This dataset has been created for the purpose of generating similes from literal descriptive sentences.
The process involves a two-step approach: firstly, self-labeled similes are converted into literal sentences using structured common sense knowledge, and secondly, a seq2seq model is fine-tuned on these [literal sentence, simile] pairs to generate similes. The dataset was collected from Reddit, specifically from the subreddits WRITINGPROMPTS… See the full description on the dataset page: https://huggingface.co/datasets/CreativeLang/scope_simile_generation.
