A set of python modules for cornel movie-dialogs corpus with storm.

This module include some classes extending storm ORM for cornel movie-dialogs corpus data.

pip install storm                # if you not
pip install cornel-movie-dialogs-corpus-storm

1. download corpus and unzip
2. generate database and insert with generate-mdcorpus-database.py

for example:

generate-mdcorpus-database.py --corpus-dir "cornell movie-dialogs corpus" corpus.db

from mdcorpus.orm import *
from mdcorpus.parser import *

...

• MovieTitlesMetadata
• Genre
• MovieGenreLine
• MovieCharactersMetadata
• MovieConversation
• MovieLine
• RawScriptUrl

This is memo when I dealt with corpus problems.

• I ignored an alphabet following year.
  • for example, line 34, 1989/I
• I ignored duplication for genre data.
  • line 58, ['horror', 'mystery', 'mystery', 'sci-fi', 'sci-fi']

I use Python2.7 and I don't know how to use codecs module.(Unicode HOWTO — Python 2.7ja1 documentation)

convert text-code to utf-8 with Mi

cornell movie-dialogs corpus$ file --mime {(ls)}
README.txt:                    text/plain; charset=iso-8859-1
chameleons.pdf:                application/pdf; charset=binary
movie_characters_metadata.txt: text/plain; charset=iso-8859-1
movie_conversations.txt:       text/plain; charset=us-ascii
movie_lines.txt:               text/plain; charset=us-ascii
movie_titles_metadata.txt:     text/plain; charset=iso-8859-1
raw_script_urls.txt:           text/plain; charset=iso-8859-1

cornell movie-dialogs corpus$ file --mime {(ls)}
README.txt:                    text/plain; charset=utf-8
chameleons.pdf:                application/pdf; charset=binary
movie_characters_metadata.txt: text/plain; charset=utf-8
movie_conversations.txt:       text/plain; charset=us-ascii
movie_lines.txt:               text/plain; charset=us-ascii
movie_titles_metadata.txt:     text/plain; charset=utf-8
raw_script_urls.txt:           text/plain; charset=utf-8

• line 115, léon

• line 1727 - 1736, léon

sqlite> select * from movie_titles_metadata where title = 'léon';
sqlite> select * from movie_titles_metadata where title = 'l駮n';
114|l駮n|1994|8.6|204901