ben-bolte-deep-language-modeling-for-question-answering-using-keras.json

{
  "description": "Neural network models have revolutionized many areas of data analysis,\nbut have yet to make their way into mainstream usage in a number of\npopular fields. Recent advances in question-answering have come\nlargely from creative applications of deep learning. In this tutorial,\nI will demonstrate how to modify the open-source framework Keras to\nbuild some of these models.\n\nQuestion answering has received more focus as large search engines\nhave basically mastered general information retrieval and are starting\nto cover more edge cases. Question answering happens to be one of\nthose edge cases, because it could involve a lot of syntatic nuance\nthat doesn\u2019t get captured by standard information retrieval models,\nlike BM-25 or LSI. Hypothetically, deep learning models are better\nsuited to this type of task because of their ability to capture\nhigher-order syntax. Two papers, \u201cApplying deep learning to answer\nselection: a study and an open task\u201d (Feng et. al. 2015) and \u201cLSTM-\nbased deep learning models for non-factoid answer selection\u201d (Tan et.\nal. 2016), are recent examples which have applied deep learning to\nquestion-answering tasks with good results.\n\nFeng et. al. used an in-house Java framework for their work, and Tan\net. al. built their model entirely from Theano. This tutorial will\ndemonstrate how to replicate the models used by each group using the\npopular open-source framework Keras, adding custom functions to\ninclude recent advances from the neural networks community.",
  "duration": 6659,
  "language": "eng",
  "recorded": "2016-09-14",
  "speakers": [
    "Ben Bolte"
  ],
  "thumbnail_url": "https://i.ytimg.com/vi/bvZnphPgz74/hqdefault.jpg",
  "title": "Deep Language Modeling for Question Answering using Keras",
  "videos": [
    {
      "type": "youtube",
      "url": "https://www.youtube.com/watch?v=bvZnphPgz74"
    }
  ]
}