Best practices for building a recommendation engine
Through the recommend parameter, Tridion Docs offers a very flexible way to build a recommendation engine. But building a good recommendation engine doesn't just depend on understanding the recommend parameter. This topic mentions some basic best practices that help you construct a recommendation engine that's best for your situation.
- Work with your own data, and start small
- A good recommendation engine strongly depends on your data. After familiarizing yourself with the new parameter and its syntax, you are most likely to achieve good results if you start off with a small data set that you control completely, and play with it until it does what you want. You can then apply what you've learned to a full data set.
- Create reusable, multi-purpose queries
-
Because the query ignores schemes that are not in use by the concept with
id, you can create a reusable query referencing many schemes in theconceptsarray. The query will automatically apply the schemes that are used by your item, and ignore the ones that aren't. - If necessary, take a deep dive into the underlying logic
-
To understand the recommendation behavior fully, beyond what you can specify in the query, you can always refer to the documentation of the
more_like_thisquery in OpenSearch. For further reading, see here: https://www.elastic.co/guide/en/elasticsearch/reference/current/query-dsl-mlt-query.html