TeXplorer

Explore a document on your own device.
Text stays in this browser; public lookups send only the selected name.
ArchXAI browser prototype · Estonian, Finnish, Latvian and Russian

Additional anonymiser terms

⚠️ This demo uses quantised (8-bit) NER models to reduce download size and memory use. Results can be less accurate than with the original full-precision models. Please review recognised tags.

Names, places and organisations run on your device in all four languages. Only the selected model downloads on first use (Finnish ~82 MB, Estonian/Latvian ~393 MB, Russian ~176 MB). Click a highlighted tag for its evidence; edit here and Explore again to refresh. Import text, Markdown, CSV, PDF or EPUB. Scanned PDFs require an existing OCR text layer.

Options
No model used yet.

Ready to explore.

Exploring your document

Starting…

Read text from an image

OCR runs on this device. Images are not uploaded. The first use downloads about 77 MB of model weights from Hugging Face.

Review the transcription before exploring it. Swedish OCR does not add a Swedish NER model.

Choose the document language

No model has been selected. Your text is unchanged.

Choose how much to analyse

Longer documents take more time and memory. The complete text remains in the editor.