Analyze moral framing in texts.

Moralization Detection with Dictionaries of Morality Indicating Words (DiMi)

DiMi is a dictionary-based preprocessing step for detecting moralized language in text. It looks for curated lemma matches in four languages and returns the matched sentence plus two sentences of context before and after the match. The lexicon was used to preprocess the data in the Moralization Corpus (Becker et al., 2026). You can read more about DiMi in Detection and Analysis of Moralization Practices Across Languages and Domains (Becker et al., 2023).

Moralization Analysis with Language Models

This section uses machine learning models to predict whether a text contains a moralization.
You can choose between a fine-tuned model: "XLM-RoBERTa", and two large language models (LLMs): "Claude Haiku 4.5", and "OpenAI GPT-5-mini".
The LLMs provide a short explanation plus extracted protagonists and moral values, while "XLM-RoBERTa" returns only a prediction confidence because it is a classification model. "Claude Haiku 4.5" and "OpenAI GPT-5-mini" are general-purpose models for multiple languages, while "XLM-RoBERTa" was fine-tuned on the Multilingual Moralization Corpus (pending publication) and supports all available languages. Each user receives 20 free credits per day for external model predictions. XLM-RoBERTa is free, while Claude Haiku 4.5 and OpenAI GPT-5-mini use 1 credit per prediction.
Language models can and will make mistakes so please use results with caution!