ffnet-summary-prediction

NLP model for predicting FFnet genre tags from summary.

dangerzone/ contains files only of historical value, used during development; they are not meant to be used further.

---

ff_analysis_tools

use class FFWebtext to reference raw web-scraped text, whether in string form or as a reference to a .txt file. Each work is represented by a Summary object, which initially contains information about the summary and genres of the work; the predicted genres (pred_genre1/2) can be modified by a SummarySet object, which is used to access HF models to use to predict genre values.

---

Example Summary object:

s1 = Summary('She had a tiny, tiny, tiny crush on him. Maybe.', 'Romance', 'Drama', 'Romance', 'Hurt_Comfort')

s1.pred_acc()
>>> '2t1r'

Example FFWebtext object:

ffwt1 = FFWebtext(filename='ffdump7.txt')

ffwt1.to_csv('ffdump7.csv)

pd.read_csv('ffdump7.csv)
>>> <DataFrame>

Example SummarySet object:

ffwt1 = FFWebtext(filename='ffdump7.txt')
slist1 = ffwt1.to_summarylist()

ss1 = SummarySet(slist1)
ss1.predict('zdreiosis/ff_analysis_5', revision='1454370')

ss1.summarylist[52].pred_genre1
>>> 'Romance'

ss1.summarylist[52].pred_genre2
>>> 'Friendship'

---

other examples can be found in the examples/ directory.