Elasticsearch附件插件与自己的Tika实现

Question

I want to use the Tika toolkit to index content of documents files (pdf, docx...) and images (via tesseract plugin).

I tried elastic ingest attachment plugin ( https://www.elastic.co/guide/en/elasticsearch/plugins/master/ingest-attachment.html ) it works pretty good but without OCR build-in. And I have to send base64 of my file, so high memory usage + elastic index the "data" (base64) field which is useless.

I'm thinking of using directly Tika toolkit and then index the content in ElasticSearch.

So I'm wondering if it's a better way or not ?

Answer 1

We've created a system to process files (Crawl -> OCR -> Index -> Search). It's called Ambar . We built it with idea to create a good and solid replacement for Ingest Attachment.

As a search engine we use ElasticSearch, as a context extractor: Tika + Tesseract + ImageMagick + Custom extractors for PDF.

We made it to provide a simple, but yet powerful alternative to own Tika + ES implementation.

Check out Github to get more details.

Answer 2

At the time of writing, there is little to no documentation about enabling OCR via Tesseract in the elasticsearch-mapper-attachments plugin.

Everything is pointing to you handling the task of OCR outside of Elasticsearch, and then indexing the content separately.

Reference: https://github.com/elastic/elasticsearch-mapper-attachments/issues/10