Machine Learning Search Token Prediction for Database Query Generation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Generating search queries from user-provided instructions in plain English is computationally intensive, leading to high resource utilization and reduced throughput in database systems, as existing technologies struggle to interpret and process natural language inputs effectively.
Innovation Solution
A system combining machine learning and natural language processing to identify search tokens within user instructions, reducing the search space and processing resources required, and refining these tokens to generate efficient search queries for database retrieval.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If an exhaustive search of user-provided words and phrases is performed to generate search queries, then the system can accurately interpret user instructions, but the computational resources required increase significantly and processing time is extended
Solution Approach 1:
The patent segments the search query generation process into distinct phases: first identifying candidate search tokens from user instructions, then evaluating and selecting the most relevant tokens. This segmentation allows the system to process user instructions in manageable steps rather than performing a single exhaustive search, thereby improving both accuracy and throughput by reducing the computational burden at each stage.
Solution Approach 2:
The patent applies preliminary action by pre-processing user instructions to identify and extract potential search tokens before the actual search query generation. This preliminary extraction phase prepares the data in advance, so that when search queries need to be generated, the system already has a refined set of candidate tokens to work with, significantly reducing the computational resources required during the query generation phase itself.
2Ease of operation
If processing resources are allocated to interpret plain English instructions and generate search queries, then user-friendly data retrieval is enabled, but fewer resources remain available for processing other data requests
Solution Approach 1:
The patent introduces an intermediary processing layer that translates plain English user instructions into structured search queries. This intermediary layer acts as a mediator between the user-friendly input interface and the database search system, enabling the system to maintain ease of operation while optimizing resource utilization. The intermediary efficiently processes instructions using targeted algorithms rather than exhaustive methods, preserving throughput for other requests.
3Reliability
If the system processes user instructions to identify search tokens using traditional methods, then search queries can be generated, but the processing duration increases and resource availability decreases
Solution Approach 1:
The patent changes the parameters of the search token identification process by using scoring mechanisms and ranking algorithms to evaluate candidate tokens. Instead of traditional exhaustive matching, the system assigns scores to potential search tokens based on their relevance to user instructions and selects the top-ranked candidates. This parameter change from exhaustive search to scored selection dramatically reduces processing duration while maintaining reliable search query generation.
Data Source
AI summary
A device configured to receive a data request comprising instructions for retrieving data. The device is further configured to input the instructions into a machine learning model that is configured to output a plurality of search token predictions for an entity based on the instructions. The device is further configured to generate a set of search query candidates based on the plurality of search token predictions. Each search query candidate comprises zero or more search tokens. The device is further configured to select a search query candidate based on a score that is associated with the search query candidate and to generate a search query using the search tokens from the selected search query candidate. The device is further configured to receive search results in response to sending the search query to the database and to output search results based on the search.