Improving Search Query Processing for a More Natural UX
Current Behavior
Currently, users must manually craft search queries by writing them in a format that resembles a desired post. However, this is unnatural, as users typically express what they want to find, rather than attempting to generate content themselves.
Examples of the Problem
A user searching for information about hiking with dogs in the mountains might currently need to phrase their search like this:
A scenic mountain hike with my dog, exploring beautiful trails and enjoying
nature.
However, users naturally search by expressing intent, such as:
I want to find posts about hiking with dogs in the mountains.
Another example:
Cooking a delicious Italian pasta dish with fresh ingredients at home.
Looking for recipes for homemade Italian pasta.
This requirement to format queries as content creates friction and makes the search experience unintuitive.
Proposed Solution
To make search more user-friendly, we could use a generative model that transforms user queries into well-structured prompts optimized for search. This approach would:
- Allow users to search in a natural, conversational way.
- Automatically convert intent-based queries into a structured prompt for better search results.
- Reduce the learning curve for new users who may not know how to structure their search effectively.