DReAMy: a library for the automatic analysis and annotation of dream reports with multilingual large language models
Date:
Dream researchers have long faced the challenge of automatically annotating large sets of textual reports describing oneiric experiences in a consistent manner. While algorithms leveraging machine learning (ML) methods for the automatic analysis of sleep-related physiological data have remarkably improved, the automatic analysis of dream reports has lagged behind. Aside very few notable examples, they remain largely based on dated approaches lacking the ability to reason over the full context of a report and required considerable data pre-processing (e.g., distributional semantics models or dictionary-based analyses). In this work, we introduce DReAMy (Dream-Report Analysis Methods with python), a fully open-source python-based toolkit, summarised in Figure 1. Using state-of-the-art large language models (LLMs), DReAMy enables the analyses and a fast annotation of vast numbers of dream reports with respect to emotions and characters in each report.