git hub is serialized using google protobuf. Can you please guide me to modify training data, so that I can add some training data and corresponding label in types.map of it?

[xiaoling]

please take a look at #5 #7

[praneethgb]

Ok, I will try this and update you if I have any issues.

Thanks

[praneethgb]

Hi xiaoling,
My objective is to add and modify the existing training data attached in https://drive.google.com/file/d/0B52yRXcdpG6MMnRNV3dTdGdYQ2M/view.
I am able to access the training data using Mention, can you point me to the file u used for creating mention object for a sentence? It would be really helpful if you explain how to add/modify training samples to existing training data.

To Modify Training data:

I will access the training data through mention object but after modifying, how to rewrite this in the same location.

[xiaoling]

(#5)

Please take a look at StanfordCoreNLP, which is the tool we use for syntactic analysis. Try using it on one of the sentences in the training set. Compare the output from StanfordCoreNLP with the ? fields in the protocol buffer file. It would be easy to see how they correspond.
Make sure that you use the Stanford Dependencies not the new universal ones.

If you prepare the new data in the same way as in the provided training data file, the system will generate features and so on for you. I currently don't have cycles to write the preparation code unfortunately.