4 Sentence Classification

In this chapter, we are going to study the task of sentence classification, where an NLP model receives a sentence and assigns some label to it. A spam filter is an application of sentence classification where it receives an email message and assigns whether it is a spam or not. If you want to classify news articles into different topics (business, politics, sports, etc.), it’s also a sentence classification task. Sentence classification is one of the simplest NLP tasks that have a wide range of applications including document classification, spam filtering, and sentiment analysis. Specifically, we are going to revisit the sentiment classifier we introduced in Chapter 2 and discuss its components in detail. At the end of this section, we are going to study another application of sentence classification—language detection.

4.1 Recurrent neural networks (RNNs)

4.1.1 Handling variable-length input

4.1.2 RNN abstraction

4.1.3 Simple RNN and Nonlinearity

4.2 Long Short Term Memory Units (LSTMs) and Gated Recurrent Units (GRUs)

4.2.1 Vanishing gradients problem

4.2.2 Long Short Term Memory (LSTM)

4.2.3 Gated Recurrent Units (GRUs)

4.3 Accuracy, precision, recall, and F measure

4.3.1 Accuracy

4.3.2 Precision and Recall

4.3.3 F measure

4.4 Building AllenNLP training pipelines

4.4.1 Instances and fields

4.4.2 Vocabulary and token indexers

4.4.3 Token embedders and RNNs

4.4.4 Building your own model

4.4.5 Putting all together

4.5 Configuring AllenNLP training pipelines

4.6 Case study: language detection

4.6.1 Using characters as input

4.6.2 Creating a dataset reader

4.6.3 Building the training pipeline

4.6.4 Running the detector on unseen instances

4.7 Summary