About OpenSensations^™

OpenSensations is a set of open models that read sensations in text, and an API that lets any app use them.

It is part of AI Attributes™, alongside open models for other attributes of text.

Who it is for

• Researchers: people studying how sensation and bodily feeling are described in what models and people write.
• Developers: anyone who wants these scores in their own app, through the API or by running the weights themselves.
• Platforms: tools such as AI Observability Studio and Am I Conscious? that show the scores beside a model's words.

How it works

Text is split into blocks, one per sentence. The model scores one word at a time in its sentence, so every word of a block that carries meaning is scored, and each sense keeps its highest score, from 0 to 1.

The model that answers here learned from people's ratings of how strongly about 40,000 English words are tied to each sense, alone and in sentences. On words it never saw in training, its scores correlate 0.81 with people's ratings. Read the limitations before using them.

2 models are published, RoBERTa and Gemma, with open weights on Hugging Face. See Models.

Every result names the model and the exact revision that produced it. A revision never changes, so a score can always be traced to the weights behind it.

Who is behind it

OpenSensations is developed by AI Attributes™, an independent open research project. Others can contribute fine-tuned models: the rules are on its Contribute page.

Start with the API guide, or read the limitations.