Counting Words That Count: NLP for exploring Romanian Parliament Transcripts

The data is obtained by scraping the cdep.ro website and contains 500k+ instances of speech from the parliament podium from 1996 to 2019. (Up to 2001 only the Chamber of Deputies published transcripts, after jan. 2001 Senate data is also included.) <br> <br> Columns: 'index' - incremented integer as row number in order of scraping 'title', - title of the scraped page, usually contains the name of the chamber and the exact data 'name', - the name of the speaker, preappended with Mr. or Mrs. 'speech', - the content of the speech, 'gender', - the gender of the speaker 'url' - the url to the profile of the speaker (useful for extending the data) CDEPs2.csv - Contains all transcripts, prone to parsing errors. 100% of data. validated-1.csv - Consists of 99% of original data. Less than 1% dropped for convenience. Ready to use.