twpos.json

{
    "task": "tagger",
    "basedir": "./twpos",
    "conll_output": "twposresults.conll",
    "charsz": 16,
    "unif": 0.1,
    "preproc": {
    },
    "backend": "tensorflow",
    "dataset": "twpos",
    "loader": {
        "reader_type": "default",
	"named_fields": {
	    "0": "text",
	    "-1": "y"
	}
    },
    "features": [
	{
	    "name": "word",
	    "vectorizer": {
		"type": "dict1d",
		"fields": "text",
                "transform": "baseline.lowercase"
	    },
	    "embeddings": { "label": "glove-twitter-27B"}
	},
	{
	    "name": "word2",
	    "vectorizer": {
		"type": "dict1d",
		"fields": "text",
                "transform": "baseline.lowercase"
	    },
	    "embeddings": { "label": "glove-42B"}
	},
        {
            "name": "word3",
            "vectorizer": {
                "type": "dict1d",
                "fields": "text",
                "transform": "baseline.web_cleanup"
            },
            "embeddings": { "label": "w2v-twitter-30M" }
        },
	{
	    "name": "char",
	    "vectorizer": { "type": "dict2d" },
	    "embeddings": { "dsz": 30, "wsz": 30, "type": "char-conv" }
	}
    ],
    "model": {
        "model_type": "default",
        "hsz": 200,
        "dropout": 0.5,
        "rnntype": "blstm",
        "layers": 1,
        "crf": true 
    },
    "train": {
        "batchsz": 20,
        "epochs": 40,
        "optim": "sgd",
        "eta": 0.01,
        "mom": 0.9,
        "patience": 20,
        "early_stopping_metric": "acc",
        "span_type": "token",
        "clip": 5.0
    }
}