com.johnsnowlabs.nlp.annotators.ner.dl
NerDLApproach
Related Docs:
object NerDLApproach
| package dl
com.johnsnowlabs.nlp.annotators.ner.dl
Batch size (Default: 8)
Whether to check F1 Micro-average or F1 Macro-average as a final metric for the best model This will fall back to loss if there is no validation or test dataset
ConfigProto from tensorflow, serialized into byte array.
ConfigProto from tensorflow, serialized into byte array. Get with config_proto.SerializeToString()
Trains Tensorflow based Char-CNN-BLSTM model
Trains Tensorflow based Char-CNN-BLSTM model
Dropout coefficient (Default: 0.5f)
Whether to optimize for large datasets or not (Default: false).
Whether to optimize for large datasets or not (Default: false). Enabling this option can
slow down training.
Whether to output to annotators log folder (Default: false)
Entities to recognize
Entities to recognize
Whether logs for validation to be extended (Default: false): it displays time and
evaluation of each label
Batch size
ConfigProto from tensorflow, serialized into byte array.
ConfigProto from tensorflow, serialized into byte array. Get with config_proto.SerializeToString()
Dropout coefficient
Memory Optimizer
Whether to output to annotators log folder
whether to include confidence scores in annotation metadata
input annotations columns currently used
Learning Rate
Maximum number of epochs to train
Maximum number of epochs to train
Minimum number of epochs to train
Minimum number of epochs to train
Gets annotation column name going to generate
Gets annotation column name going to generate
Folder path to save training logs
Learning rate decay coefficient.
Learning rate decay coefficient. Real Learning Rage = lr / (1 + po * epoch)
Random seed
Random seed
useBestModel
Whether to use contrib LSTM Cells.
Whether to use contrib LSTM Cells. Not compatible with Windows. Might slightly improve accuracy.
Choose the proportion of training dataset to be validated against the model on each Epoch.
Choose the proportion of training dataset to be validated against the model on each Epoch. The value should be between 0.0 and 1.0 and by default it is 0.0 and off.
Level of verbosity during training
Level of verbosity during training
Folder path that contain external graph files
whether to include all confidence scores in annotation metadata or just score of the predicted tag
Whether to include confidence scores in annotation metadata (Default: false)
Input annotator types: DOCUMENT, TOKEN, WORD_EMBEDDINGS
Input annotator types: DOCUMENT, TOKEN, WORD_EMBEDDINGS
columns that contain annotations necessary to run this annotator AnnotatorType is used both as input and output columns if not specified
columns that contain annotations necessary to run this annotator AnnotatorType is used both as input and output columns if not specified
Column with label per each token
Column with label per each token
Learning Rate (Default: 1e-3f)
Maximum number of epochs to train
Maximum number of epochs to train
Minimum number of epochs to train
Minimum number of epochs to train
Output annotator types: NAMED_ENTITY
Output annotator types: NAMED_ENTITY
Folder path to save training logs (Default: "")
Learning rate decay coefficient (Default: 0.005f).
Learning rate decay coefficient (Default: 0.005f). Real Learning Rate calculates to lr /
(1 + po * epoch)
Random seed
Random seed
Batch size
ConfigProto from tensorflow, serialized into byte array.
ConfigProto from tensorflow, serialized into byte array. Get with config_proto.SerializeToString()
Dropout coefficient
Whether to optimize for large datasets or not.
Whether to optimize for large datasets or not. Enabling this option can slow down training.
Whether to output to annotators log folder
Entities to recognize
Entities to recognize
Whether logs for validation to be extended: it displays time and evaluation of each label.
Whether logs for validation to be extended: it displays time and evaluation of each label. Default is false.
Folder path that contain external graph files
whether to include confidence scores for all tags rather than just for the predicted one
Whether to include confidence scores in annotation metadata
Overrides required annotators column if different than default
Overrides required annotators column if different than default
Column with label per each token
Column with label per each token
Learning Rate
Maximum number of epochs to train
Maximum number of epochs to train
Minimum number of epochs to train
Minimum number of epochs to train
Overrides annotation column name when transforming
Overrides annotation column name when transforming
Folder path to save training logs
Learning rate decay coefficient.
Learning rate decay coefficient. Real Learning Rage = lr / (1 + po * epoch)
Random seed
Random seed
Path to test dataset.
Path to test dataset. If set, it is used to calculate statistics on it during training.
Path to test dataset.
Path to test dataset. If set, it is used to calculate statistics on it during training.
Whether to use contrib LSTM Cells.
Whether to use contrib LSTM Cells. Not compatible with Windows. Might slightly improve accuracy.
Choose the proportion of training dataset to be validated against the model on each Epoch.
Choose the proportion of training dataset to be validated against the model on each Epoch. The value should be between 0.0 and 1.0 and by default it is 0.0 and off.
Level of verbosity during training
Level of verbosity during training
Level of verbosity during training
Level of verbosity during training
Path to test dataset.
Path to test dataset. If set, it is used to calculate statistics on it during training.
requirement for pipeline transformation validation.
requirement for pipeline transformation validation. It is called on fit()
required uid for storing annotator to disk
required uid for storing annotator to disk
Whether to restore and use the model that has achieved the best performance at the end of the training.
Whether to restore and use the model that has achieved the best performance at the end of the training. The metric that is being monitored is F1 for testDataset and if it's not set it will be validationSplit, and if it's not set finally looks for loss.
Whether to use contrib LSTM Cells (Default: true).
Whether to use contrib LSTM Cells (Default: true). Not compatible with Windows. Might
slightly improve accuracy. This param is deprecated and only exists for backward
compatibility
takes a Dataset and checks to see if all the required annotation types are present.
takes a Dataset and checks to see if all the required annotation types are present.
to be validated
True if all the required types are present, else false
Choose the proportion of training dataset to be validated against the model on each Epoch
(Default: 0.0f).
Choose the proportion of training dataset to be validated against the model on each Epoch
(Default: 0.0f). The value should be between 0.0 and 1.0 and by default it is 0.0 and off.
Level of verbosity during training (Default: Verbose.Silent.id)
Level of verbosity during training (Default: Verbose.Silent.id)
A list of (hyper-)parameter keys this annotator can take. Users can set and get the parameter values through setters and getters, respectively.
Required input and expected output annotator types
This Named Entity recognition annotator allows to train generic NER model based on Neural Networks.
The architecture of the neural network is a Char CNNs - BiLSTM - CRF that achieves state-of-the-art in most datasets.
For instantiated/pretrained models, see NerDLModel.
The training data should be a labeled Spark Dataset, in the format of CoNLL 2003 IOB with
Annotationtype columns. The data should have columns of typeDOCUMENT, TOKEN, WORD_EMBEDDINGSand an additional label column of annotator typeNAMED_ENTITY. Excluding the label, this can be done with for exampleFor extended examples of usage, see the Spark NLP Workshop and the NerDLSpec.
Example
NerConverter to further process the results
NerCrfApproach for a generic CRF approach