Since u ae using the term big data, hardware requirements won't be an issue. Since most Big Data technologies can work on commodity hardware. However, To process pdf & web-pages using ontology. I guess a machine with 2GHz dual-core processor and 4 GB RAM should be suffice to make up a single system. U can combine as many systems as you want to reduce the processing time of documents by adding more systems.

Its seem that you need a unified big data platform capable of the following two major operations.

1- Text extraction (IBIE based) from documents and storing 

For text extraction, indexing, etc you can use any normal server with many ordinary machines (as slaves, if needed) to handle big data operations. Use Hadoop ecosystem if the operations are acyclic and non-interactive else use Spark ecosystem. 

In this situation, PowerEdge T630 Tower Server can be a good solution option (cost up to $1400).

• Chipset Intel C610 series chipset
• Storage capacity : 144TB via 18 x 8TB 3.5” HDDs
• Memory supports Up to 1.5TB

or 

HP ProLiant ML150 Gen9 Special Server

2- Searching

If you are expecting normal and limited traffic for searching then virtualize the above server i.e, one for crawling/indexing and the second one is to handle query processing and searching while allocating resources.  

If you are expecting huge traffic then try  to use a cloud for searching.  

Since u ae using the term big data, hardware requirements won't be an issue. Since most Big Data technologies can work on commodity hardware. However, To process pdf & web-pages using ontology. I guess a machine with 2GHz dual-core processor and 4 GB RAM should be suffice to make up a single system. U can combine as many systems as you want to reduce the processing time of documents by adding more systems.