Project: Bachmannpreis

Background

The 'Tage der deutschsprachigen Literatur' in Klagenfurt, Austria, is a major literary festival and basically the only one that is being televised on German television. This event consists of readings of 30 minutes by a number of invited writers who are being evaluated by a circle of critics. The culmination of the event is a live voting by the critics to determine which writers receive an award, most notably the Ingeborg-Bachmann-Preis. For several germanspeaking writers this event has been a starting or ending point for their literary career.

Structure

This project consists of five parts:

1. Getting the data / Wrangling the data / Storing the data

• Scrape data about the event (Wikipedia, official website of Bachmannpreis) using BeautifulSoup
• Use Goodreads API and geopy to acquire additional data
• Data Wrangling with pandas
• Set up database with PostgreSQL

2. NLP

• Sentiment analysis for jury discussion (with spacy and SentiWS)
• NER and POS-tagging for texts using flair
• Topic modeling with NMF and TFIDF

3. Feature Engineering / Prediction

• Feature engineering for prediction within pandas
• Predict winning authors with Random Forest model using AutoSKLearn

4. Web application

• Deploy website using Flask and chart.js
• Live website: http://countbachmann.herokuapp.com

5. Chatbot

• Deploy chatbot for user interaction (Rasa)