Description
Read the following paper and write a critical summary of it:
Zhang et al., 2015. Character-level Convolutional Networks for Text Classification. Advances in
Neural Information Processing 28. http://papers.nips.cc/paper/5782-character-levelconvolutional-networks-for-text-classifica
Your summary should include a description of the paper’s contents, its strengths and limitations, and
any points that you did not understand. Since this paper discusses techniques that are not covered in
the lectures, I expect that there will be concepts that you do not understand! Do not worry, and
document this in your write-up.
Also, discuss the following issues in your writeup:
1. How does the proposed method handle OOV items? How does this compared to the methods
that we have discussed in class?
2. There does not seem to be a single method that does well across all the datasets. Was this a
surprise to you? What are some possible explanations for this? What does this tell us?
3. Table 4 presents the results of many model variants next to each other. Is this justified?
Your summary should be at most 1 page. It will be graded on the basis of its coverage, linguistic quality,
and clarity of argumentation.
Resources
• Jason Eisner’s guide to reading papers: https://www.cs.jhu.edu/~jason/advice/how-to-read-apaper.html
• It may help to read this tutorial on convolutional neural networks for NLP by Denny Britz before
reading the paper: http://www.wildml.com/2015/11/understanding-convolutional-neuralnetworks-for-nlp/