Thread: Extracting listeners sentiment from YouTube comments

Hello,

I would like to be able to extract the sentiment of listeners on YouTube. More specifically, I would like to perform natural language processing majik on the comments of various songs. These songs come from Alpha Pup Records an independent record label.

If I were to successfully determine the sentiment of listeners for each song, then I would be able to compare the relative sentiment of various songs. Which song on artist x's new album receives the best sentiment?

Has anyone worked on projects of this nature? I would like to hear any anecdotes you have to offer; I am in the mood for any interesting read.

I am a stupid college student interested in various things. Where should I start, if were to embark on this project?

Here is an interesting YouTube video I found:
Text Mining Twitter, a STATISTICA Case Study - YouTube

While it does turn out to be an advertisement for statistical software, the case study is rather interesting. I need to be able to make use of similar tools in order to begin mining the comments. I've heard that R is a programming language that comes with these sorts of capabilities 'built-in'. Has anyone made use of R in these kinds of projects? I would love to hear!

Use the YouTube Comment Quality Improver for Greasemonkey

I wish I could help you more, but I've never worked with R or STATISTICA. I don't think R has text mining "built in", but Google turned up some free plugins, which should suffice. If they're anything like the STATISTICA package, they will remove a lot of the tedious, pain in the rear work of parsing text, extracting keywords, etc. A good part of your work would then be tuning the parameters, keyword lists, etc to get the most accurate representation of user comments/sentiment. The other big chunk of your work will be figuring out the YouTube API to get the comments you want. There appear to be some R plugins for network/HTTP programming (e.g. RCurl), to access the YouTube APIs.

I would start by playing with R a little bit, and master some of the basics. Then I would investigate some text mining and network plugins for R, see which seem to have the most features, are the most reliable, are currently developed/supported, etc. Once you pick out the packages you want to use, start in on either getting comments from YouTube, or working on the text mining. If you start with text mining, you can probably use basic, hand-created text files as sample input. Once you have those two parts working, you can integrate them and begin tuning the text mining parameters to get the data you want.

Remember, with bigger projects, it helps immensely to be organized and methodical. Come up with a plan. Implement small parts at a time, compile and test often. You may want to invest a little bit of time setting up a version control system, to keep track of all the changes to your code, and allow you to revert back to older versions if you find, e.g., a particular path you're headed down wont work. It's largely a matter of opinion/preference, but I'm partial to Git, it's easy to set up and get going, and it's branching capabilities are excellent, both in terms of speed and features, it's free and there's excellent documentation. If you have a VCS you are already familiar with and like, use that.

The comment improver is hilarious!

@anduril462

Your advice is wonderful! It makes me excited to get to work on this. I'll investigate Git, it makes sense it become organized and methodical from the start.