Current state of indexing office documents? collective.solr vs. collective.elasticsearch

What is currently the best option to index structured content (e.g. DOCX documents but also HTML (with their structure) in Plone using Solr or ES? By "structured content" I mean that it should be possible search in heading only instead of searching over a control document. The standard approach of Plone (throwing all content into a hug bag which dropping its structure) is not an option here.