This supports the classification of text files.
Refer to classification dataset You must be aware of the number of lables used for classification.
As of December 13, 2018, there are 9 labels used. Set ENV NUMBER_OF_LABELS=9 in the Dockerfile
Classification model is trained currently with large number of ARC documents, therefore it has bias towards that class. If the document given is ARC, you shuold get a 100% probability. If it is less than 1, (100%), normally it is UNKNOWN.
./deploy.sh DEV classify-fasttext
//add the following header if you dont want the chaining action of the pipeline.
-H "strategy:{}"
//this will work only in dev.
//the reason is, async-function is exposed only in DEV.
//in production the orchestration happens via the orchestration overlay network.
arc="$(cat sample_arc.txt)"
//if you want to do only upload - without doing any further steps
curl -v http://localhost/async-function/classify-fasttext \
-d "$(echo "$arc")" \
-H "page:1" \
-H "X-Callback-Url: http://webhook/" \
-H "X-Api-Key:$api_key"
{
"action": "classify",
"endpoint": "classify-fasttext",
"page": 1,
"result": {
"classification": ["ARC"],
"details": {
"ARC": 1.0000100135803223,
"TASKCARD": 1.0000003385357559e-05,
"TESTCELL": 1.0000003385357559e-05,
"LLPLIST": 1.0000003385357559e-05,
"MODLIST": 1.0000003385357559e-05,
"ADSB": 1.0000003385357559e-05,
"ACCESSORYLIST": 1.0000003385357559e-05,
"SERVICEABLETAG": 1.0000003385357559e-05,
"BORESCOPE": 1.0000003385357559e-05
}
},
"status": 200,
"total_time_taken": "2.972723960876465",
"next_action": "extract-fields",
"next_action_response": 202
}
Content type
Image
Digest
Size
808.3 MB
Last updated
about 6 years ago
docker pull blockaero/classify-fasttext:a16559384be15ed4c5a33601187fb549df80e88b