siyan-sylvia-li / ColorLit

This repo contains the code and data for the paper Color Me Intrigued: Quantifying Usage of Colors in Fiction, accepted into the Creative AI Across Modalities workshop, AAAI 2023.

ColorLit

This repo contains the code and data for the paper Color Me Intrigued: Quantifying Usage of Colors in Fiction, accepted into the Creative AI Across Modalities workshop, AAAI 2023.

Top level .py files

• classify_glasgow.py: Given an input file and a Glasgow Norm Label (in this work we focus on IMAG, CNC, and VAL), load the corresponding trained Glasgow norm model and predict the corresponding value for each word.
• compute_glasgow_means.py: Compute the average Glasgow Norm values per author.
• plots.py: Plot the separate Glasgow Norm plots over time.
• plot_per_author.py (not shown in paper): Plot the GloVe embeddings of color-dependent nouns between two different authors.

final_data folder

All the data used for plotting; the final average Glasgow Norm values are in the *_author_means.csv files for each type of Glasgow Norm labels.

litbank_color_lines folder

Contains the lines containing color words (or synonyms) and the dependent noun extracted from all LitBank novels.

color_extraction folder

find_dependent_noun.py: Use stanza to extract color-dependent nouns; assuming the input file contains all individual sentences containing a color word / synonym of interest.

glasgow_models folder

Training and evaluation code for the Glasgow Norm MLP's. The trained models are also stored in this directory.