04 August 2016

SharePoint Document Tagger With Jeopardy Winning Analysis Tool?

Document Tagger logotypeA couple of years ago developed a SharePoint solution that tagged SharePoint documents with metadata. We never released the final product on the market, though, as it requires a really good text analysis service. At that time we could not find a service that met our requirements, but today we have better options for Document Tagger.

Metadata – great to have but a pain to add
Everything that describes the contents and context of files is called metadata, like file names, location and of course keywords. The SharePoint search works much better when documents are tagged with appropriate metadata, and  with the help of metadata you can filter and create views that show files from different aspects.

 To manually tag documents with metadata is however boring and time consuming, so people often avoid doing it. In the developer group we have discussed several solutions to this problem.

Metadata solutions
Templates Manager logotypeThe product Templates Manager for SharePoint makes it easier to remember adding metadata. This is because we have made a change in how documents are opened in SharePoint. When you click on a template to create a new document from it, the document will not open immediately. Instead the Properties dialog will be displayed first, which is an important enhancement compared to the standard SharePoint behavior.

An even better solution to the problem of adding metadata is to let an analysis service give metadata suggestions which the users can accept or reject. That is the idea behind Document Tagger, a solution that sends documents to an analysis service in a secure way and gets metadata suggestions back to the SharePoint library. But of course Document Tagger will not be useful until it is combined with an analysis service that really gives relevant metadata to add to each document. Too many irrelevant suggestions will only make users frustrated.

In April I wrote a blog post about the Microsoft Azure Text Analytics, which we have found promising in tests. Now we have seen that also IBM's AlchemyLanguage seems to meet our standards.

AlchemyLanguage iconAlchemyLanguage is part of Watson, a question answering computer system that can answer questions posed in natural language. Some years ago the program beat all human players in a Jeopardy show and thus made the amazing prospects of cognitive systems clearer to many of us.

We can use both
The analysis service is crucial to Document Tagger, but technically it is not a big part of the solution and we can very well build several editions that send documents to different services.

We will continue evaluating the two analysis services I have mentioned above and maybe some more,  before we decide the future path for Document Tagger. Please contact us if you have comments!

By Peter Kalmström
CEO and Systems Designer Business Solutions