Part-of-Speech and Sequence Annotation

This chapter describes a selection of neural network architectures to annotate the words of a text with their part of speech. Starting from feed-forward networks, we will train models on annotated datasets and apply them to a text. We will move on to recurrent networks that fit well the sequential nature of sentences. We will illustrate each architecture with a program. As application, we will consider the annotation with parts of speech and then the recognition of names in sentences, either people, countries, or organizations.