Relevance is the data-quality property of being pertinent to the specific business question or problem being addressed. It keeps analysis focused on the slice of reality that matters instead of sending you after the wrong rabbit. section 2.2 section 5.2.1
Before analyzing text data, check whether it belongs to the question you are asking:
- Confirm that reviews concern your product rather than a competitor’s product. section 5.2.1
- Check whether the data covers the period being analyzed rather than an outdated period. section 5.2.1 section 2.2
- Establish whether comments are tied to orders or are general impressions from a review site. section 5.2.1
- Inspect duplicates before analysis, because copied comments can give one opinion disproportionate weight. section 5.2.1
A misleading review dataset
A collection labeled as product reviews may contain competitor reviews, stale comments, or one complaint copied repeatedly. Treating it as relevant can distort the analysis before any text-cleaning or modeling begins. section 5.2.1
