Semantic Playbook Ranking for Next Best Search Actions
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Users often struggle to generate effective search filters and refine queries in online searches due to a lack of understanding of optimal filters and search tactics, requiring domain expertise and iterative refinement without guidance.
Innovation Solution
An embedding model generates embeddings for user intents and playbooks, using a semantic relationship score to rank playbooks, recommending next best actions through a ranking machine learning model, and allowing user interaction to refine search results.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If users perform online searches without guidance, then they can conduct searches independently, but they struggle to generate effective search filters and refine queries due to lack of domain expertise
Solution Approach 1:
The system enables self-service by automatically generating search filter recommendations and next best action suggestions without requiring users to have domain expertise. The machine learning model analyzes user intent and playbook data to provide guided search assistance, allowing users to conduct effective searches independently while the system handles the complexity of filter generation and query refinement
Solution Approach 2:
The patent introduces an intermediary machine learning system that mediates between user search intent and optimal search filters. This intermediary component (the recommendation engine) translates user goals into effective search queries by leveraging playbook data and semantic relationships, thus bridging the gap between user capability and search effectiveness
2Measurement precision
If users iteratively refine search queries without guidance, then they can improve search results, but they require significant time and domain expertise to understand optimal filters
Solution Approach 1:
The system performs preliminary action by pre-processing playbook data and pre-computing semantic relationships between user intents and available filters. The machine learning model is trained in advance on historical playbook data, enabling it to quickly recommend optimal search filters and next best actions without requiring users to engage in time-consuming iterative refinement processes
Solution Approach 2:
The patent implements feedback mechanisms where the system learns from user interactions with search results and playbook recommendations. The model continuously refines its recommendations based on user behavior patterns, improving the precision of search results over time while reducing the time users need to spend on query refinement through adaptive learning
3Productivity
If the system recommends next best actions using machine learning, then search efficiency is enhanced, but the system complexity increases due to embedding models and ranking algorithms
Solution Approach 1:
The patent segments the recommendation system into distinct modular components: an embedding model component that converts text to vectors, a ranking model component that scores playbooks, and a recommendation component that generates suggestions. This segmentation allows each component to be independently optimized, trained, and maintained, managing system complexity while enabling sophisticated machine learning-based search efficiency improvements
Data Source
AI summary
In an example embodiment, an embedding model is used to generate an embedding of a natural language searching goal specified by a user, the embedding representing user intent of the user. Playbooks in a database of playbooks are also run through the embedding model to generate an embedding for each playbook indicative of a meaning of each playbook. A semantic relationship score can then be computed for each combination of the natural language search goal and a playbook, using the embeddings. These semantic relationship scores can then be passed into a ranking machine learning model, along with measured success rates for the playbooks, to generate a ranking of the playbooks. Based on this ranking, a set of filters and action corresponding to at least one of the playbooks may then be recommended to the user.