Data extraction for procedures
Potato's API is designed to extract data from scientific content in a way that is specific to the method used.
We generate Python-like data structures that cleanly encode key facts from scientific content.
Potato's API is designed to extract data from scientific content in a way that is specific to the method used.
We generate Python-like data structures that cleanly encode key facts from scientific content.
In this example, we show a paper that describes a neuroscience experiment, extract the methods, and provide the data structure.
Build higher quality search, clustering, and classification using pre-structured data we produce. Or work with us to build your own fine-tuned models.
| ACCURACY | ||||
|---|---|---|---|---|
| Model | Length | Search | Category | Classify |
| Raw Content | 100% | 0.37 | 0.41 | 0.76 |
| Abstractive gpt-4o | 22% | 0.38 | 0.40 | 0.78 |
| Abstractive o1-preview | 44% | 0.39 | 0.43 | 0.78 |
| Potato Structured | 70% | 0.44 | 0.48 | 0.81 |
| Potato Optimized | 24% | 0.46 | 0.53 | 0.81 |
One JSON API for uploading text content and one for fetching structured data. All that's required is the text and an API key.
Upgrade to Potato+ to get access to the API.
