Skip to main content
Elephas searches your documents in eight languages: English, German, Spanish, French, Hindi, Tamil, Japanese and Chinese. You can ask in one language and find passages in another, for example ask in English about a German contract. Elephas also matches a question that is worded differently from your document.

​
Before you start

​
Supported languages

Only these eight languages are tested and supported. Documents in other languages, such as Italian or Dutch, may still be found, but results are not promised.

​
Steps

1

Add your documents as usual

There is no language setting. Add your files and folders, and Elephas recognises the language of each document on its own. See Add files and folders.
2

Ask in the language you prefer

Ask in any of the eight languages. Elephas finds matching passages in your documents, whatever language they are in.
3

Check the sources

Open the N sources button under the answer to read each passage in its original language. See Check every answer with citations.

​
How it works

​
Search works by meaning

Elephas finds passages by what they mean, not only by the words they share with your question. That is why a question in English can find a passage in German, and why “how much notice do we need to give” can find a clause about the “termination period”. Exact codes and numbers, such as invoice numbers, are still matched word for word. Search is built on your computer. The search model runs locally and never uses the network.

​
After you update to 1.0.2

Your existing documents are re-indexed in the background. You don’t need to add them again, and you can keep asking while it runs. For most libraries it takes seconds. Until a document is re-indexed, it is found by keyword search only. That works less well for Japanese and Chinese. Each computer re-indexes its own copy. Nothing about the re-index is synced.

​
Limits you should know

  • Search across languages is good, but same-language search is stronger. If you know the language of the document you need, asking in that language can find more.
  • Japanese and Chinese rely on meaning. Searching them for an exact word is weak. Ask about what the passage says instead.
  • German compound words are not split. A keyword search for “Vertrag” does not match “Mietvertrag”. Searching by meaning still finds related passages.
  • Older documents keep their original sentence splits. Japanese, Chinese and Hindi documents added from 1.0.2 onwards are split into sentences correctly. Documents added before keep their earlier split. To use the new split, remove the document and add it again.
  • Scanned pages are read as English. Text recognition for scanned PDFs and images reads English. See Scanned PDFs and images.

​
Troubleshooting

Check that it is in one of the eight supported languages. If you just updated, wait a moment for the background re-index to finish, then ask again. Narrowing the scope to that document can help. See Ask about specific documents only.
Exact-word search in Japanese and Chinese is weak. Ask about what the passage means instead of quoting a single word.
The answer would use a document in a language that personal information protection does not cover yet, such as Tamil, so nothing was sent to cloud AI. Use Local AI, or turn protection off if you accept sending that text as it is. See Personal information protection.

​
FAQ

No. Elephas recognises the language of each document and of your question on its own.
Yes. One Brain can mix languages, and one question can find passages in all of them.
Say which language you want in your question, for example “Answer in English.” Citations always show the passage in its original language.

Supported file types

Personal information protection

Tips for better answers

Scanned PDFs and images