haotianchi / episode-miner

episode-miner

Provides methods to find events from text based on user defined vocabulary. Stores the result in 'events' layer of EventText object. Uses Winepi algorithm to find frequent serial episodes in event sequence.

Installation

git clone https://github.com/estnltk/episode-miner.git
cd episode-miner
python setup.py install

# or pre-install requieremets separately: 
pip install -r requirements.txt

Usage

For details see the docs.

Quick example

Find frequent episodes from text.

from estnltk.taggers import EventTagger
from episode_miner import EventText, EventSequences, Episode, Episodes

event_vocabulary = [{'term': 'üks'}, 
                    {'term': 'kaks'}]    
event_tagger = EventTagger(event_vocabulary, case_sensitive=False, return_layer=True)
event_text = EventText('Üks kaks kolm neli kolm. Kaks üks kaks kolm neli kolm üks kaks.', event_tagger=event_tagger)
event_sequences = EventSequences(event_texts=[event_text], 
                                 classificator='term', 
                                 time_scale='start')
frequent_episodes = event_sequences.find_serial_episodes(window_width=31, 
                                                         min_frequency=0.3, 
                                                         only_full_windows=False, 
                                                         allow_intermediate_events=True)
frequent_episodes

[('üks'),
 ('kaks'),
 ('üks', 'kaks'),
 ('kaks', 'üks'),
 ('kaks', 'kaks'),
 ('kaks', 'üks', 'kaks')]