As your search has already found Apache Tika as the best solution, this is because the task to wrangling data from many different sources is complicated and that is what Tika seeks to solve. To be clear Apache Tika uses other open source tools to do that actual data extraction: Such as Apache POI to extract data from Word documents, and I think it might use PDFBox to extract text from pdf documents.
It isn't hard to extract text from word documents, or pdf documents but you need to handle each case and to build a robust way of managing these other tools will involve recreating what is Apache Tika. When you come across more exotic document types (for which there is no parser) Tika provides interfaces for defining your own extractor and adding that to the set of documents it can work with.
So you can either use tika or hand role a solution. This said there are tools which extend tika and provide all-in-one document search features that being Apache Solr. Apache Solr is a server which provides document parsing and indexing features using a restful api (even to feed it the documents), under the hood it makes use of the Lucene framework. If you can't stand working though a restful api and want to work directly in Java then use Lucene directly.