This is the model from our paper "Story Embeddings — Narrative-Focused Representations of Fictional Stories". We publish the variant trained on augmented data.
You can use this just like the base model intfloat/e5-mistral-7b-instruct.
We trained with the task prefix "Retrieve stories with a similar narrative to the given story: " so we recommand that you use this for story retrieval.
If you are interested in just the adapter weights (e.g. for continued fine-tuning), check out the directory adapter-weights.
Do not hestitate to reach out if you encounter any issues! You are likely one of the first few people actually downloading this :)
Citation
If you are making use of this model in your publication, please cite our paper:
@inproceedings{hatzel-biemann-2024-story-embeddings,
title = "Story Embeddings -- Narrative-Focused Representations of Fictional Stories",
author = "Hatzel, Hans Ole and Biemann, Chris",
booktitle = "Proceedings of the 62st Annual Meeting of the Association for Computational Linguistics",
year = "2024",
address = "Miami, Florida",
publisher = "Association for Computational Linguistics",
}
