%20logo.svg?alt=media&token=8a329413-2638-4c59-9fe9-0dac942ffa3c)
Textual relevance
Textual relevance measures how well a document matches a query. However, since not all fields in a document are equally important, the relevance of each individual field can be adjusted.
Matching against neural indexes in your query pipeline allows you to search your collection using Neuralsearch. Neural indexes must first be created from within the Indexing section of the console.
When creating a neural index, choose the field(s) which most accurately describe the concept of the result. For example, selecting a field which contains descriptive values such as “large outdoor patio furniture set” will provide more accurate results than a field containing SKU or model values such as “Samsung 3445xc 32L Econoblast”.
After creating a neural index, all records in the collection must be re-indexed for the value to be generated. For Site Search or Shopify collections, this can be performed from within the Console, and is prompted upon index creation.
The precision of a neural index determines how closely the record must match the query for it to be included in the result set; a lower precision value will include more results, but may include results which are less relevant, whereas a higher precision value will include fewer results, but ensures that results are more relevant.
Field boosts allow for fine-tuning how much weight a match for a given field contributes to the overall index score. Unless a field has a field boost assigned, it won't be searched.
Weighting is relative; "0.2" is twice as important as "0.1", as is "0.4" to "0.2", and so on.
Example
Assuming a product has three fields and the following scores assigned to them.
Field | Score |
|---|---|
Title | 0.8 |
Description | 0.4 |
Category | 0.2 |
A keyword match in the title field has the highest weight, where a match in the category field is considered less important. If your customers tend to search for categories rather than specific products, increasing the score for the category field is likely to improve the search results.
Tip: Score fields by importance to your business. Aim to have few fields with scores above 0.5 to avoid over-saturation. When everything is important, nothing is.
The minimum index score determines which products should be included in the search result set. Products with an index score lower than the minimum index score are not be included.
Example
Assuming a search for "watch". The top results match the keyword watch in the title or other high scoring fields like category. Due to it being a perfect match the resulting index score is "1". Accessories, like wrist straps, are likely to mention the term "watch" in the description. But due to the lower score assigned to the description field, they have a lower overall index score. This could be a "0.4" for example.
The minimum index score enables you to decide whether you want those lower scoring results included in the search results are not. If set to "0.6", none of the accessories that scored lower than 0.6 are included in the results. if set to "0.4", they would be shown.
Tip: Whether it makes sense to surface lower scoring results depends on your requirements. We recommend experimenting with different scores to identify what setting yields the best conversion rate for your business.
Last modified 7mo ago