Scoring Theory
Once we have a list of matching items, they need to be ranked by a relevance score. This score is determined by three factors:
- Fuzziness score
- Key weight
- Field-length norm
Fuzziness Score
The fuzziness score is internally calculated via a modified implementation of the Bitap algorithm.
Distance, Threshold, and Location
The calculation for something to be considered a match (whether fuzzy or exact) takes into account how far the pattern is from the expected location, within a threshold.
To illustrate, consider the following options:
locationdefaults to0distancedefaults to100thresholddefaults to0.6
With the above options, for something to be considered a match, it would have to be within (threshold) 0.6 x (distance) 100 = 60 characters away from the expected location 0.
For example, consider the string "Fuse.js is a powerful, lightweight fuzzy-search library, with zero dependencies". Searching for the pattern "zero" would not match anything, even though it occurs in the string. The reason is that with the above defaults, for it to be considered a match it would have to be within 60 characters away from the expected location 0. However, "zero" appears at index 62.
TIP
If you don't care where the pattern appears in the string, and you still want to consider it a match, set ignoreLocation to true.
Key Weight
User inputted weight of the key. The higher the weight, the higher its relevance score. This is optional, although Fuse.js will internally default it to 1 if one isn't provided. This boosting is applied at query time.
Field-length Norm
The shorter the field, the higher its relevance. If a pattern matches a short field (such as a title field) it is likely to be more relevant than the same pattern matched with a bigger field. This is calculated at index time.
TIP
You can ignore the field-length norm by setting ignoreFieldNorm to true. Alternatively, you can configure how much the field-length norm affects your scoring by setting fieldNormWeight to a value other than 1 (the default).
Donate