Course talk:CPSC522/Graph Based keyword extraction

Great work Jiahong. Your page is a long but very interesting read. GDP vs centrality in country's top-corporates, who would have thought that :-). Here are couple of queries/suggestions:

You say corporate graph is like social graph in your context. Now social graph is easy to get these days, but how to get such graphs. The linking criteria you mentioned is often a private information. How the graph qualifies in terms of truly reflecting the relatedness between companies? any way to measure this?

Betweenneess Centrality one 'e' extra :-)

Hi Prithu,

Good questions! The linking criteria is based on board member information, which should be public for companies listed on stock exchange. The result of graph is compared to profit ranking, which is frequently used to rank companies, with the help of rank correlation algorithm. And higher correlation between two ranks means graph based result will be more accurate. Hope these answers helps, if you have further questions, please do not hesitate to contact me.

Best regards,

Jiahong Chen

Hi Jiahong, Thank you for your page. Your page is really detailed in analysis and also gives enough information for the naive user. I would like to provide the following feedback in the interest of improving the page:

• Some of the paras are a bit long. Breaking them into smaller paragraphs and/or modules would greatly help the reader in my opinion.
• The page as many have rightly mentioned is a bit longer. I suggest, you could put external links for concepts which are more widely known (e.g. Basics of Graph theory and graphical notations) while keeping the more convoluted concepts as it is.
• The work is really interesting. A bit more of your own personal thoughts on the future work section to point in the direction where this work might go in your own understanding would be a huge boost to a already great page.

Great job on the development of the WIKI and thank you

Hi Mehrdad,

Thank you for your kindly suggestions! I will cut off some useless paragraphs to shorten the page, and add more external links. I will also trying to add some interesting future works.

Many thanks, Jiahong Chen

Hi Jiahong Chen
Great page! It is quite an interesting topic. Some suggestions and questions:

• In the introduction you have mentioned "And then choose top-K companies from those centrality method results, where k should be larger than K." How does k differ from K?
• In the Graph Notation Examples m =12 (counting (symmetric) directed links), are you considering 2-way edges making it 6*2=12 ?
• I wasn't able to visualize how projection works. Maybe a figure would have helped!
• In the Degree Centrality section v is the number of nodes, what is deg(v)?
• Is possible that you provide a link for the data set?
• In table 2, do the nodes refer to companies?
• Is there any reference for the tail of the data containing a lot of noise?
• You can put an external link for Spearman`s rank correlation.
• Figure 8 is a little small. You might consider increasing the size.

Great job on the methodology, results and case study.

Hi Samprity,

Thank you for your kindly critique!

1) the difference between k and K is that they are different variables, it seems it would be better to choose another expression to not confuse them.

2) Yes, it is a 2-way edges as considering it as a directed graph.

3) I will try to find out a figure that represents it.

4) deg(v) it the degree of that point, and it will be the in-degree and out-degree if it is a directed graph.

5) Sorry, it is not a open source data.

6) Yes, they are companies.

7) Thank you! I forgot to add the reference, will add it soon.

8) Sure! Will add it soon.

9) Did you mean the figure Other "Important" Companies In China? Figure 8 seem like figure 7 and I suppose it might be big enough. I will increase the scale of figure Other "Important" Companies In China.

Thank you again for your kindly suggestions! I will correct all these errors soon.

Best regards,

Jiahong Chen

Thank you for the clarifications! Yes I did mean Figure 9.

Hi Jiahong,

Good job!

Here are a few suggestions:

>Centrality Ranking Based Corporate Network Analysis And Top-K Corporation Extraction

This subject is too long. It might be better to choose a shorter subject for this section.

What are the contents of this subject? I did not get what you have tried to put here. In the suggested template we were asked to explain the page in one sentence. The content of this section looks like what we usually put in the abstract section. In this case you can add it to the abstract part.

>Abstract

This section is great. I got the whole idea in just one paragraph.

>Introduction

What I understood from this section is that because of some reasons (mentioned in your page) in order to make a weighted graph lots of things should be considered. But I did not get why you build an un-weighted graph? (Is it just for simplicity? Or maybe you did not have enough data to decide about the weight.)

I liked how you explain the methodology and experimental results.

A quick proofread might help you.

Hi Bahare,

Thank you for your suggestions!

1) Yes, it seems a bit too long, I will try to figure out a shorter one. Also, I will try to write more brief summary for the page. Thanks.

2) I assume that the weighted graph is as important as the unweight one. However, due to the lack of time, I only able to carry out tests and evaluations on the unweighted graph. Sorry about that.

Again, thank you so much for your kindly suggestions!

Best regards,

Jiahong Chen