CSFD comments sentiment match

Semestral project for data mining from web.

CSFD is Czech imdb-like movie database.

Goal of this project is to compare how the comment sentiment matches the stars rating.

Method used

1. Crawler gets comments from the first 300 most incostintent-rated movies (to have all types of ratins)
2. We run affin sentiment analysis on the comment text, normalize it to have a value from 0 to 5 and compare it with the number of stars assigned to comment.
3. For sentiment analysis I'm using https://github.com/VilemR/affin.cz affin dictionary.

Results

Match in 1688 cases, star rating matched with sentiment in 14.8788012340238% of comments
Rating difference average: 1.6520052886734244

Interesting results - the comments with maximal difference between thee stars rating and sentiment were homonyms, irony or user errors - for example:

Slušnej nářez.

has maximal sentiment value, but the user rated it as total trash.

How to run

0. Install dependencies

yarn

1. crawl comments data

yarn start:crawl

2. run sentimental analysis on data

yarn start:sentiment