We'll have a short presentation on deep NLP and review the paper "StarSpace: Embed All The Things!"
The major vendors in cloud-based services are starting to provide machine learning as a service, such as Google Cloud AutoML Natural Language (https://cloud.google.com/natural-language/automl/docs/) and Azure Language Understanding (LUIS) (https://azure.microsoft.com/en-us/services/cognitive-services/language-understanding-intelligent-service/). Jim Tyhurst will give a brief demonstration of IBM Watson Natural Language Classifier service (https://www.ibm.com/watson/services/natural-language-classifier/), which enables you to build a custom classifier with no programming. Just create a new instance and submit training data. When the system has finished training, you can submit a document through a web API. The system responds with JSON, giving a list of some possible categories with a confidence score associated with each category. We will discuss some of the advantages and disadvantages of this service.
StarSpace: Embed All The Things! - describes a general-purpose neural embedding model that can solve a wide variety of problems: labeling tasks such as text classification, ranking tasks such as information retrieval/web search, collaborative filtering-based or content-based recommendation, embedding of multi-relational graphs, and learning word, sentence or document level embeddings. https://arxiv.org/abs/1709.03856