Course talk:CPSC532:StaRAI2020:Convolutional2DKnowledgeGraphEmbeddings

You were supposed to describe 2 papers, one of which extended the other. This reads like a description of one paper. What did the first paper do? What was the incremental contribution of the second paper? What is the intuition of how/why it works? I couldn't work out what it does from the description.

If the figure is not yours, it is important to explicitly cite it in the text. It needs to be clear that it is not yours. We should be able to tell without clicking on it.

Thank you for your feedback. I did not get the idea of how to write the article very well and your feedbacks and reading other's articles helped me to get the idea. I made some edits and I hope it answers these questions now.

Hi, I have a small doubt. How did convolution rectify issues of TransE embedding?

Also, I think this line: "The main characteristic of our model is that the score is defined by a convolution over 2D shaped embeddings" needs to be edited to third person sense.

Thank's for your feedback. I have edited my article and added some paragraphs and I think it may be more clear now.

I think providing more information about TransE could be helpful in understanding what ConvE changed. This will also allow you to provide more detail on ConvE results as Lucca suggested.

Thank you for your feedback. I've added a paragraph about TransE and hope that's better now.

Interesting choices. I believe it'd be nice to briefly go into more detail about the results of ConvE, and how it does better.

Their results and experiments had a lot of details in hyperparameters and different datasets they have been used. I mentioned the related table in the source and also briefly summarized their results. Hope that's better now. Thanks for your feedback!