Text data has certain unique characteristics that makes it a very different type of unstructured data compared to image data. As mentioned earlier, unstructured data requires extra processing steps to arrive at a structured format that can be used for developing a deep learning classification network. One of the applications of deep learning with text data involves developing a deep neural network sentiment classification model.

To develop a sentiment classification model, labels capturing sentiment related to the text data are needed. For example, we may use text data on movie reviews and a related sentiment label (positive review or negative review) to develop a model that can be used to automate the process. Another example could be the development of a sentiment classification model using text data on tweets. Such a model can be useful in comparing sentiments contained in thousands of tweets or and after an important event. Examples of such events where sentiment classification models can be useful include sentiments contained in tweets before and after the release of a new smartphone by a company, and sentiments contained in tweets before and after the performance of a presidential candidate in a live debate. A deep network for a sentiment classification model using text data is illustrated in Chapter 9, Deep Networks for Text Classification.