Is search the next industry to be revolutionized by deep learning? Lately researchers have been applying neural networks to search applications, with impressive gains. Search users use different language than what's contained in the corpus. For example, doctors create articles discussing jargon like 'myocardial infarction' but patients search use lay-terms like 'heart attack'. Mapping vocabularies using expert created taxonomies or word embeddings (word2vec, LDA, etc) can help. Manual approaches can take a great amount of work or don't map between searcher and document vocabulary. When clear associations between relevant documents and queries can be made, neural search can learn the patterns between query and document language embeddings, with tremendous gains on text search. Such embeddings can also be used to provide alternative representations of the user queries in order to better capture the user intents.
Join Doug Turnbull, author of 'Relevant Search', and Tommaso Teofili, author of 'Deep Learning for Search' as we explore this promising frontier. Is it a silver bullet? What are the pros and cons? And how can it fit into your search infrastructure using Solr, Elasticsearch or Lucene?