Computing relevance indicators for each result - Fluid Topics - Latest

Understand Relevance in Fluid Topics

Category
Technical Notes
Audience
public
Version
Latest

The main challenge with relevance is how the sorting options are used. In Fluid Topics, AFS ranks every matching document by using several sorting strategies:

  1. AFS native relevance (descending order)
  2. Publication type: (book or document) > article > book topics
  3. Publication/topic title (ascending order)
  4. Publication/topic revision date (descending order)

It works the following way:

  • All the matching documents are sorted according to the first criterion.
  • If at least two documents have equal relevance, they are sorted according to the second criterion.
  • If documents are still tied, they are then sorted according to the third criterion.
  • And so on, until each document gets a distinct position.

If documents have the same sorting result after assessing all the strategies, they are ranked by their document ID corresponding to the indexing order, i.e. the latest ones are displayed first.

AFS ranks every matching document by using these previous options in the listed order. The higher the criterion on the list, the more important it is for ranking. Moreover, the ranking algorithm assesses each matching document from several criteria, and AFS native relevance can also be divided into attributes as listed below:

  1. Words
  2. FieldMatch
  3. Pathlen
  4. Weight

The logic seen in sorting strategies is also applied if documents have the same Words value, ranking algorithm assesses FieldMatch value. If documents have the same FieldMatch value, ranking algorithm assesses Pathlen value, and so on.