14 comments

[ 4.6 ms ] story [ 53.7 ms ] thread
Buckaroo Banzai is #2, and #1 by Fog score. Helps to have your main character named "Buckaroo", I guess. And your mcguffin named the Oscillation Overthruster.
You could improve the score by mitigating the effect of words that occur often, e.g., by dividing by the log(word frequency). This would cover both names and effects of movie-specific language (think Star Trek)
Scripts let you do interesting analyses. Dialog, action (non-dialog), and directions all have their own syntax. I once wrote a tool that let you pick movies based on dialog complexity vs. ration action/dialog.

You also have to be careful to use the shooting script. The scripts available online are often many generations removed and quite different from the actually filmed scenes.

That sounds interesting. Consider putting it up on Github or some such.
Wouldn't using a database of subtitles remove most of the bias?
mostly. some subtitles contain Closed Captioning expressions like [Dogs Barking] or [Crowd Murmurs].

It'd definitely be less biased than the full scripts+stage direction though.

I love the concept.

This could be improved by using multiple reading level tests and averaging the results,

Kind of like this guy does with his open source project:

https://readability-score.com/

Second, the analysis is unfairly favorable to scripts with detailed stage directions, which tend to be more complex than dialogue.

Yeah, this basically makes it useless for me. Description can convey a mood but there's no telling how that will translate to what the audience sees. I used to like writing impressionistic, evocative descriptions; over time I shifted towards a very minimal style where each sentence describes a single action or visual subject.

They should rename it the Fletch-Kincaid score.

[During a proctological exam] Fletch: You using the whole fist, Doc?

I'm amazed that a talking head movie like Glengary Glen Ross (0.11) ranks lower than The Shining (0.73), with its long tracts of imagery and music.

http://www.imsdb.com/scripts/Glengarry-Glen-Gross.html http://www.imsdb.com/scripts/Shining,-The.html

I guess that's a testament to David Mamet's distinctive writing style, and his technique of characterization (...and probably the stage directions, as mentioned in the article).

Also, out of 956 movies, nothing ranks above 6th grade.

It's much more likely that it's a testament to the unreliability of this superficial readability measure. A long sentence may be less complex than a perhaps cryptic shorter sentence. A long, rare word may be more descriptive than a long-winded circumscription using common words.
Very cool. @dfkoz, did you do any pre-processing to remove script boilerplate (e.g., I'm looking at Zero Dark Thirty and there's a lot of CUT TO ... CUT TO ... CUT TO) or try to extract dialogue vs. stage direction?
Thanks! I removed all of the ALL CAPS words to eliminate many instances of character names and some stage directions.
Author here: all of the comments are spot on. This is what I love about Hacker News.