6 Episodic memory and NLP

 

This chapter covers:

  • How strongly supervised end-to-end memory networks can be applied to other sequential NLP problems than just standard Question Answering.
  • How to implement a multi-hop memory network that allows for semi-supervised training.
  • How strongly supervised memory networks compare to semi-supervised memory networks.

For the data we use, we will find that strongly supervised memory networks easily produce above-baseline results with very little effort. Semi-supervised memory networks produce better accuracies in a number of cases, but not consistently.

The following picture displays the chapter organization:

Figure  6.1. Chapter organization.
mental model chapter6 all

6.1  Memory networks for sequential NLP

mental model chapter6 memnet

6.2  Data and data processing

6.2.1  PP attachment data

6.2.2  Dutch diminutive data

6.2.3  Spanish part-of-speech data

6.3  Strongly supervised memory networks: experiments and results

6.3.1  PP-attachment

6.3.2  Dutch diminutives

6.3.3  Spanish part-of-speech tagging

6.4  Semi-supervised memory networks

6.5  Semi-supervised memory networks: experiments and results

6.6  Summary

6.7  Code and data

6.8  Further reading

sitemap