0 ×

03_​Weak_​Supervision_​for_​Document_​Classification

Workflow

Weak Supervision for Document Classification
human-in-the-loopguided analyticsdocument classificationsentiment analysismovie reviewsweak supervisionweak modelgradient boosted treeprobabilistic outputcompare modelsbinary classification
Weak Supervision for Document ClassificationThis workflow defines a fully automated web based application that will label your data using weak supervision. Theworkflow was designed for business analysts to easily go through documents to be labeled in any number of classes.The user provides euristhics, that is simple labeling functions. Some documents gets labeled, a generative model isapplied, and a model can be trained using the a probabilistic input. Gradient Boosted Trees was used after the WeakLabel Model. The Process Step by Step1. Upload your documents2. Input labeling functions3. Train weakly-supervised model and final discrimative model4. Visualize the results and compare models using an interactive view Train Discriminative ModelA logistic regression is ued but any classifier canbe used (e.g. Gradient Boosted Trees) Compare ifDiscriminative model isbetter than Generativemodel Train Generative Model Ask to the user a number oflabeling functions to be refined. (e.g."movie is horrible" thenlabel = "bad review") Score predictions without Discriminative 0: train1: testkeep only weak signals Partitioning Text Preprocessing Weak LabelModel Learner Weak Label ModelPredictor Weak Label ModelPredictor Nominal ProbabilityDistribution Splitter Gradient BoostedTrees Learner Gradient BoostedTrees Predictor Rules Upload Column Splitter Compare Weak Supervision for Document ClassificationThis workflow defines a fully automated web based application that will label your data using weak supervision. Theworkflow was designed for business analysts to easily go through documents to be labeled in any number of classes.The user provides euristhics, that is simple labeling functions. Some documents gets labeled, a generative model isapplied, and a model can be trained using the a probabilistic input. Gradient Boosted Trees was used after the WeakLabel Model. The Process Step by Step1. Upload your documents2. Input labeling functions3. Train weakly-supervised model and final discrimative model4. Visualize the results and compare models using an interactive view Train Discriminative ModelA logistic regression is ued but any classifier canbe used (e.g. Gradient Boosted Trees) Compare ifDiscriminative model isbetter than Generativemodel Train Generative Model Ask to the user a number oflabeling functions to be refined. (e.g."movie is horrible" thenlabel = "bad review") Score predictions without Discriminative 0: train1: testkeep only weak signalsPartitioning Text Preprocessing Weak LabelModel Learner Weak Label ModelPredictor Weak Label ModelPredictor Nominal ProbabilityDistribution Splitter Gradient BoostedTrees Learner Gradient BoostedTrees Predictor Rules Upload Column Splitter Compare

Download

Get this workflow from the following link: Download

Nodes

03_​Weak_​Supervision_​for_​Document_​Classification consists of the following 328 nodes(s):

Plugins

03_​Weak_​Supervision_​for_​Document_​Classification contains nodes provided by the following 12 plugin(s):