Understanding MongoDB Atlas Search Scoring for Better Search Results

Recently, while implementing hybrid search functionality in our application, I encountered a challenge in improving search result relevancy. While vector search and text search provided good results, we needed finer control over how results were ranked. MongoDB Atlas Search provides powerful features - boost and bury - that allow developers to fine-tune search rankings.
In this post, I'll give a quick overview of how you can implement score optimization in Mongo to help improve your search results.

If you're working with hybrid search in MongoDB, you might be interested in my previous posts about implementing semantic search (https://dev.to/shannonlal/building-blocks-for-hybrid-search-combining-keyword-and-semantic-search-236k) and hybrid search with filtering (https://dev.to/shannonlal/navigating-hybrid-search-with-mongodb-ugly-approach-1p9g).

Let's explore how we can leverage MongoDB's scoring features to create more relevant search results for your users.

Understanding Score Components
MongoDB Atlas Search uses the BM25 algorithm for scoring, which considers:

• Term frequency in the document
• Inverse document frequency (term rarity across all documents)
• Field length normalization

Implementing Score Modification
Here's an example of how to implement boost and examine score details:

db.collection.aggregate([
  {
    $search: {
      index: "my_index",
      compound: {
        should: [{
          text: {
            query: "ocean",
            path: "description",
            score: { boost: { value: 3 } }
          }
        }],
      },
      scoreDetails: true
    }
  },
  {
    $project: {
      description: 1,
      score: { $meta: "searchScore" },
      scoreDetails: { $meta: "searchScoreDetails" }
    }
  }
])

Understanding Score Details
The scoreDetails provide a breakdown of how the final score was calculated:

{
  value: 14.02,
  description: "sum of:",
  details: [
    {
      value: 8.92,
      description: "weight(description:ocean) [BM25Similarity]",
      details: [
        {
          value: 3,
          description: "boost"
        }
        // Additional scoring factors
      ]
    }
  ]
}

Optimizing Search Rankings

{
  compound: {
    should: [
      {
        text: {
          query: "search_term",
          path: "title",
          score: { boost: { value: 3 } }
        }
      },
      {
        text: {
          query: "search_term",
          path: "description",
          score: { boost: { value: 1 } }
        }
      }
    ]
  }
}

Understanding and implementing proper score modification in MongoDB Atlas Search can significantly improve search relevance for your users. As you implement these techniques, remember to:

• Start with score analysis using scoreDetails
• Test with real-world queries
• Monitor user interaction with search results
• Iterate based on user feedback

With these tools and techniques, you can create more precise and relevant search experiences in your MongoDB-based applications. Remember that scoring optimization is an iterative process - start simple, measure impact, and refine based on your specific use case.