Depth study has been widely used in natural language processing (NLP) field, in the traditional areas in the past seemingly complex challenges, but also have made very significant progress. This article from a Kaggle talking about the actual topic, and take this as an example, to show an application in depth study on the NLP. The real problem may be considered a typical example of semantic understanding or nerve translation, in the process of solving this problem, we will be specifically used include LSTM [5], Word Embedding [4], Siamese network [6], and many other technologies If readers do not understand a technique in which the end of the text can be found in the references given in several blog entry before I write. As used herein, the framework is still very concise Keras (Of course, if you're more familiar with TensorFlow or PyTorch, based on a simple rewrite of this article is very simple).
First, the problem description
First look at the problems we have to solve. The problem comes from the Kaggle site once held a contest [1], as shown below. As we all know, Quora is similar to the English version of "know peace," the website. Every day there will be a lot of people on Quora questions, many of these issues, although the question is asked is not exactly the same, but the essence is actually a problem. E.g:
- How can I see all my Youtube comments?
- How do I read and find my Youtube comments?
as well as
- How can I be a good geologist?
- What should I do to be a great geologist?
Now this is the field of NLP in a very common problem, it can be grouped into semantic understanding (understanding the meaning of two sentences and contrast) or nervous translation (translation from one language to describe the way or manner or another description language) category .
