Mechanism for tracking accesses to ontologies/terms

[mmisw]

From paul.r.a...@gmail.com on July 27, 2009 16:00:19

The uri resolving mechanism should also be capable of tracking accesses to an
ontology and its terms. Most likely we'll want to know the user agent or some
other identifying mechanism. Not sure of other useful things to track.

Original issue: http://code.google.com/p/mmisw/issues/detail?id=158

[mmisw]

From paul.r.a...@gmail.com on July 27, 2009 16:01:39

Labels: ont

[mmisw]

From jgrayb...@ucsd.edu on April 05, 2010 17:25:59

I think this is a critical capability, one that we need to incorporate at a high priority.

Labels: -Type-Defect -Priority-Low Type-Enhancement Priority-High mmiorr Metrics

[mmisw]

From caru...@gmail.com on April 13, 2010 10:06:09

[from John's 2010-04-13 email]

We need to be able to see which ontologies are of the most interest and are most
used. Ideally, it should be tracking at the core URI level -- so whether I'm
searching by primary URI, or by some composite URI (with '#http://....' or by some
format request (.html, ?format=), the system figures out what ontology or term is
being requested and increments that counter by 1.

Owner: carueda
Labels: Milestone-Beta1

[mmisw]

From caru...@gmail.com on April 25, 2010 21:50:41

[from discussion with Luis on 4/13/10]

A possible implementation approach is to use an analytics service (eg., google
analytics) so it is triggered for each ontology/term URI request. The effected module
here is the "Ont" service, and this part of the solution will be automatic. Then, a
panel in the ORR portal can be updated manually with a report from the analytics
service to include the desired statistics. In a second phase, some API that the
analytics service provides can be used to also update the ORR panel automatically on
a regular basis.

[mmisw]

From caru...@gmail.com on June 16, 2010 23:22:02

Started preparations to track visits to ORR pages using Google analytics.

[mmisw]

From caru...@gmail.com on June 17, 2010 17:12:40

Under testing with MMI Portal 1.9.4.alpha (20100617170207):

Using Google analytics. The mechanism is activated if the property 'ga.uanumber' is defined in the build.properties file.

Besides the main ORR page itself (ie., index.html), the tracking is applied upon every history token change as the user navigates the various options in the portal. Currently this is always done with a _trackPageview call. Later on, _trackEvent might a better mechanism in certain operations.

NOTE: the current work for this feature request is on the ORR interface. Later on, the "Ont" URI resolver service should include also its own tracking mechanism as that's the core service that supplies ontologies and terms to clients in general. I've spent some time investigating how to apply google analytics tracking on the server side, so it could be used byt the Ont service, but it seems there is no straightforward way to do it, especially because the tracking code is mainly intended to be run on the browser side (javascript).

[mmisw]

From caru...@gmail.com on June 21, 2010 22:31:01

I started a possible mechanism to trigger events against Google Analytics from the Ont module using Rhino and EnvJs.

MMI Ontology and Term URI Resolver. Version 1.9.5.alpha (20100621222328)
Under preliminary testing.

[mmisw]

From caru...@gmail.com on June 24, 2010 20:55:51

Apparently, the Rhino+EnvJs mechanism is not working as no data with the prefix "/uri:" is showing up in the google analytics reports. Will look again into this next week.

Status: Started

[mmisw]

From caru...@gmail.com on June 28, 2010 22:02:50

Did various changes fixing some set-up to trigger the request to GA. The google image is being requested. Will be confirming the reports next.

[mmisw]

From caru...@gmail.com on June 29, 2010 12:09:10

"/uri:" entry now showing up in the GA report (see attached), which confirms the pageview is properly notified to GA from the Ont service.

In summary:

There are two places where the GA pageview events are generated:

1. In the ORR portal as the user opens particular ontologies. The argument for the pageView notification is the history token prefixed with '/'. So, not only ontologies but other actions are notified, eg., "/st" (search terms), "/admin", "/b" (browse ontology table), etc.

2. In the Ont service, which is the central URI resolver, so requests from all clients are now notified.

NOTE: these notifications will all be generated wrt to the Ont service itself, ie., there are no any association with the actual User Agent that requested the URI, so the GA report will not show info about countries, browsers capabilities, etc. I think this is OK at this initial phase of the tracking (at least for beta1).

Marking as fixed.
MMI Portal 1.9.4.alpha (20100617170207)
MMI Ontology and Term URI Resolver. Version 1.9.6.alpha (20100628171049)

==For a next version==

In general, some of the TODOs for a later version include:

• Determine where _trackEvent [1](instead of _trackPageview [2]) would be more appropriate
• At the Ont service, transfer User Agent information to the tracking for better reports. (Note that from the ORR portal this is handled appropriately.)
• At the Ont service, dispatch the GA notification asynchronously to avoid a potential long response to the client.

[1] https://code.google.com/apis/analytics/docs/gaJS/gaJSApiEventTracking.html#_gat.GA_EventTracker_._trackEvent [2] https://code.google.com/apis/analytics/docs/gaJS/gaJSApiBasicConfiguration.html#_gat.GA_Tracker_._trackPageview

Status: Fixed
Cc: jgrayb...@ucsd.edu bermud

Attachment: Analytics_mmisw.org_20100529-20100628_(TopContentReport).pdf