Hi Eric,
Thanks for your great work and docs.
I'd like to train your CNN by my data in Persian.
I formatted my data as you did on news_tagged_data.txt dataset with new tags like B-PERS, B-ORG, B-LOC ant etc. Then I made wordvecs.txt on my data used gensim and formatted my wordvecs.txt as you did.
I traind CNN successfully:
Test training data:
######################################
CNN - N-Gram test result. 4809 correct of 5235 total.
Accuracy is 0.918625
######################################
Test testing data:
######################################
CNN - N-Gram test result. 1076 correct of 1155 total.
CNN - Single word test result. 1294 correct of 1337 total.
Accuracy is 0.967838
######################################
but when I test my sentences using ./NER 4 I get wrong responses. Actually I get only "O" tags in response.
alich@Alich:~/named-entity-recognition$ ./NER 4
Type a query... (end with )
شهر رشت در کشور ایران قرار دارد و احمد میرهاشمی به آنجا سفر کرده
Successfully read network information from network/info_80.xml...
شهر : O
رشت : O
در : O
کشور : O
ایران : O
قرار : O
دارد : O
و : O
احمد : O
میرهاشمی : O
به : O
آنجا : O
سفر : O
کرده : O
Totally used time: 0.503745 second
Hi Eric, Thanks for your great work and docs. I'd like to train your CNN by my data in Persian. I formatted my data as you did on news_tagged_data.txt dataset with new tags like B-PERS, B-ORG, B-LOC ant etc. Then I made wordvecs.txt on my data used gensim and formatted my wordvecs.txt as you did. I traind CNN successfully:
Test training data: ######################################
CNN - N-Gram test result. 4809 correct of 5235 total.
Accuracy is 0.918625
###################################### Test testing data: ######################################
CNN - N-Gram test result. 1076 correct of 1155 total.
Accuracy is 0.931602
###################################### ######################################
CNN - Single word test result. 1294 correct of 1337 total.
Accuracy is 0.967838
######################################
but when I test my sentences using ./NER 4 I get wrong responses. Actually I get only "O" tags in response.
alich@Alich:~/named-entity-recognition$ ./NER 4 Type a query... (end with)
شهر رشت در کشور ایران قرار دارد و احمد میرهاشمی به آنجا سفر کرده
Successfully read network information from network/info_80.xml...
شهر : O
رشت : O
در : O
کشور : O
ایران : O
قرار : O
دارد : O
و : O
احمد : O
میرهاشمی : O
به : O
آنجا : O
سفر : O
کرده : O
Totally used time: 0.503745 second
would you please to help me?