Define a interface to the text extractors

[jvanz]

To make the code clean, easy to maintain and expand, I think we should create interfaces to abstract complex and convoluted tasks. For example, in the current version of the library, we use a Apache Tika jar file to call a process and extract the text. However, we could use an Apache Tika server (#14) to the same job or even a third party tool for that. So, if build a well defined interface the client code which will extract the text does not need to know how it is done. Furthermore, it is easier to integrate with other packages in the future.