Analysing Movie Dialogues from the Cornell Corpus [Project3]

[HarshaDevulapalli]

The Cornell Corpus contains a large metadata-rich collection of fictional conversations extracted from raw movie scripts. My Pitch Number is #109

• 220,579 conversational exchanges between 10,292 pairs of movie characters
• involves 9,035 characters from 617 movies
• in total 304,713 utterances
• movie metadata included:
• genres
• release year
• IMDB rating
• number of IMDB votes
• IMDB rating
• character metadata included:
• gender (for 3,774 characters)
• position on movie credits (3,321 characters) The idea is to analyse dialogue and utterances by gender in each of these films and pitch them against one another as a test of gender dialogue equality. This is eerily similar to what has been done here : http://polygraph.cool/films/ - But the idea is to visualise it in a different way.

Data Source : https://www.cs.cornell.edu/~cristian/Cornell_Movie-Dialogs_Corpus.html

Story issue checklist

My pitch was (use the number): _

• [x] My pitch has been approved (see PITCHING.md)
• [x] My story issue links to my pitch issue
• [x] I link to my finalized (ish) data source(s)
• [x] I've included a brief summary of my story
• [x] I've included some possible headlines or findings
• [x] I've included some links or images as inspiration (if you have any)
• [ ] I have received two comments of peer feedback
• [x] I've included an update of my visualization/story in a comment
• [ ] I have received two comments of peer feedback after posting an update
• [ ] I have received editorial feedback

[HarshaDevulapalli]

Massaging the Data took its own sweet time. This data did not have too much that could be visualised fancily. So I put together a lot of relatively interesting things as an infographic. It's still pretty chaotic - I would like to organise this a little better. Comments please!

[playfairbot]

Hi there, I'm the Playfair Bot!

Thanks for posting your story issue, but would you mind adding editing the original issue to add the first draft of your image? You have my sincere apologies, but it's easier for dumb robots like me when the comments are only used for updates.

Thanks! :pray:

[ghost]

Yay this is very, very cool! The fruit of your labor is finally visualized. My comments/suggestions below.

1. I agree with @jsoma on the color vs. size hierarchy for the horizontal bars. Pick an order of colours and stick with it to make it more readable.
2. Dante having 537 lines means nothing if we don't know the context. Is that a LOT more than average? wildly more? How to tell? I would make that a visual, actually. Maybe a scatter plot with x=hours of movie and y = total lines, and show his dot as an outlier. You have a lot of words already.
3. The grids on your histograms look awful, so-- I would make them a lot tinier and really go for the grid if that is your intention-- or scrap them. entirely.
4. The titles of your histograms are inaccurate. Both the histograms just show the frequency of movies within given percentage ranges. Maybe change the titles to "Movies with low percentage of female dialogue more abundant" or something.

In general, not sure I would opt for the "one big cluster of infographics on a page" approach. Maybe breaking them up into detailed visualizations that would be able to be inserted into an article/viewed on a page separately would make more sense. I'd certainly like to see more detail on all of the charts!

Very cool work. Also post your code.

[gcgruen]

I would also think of overall visual hierarchy of the whole composition. Everytime you talk about this project, there are two aspects that you highlight to everyone (man have twice as much speech time and women's most frequent terms are "Yes Sir") So I would suggest to make these bigger in relation to the other graphics.

Also (I like you fonts!) I would suggest to put in a thousand separator to make numbers more digestible (i.e. 6,020 instead of 6020) -- highlights visually the difference to those which are "only" hundreds.

Last but not least, I would suggest to also put your name on the whole thing ;)