Useless tags being added to node

Looks like we need a bit more work on the tagging feature. I'm getting numeric tags added to nodes. For example, this story, http://www.npr.org/blogs/thetwo-way/2013/10/03/228771578/sharp-words-ove..., has "53" as a term attached to it. Not sure where that came from but I assume it's an API category or some such other machine-readable data.

Is a good place to start removing all numeric tags? Will probably create a helper function called _npr_pull_filter_tag that will filter an array of tags based on pre-determined criteria. Eventually, we could expose the criteria to an admin interface allowing the user to fine-tune our defaults.

Thoughts?