Output clustering relationships in feature generator API

[paulalbert1]

ReCiter has several phases of clustering:

• Tepid (journal, MeSH major, coauthors, targetAuthorScopusAffiliationID)
• Email-based
• Grant
• Cites or cited by
• MeSH major

It would be useful for web interface users to see how articles are related to each other during clustering as defined in issue #217. Scores need not be generated because they’re already reflected in the clustering score.

As clustering occurs (it would be more efficient to do it this way than recomputing), please track cases where Article A is clustered with ArticleB and why.

If an article is in its own cluster, return:

clusteringEvidence-tepid: null
clusteringEvidence-cites: null
clusteringEvidence-meshMajor: null
clusteringEvidence-email: null
clusteringEvidence-grants: null

Note that tepid clustering is defined in step 3 of #217

Else, return any of the following as appropriate (the numbers are PMIDs of course):

clusteringEvidence-tepid: {234234, 345345...}
clusteringEvidence-cites: {234234, 345345...}
clusteringEvidence-meshMajor: {234234, 345345...}
clusteringEvidence-email: {234234, 345345...}
clusteringEvidence-grants: {234234, 345345...}

In the web interface for the end user, we will convert these lists of PMIDs to links.

[paulalbert1]

This looks less appealing as we pursue a more machine learning clustering based approach. Closing for now.