Is you input tokenized? I don't know Japanese, but what I know of OpenNLP name finder is that it requires tokenized input.
I hope it helps. William Em ter., 28 de jan. de 2020 às 00:08, Kayak28 <kaya.ota....@gmail.com> escreveu: > Hello, OpenNLP community: > > I have trouble with the conversion between python brat annotation tool to > OpenNLP. > > I would like to create a custom model in Japanese that can be used by > OpenNLP. > > Currently, I have xxx.ann file (which I annotated using python brat), > xxx.txt (which is the original NL document) file, and an annotation > configuration file. > > And reading this answer ( > > https://stackoverflow.com/questions/39877434/creating-and-training-a-model-for-opennlp-using-brat > ), > I have tried the following command. > > opennlp TokenNameFinderTrainer.brat -bratDataDir nlp_data/ > -annotationConfig exercise/brat-1.3_Crunchy_Frog/annotation.conf -moel > output_model.bin -lang ja > (nlp_data is a directory where ann and txt files are stored.) > > Eventually, the execution ended up with the following console output. > > Training data summary: > > #Sentences: 8030 > > #Tokens: 11953 > > #LOCATION entities: 28 > > #PERSON entities: 52 > > #VEHICLE entities: 14 > > #FACILITY entities: 88 > > #PLAN entities: 7 > > #EVENT_OTHER entities: 4 > > #EVENT entities: 20 > > #FACILITY_OTHER entities: 2 > > #ORGANIZATION entities: 59 > > #DATETIME entities: 110 > > #PRINTING entities: 8 > > #PRODUCT entities: 186 > > #TITLE entities: 7 > > #ACCESS entities: 255 > > #FOOD entities: 19 > > > Writing name finder model ... Compressed 52591 parameters to 10192 > > 427 outcome patterns > > done (0.375s) > > > Wrote name finder model to > > path: /home/vagrant/output_model.bin > > but when I used the output_model.bin file with the command below, it does > not give named entities, responding like below. > ```sh > opennlp TokenNameFinder output_model.bin > Loading Token Name Finder model ... done (0.132s)スティーブ ジョブスは偉い人 > > (my input) $スティーブジョブスは偉い人 > (output) スティーブジョブスは偉い人 > ``` > So, my questions are: > Did I take the wrong steps? > If I did badly, how should I use xxx.ann file to make a custom model in > OpenNLP? > > Any help will be appreciated. > > > Sincerely, > Kaya Ota >
