Query Formatting System Using Dictionary and Learning Model
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing query formatting technologies struggle to accurately recognize named entities in short queries, leading to reduced accuracy in query formatting, as they are designed for queries of a certain length and cannot effectively handle errors in input formatting such as missing spaces between words.
Innovation Solution
A query formatting system that uses a combination of dictionary data and learning models to identify and format tokens within a query, employing matching and machine learning methods to classify tokens and adjust their format, thereby improving accuracy by considering both dictionary matches and learning model outputs.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If existing named entity recognition technologies (Bi-LSTM CRF, BERT-CRF) are applied to query formatting, then they can recognize named entities in sentences of certain length, but they cannot recognize named entities in relatively short queries, resulting in reduced query formatting accuracy
Solution Approach 1:
The patent changes the input parameters of the learning model by providing not only the query but also formatting result information from dictionary-based matching. This parameter change allows the model to learn from both the original query and the partially formatted results, improving its ability to handle short queries where named entities may be missed by traditional NER approaches.
Solution Approach 2:
The system implements a feedback mechanism where the formatting results from dictionary matching are fed back into the learning model. The learning model uses this feedback information along with the original query to generate improved formatting results, creating an iterative refinement process that enhances accuracy for short queries.
2Measurement precision
If only dictionary-based matching is used for query formatting, then the process is simple and fast, but it cannot handle tokens that do not match the dictionary data, resulting in reduced formatting accuracy
Solution Approach 1:
The patent merges two different approaches: dictionary-based matching and machine learning-based formatting. The system first performs dictionary matching to obtain initial formatting results, then uses a learning model to process both the original query and the dictionary matching results. This combination allows the system to handle both common words (via dictionary) and unknown tokens (via learning model), improving overall accuracy while maintaining reasonable system complexity.
3Measurement precision
If a learning model is used alone for query formatting, then it can handle unknown tokens, but it requires more computational resources and processing time compared to dictionary-based matching
Solution Approach 1:
The system performs dictionary-based matching as a preliminary action before applying the learning model. This preliminary formatting step processes tokens that can be quickly matched against the dictionary, reducing the burden on the more computationally intensive learning model. Only tokens that require more sophisticated processing are handled by the learning model, thereby reducing overall processing time while maintaining accuracy.
Data Source
AI summary
Provided is a query formatting system including at least one processor configured to: acquire a plurality of tokens included in a query; execute matching through use of dictionary data based on the plurality of tokens; and format a second token which fails to match the dictionary data among the plurality of tokens based on a first token which matches the dictionary data among the plurality of tokens and a learning model relating to query formatting.