Font LDL 2020

This is a repository for our ACL 2020 paper Let Me Choose: From Verbal Context to Font Selection.

Requirements

python==3.6.9 and pip install -r requirements.txt

Data

You can find the Font dataset in the following repository: https://github.com/RiTUAL-UH/Font-prediction-dataset

Settings required for the Emoji model:

Emoji Model: In this model, we use the Deep-Moji pre-trained model (Felbo et al., 2017) to generate emoji vectors by encoding the text into 2304-dimensional feature vectors. Our implementation is based on the Hugging Face Torch-moji implementation. You can find emoji vectors for the Font dataset here.

Instructions for running the code

pip install -r requirements.txt
python -m nltk.downloader wordnet
Download http://nlp.stanford.edu/data/glove.6B.zip and unzip glove.6B.100d.txt (part of glove.6B.zip) to EMBEDDINGS/glove.6B/glove.6B.100d.txt.
Download http://sentiment.nrc.ca/lexicons-for-research/NRC-Sentiment-Emotion-Lexicons.zip, then unzip NRC-Sentiment-Emotion-Lexicons/NRC-Emotion-Lexicon-v0.92/* and NRC-Sentiment-Emotion-Lexicons/NRC-VAD-Lexicon/* to DATA/emotion_lexicon.
Download http://saifmohammad.com/WebDocs/NRC-AffectIntensity-Lexicon.txt and copy NRC-AffectIntensity-Lexicon.txt to DATA/emotion_lexicon.
In config.py select the model and configurations. base_model values are "glove", "bert_seq_classification", "emoji" and "NRCfeat". (For more information about the details of the models check out the ACL paper)
Change train and test to True for training and testing respectively.
Use python main.py for running the model.

Citation

If you use this code in your work, please cite our paper as follows:

@inproceedings{shirani2020font,
  title={Let Me Choose: From Verbal Context to Font Selection},
  author={Shirani, Amirreza and Dernoncourt, Franck and Echevarria, Jose and Asente, Paul and Lipka, Nedim and Solorio, Thamar},
  booktitle={Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics},
  year={2020}
}

RiTUAL-UH/Font_LDL_2020