Support/Kitsune/Features/Forum-Data-Clustering

Title

Support forum data clustering

Problem statement/description

We have a great deal of unstructured data on the forums. This makes identifying areas of focus more difficult. This is not only a metrics issue, but it will make faceted search difficult. If our forum data is not structured (or structured differently than our KB articles), our search facets will not be useful.

Measurable outcome

An easy way to identify top issues trending in the forums. Active monitoring and alerts for breaking issues. Ability to serve accurate results from the forums in our search implementation.

This impacts our various groups in this way:

• User Advocacy - Greater insights into types of questions coming into the forums. Ability to establish problem buckets rather than problem threads. More accurate reporting of scale of problems
• Content Manager - Ability to identify which buckets drive the most traffic to the forums. Ability to create and modify KB to better address these issues.
• Community Manager - Ability to identify which buckets our unsolved issues fall into. Ability to create further training around these buckets or recruit specialists for the problems.
• User Advocacy - Ability to generate graphs based on trends for individual buckets. Greater insights into trends and bigger product issues

Possible Solutions:

• Manual tagging of questions based on a predefined set of tags in the AAQ Flow
• Manual tagging of questions based on a predefined set of tags in the forums by contributors
• Clustering through Mechanical Turk
• Automated clustering based on machine learning