Deep-Structured Neural Network for Query Noise Filtering
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Search engines face difficulties in interpreting user queries and content items due to the presence of noisy linguistic items, especially in long queries and document summaries, which can lead to irrelevant results and inefficient resource usage.
Innovation Solution
A computer-implemented technique using a deep-structured neural network to process input linguistic items, generating word embeddings, state vectors, and probability information to distinguish important from less important tokens, thereby identifying relevant content items by comparing concept vectors associated with the input and candidate content items.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional token comparison methods are used to match queries against content items, then the search engine can process arbitrary length linguistic items, but the presence of noisy tokens reduces the accuracy of identifying relevant content items
Solution Approach 1:
The patent segments the linguistic item into individual tokens and processes each token through the neural network to generate separate state vectors and probability scores. This segmentation allows the system to evaluate each token's contribution independently, identifying and down-weighting noisy tokens while emphasizing meaningful ones in the final relevance determination.
Solution Approach 2:
The patent applies local quality by assigning different probability weights to different tokens within the same linguistic item. The neural network generates position-specific probability scores that reflect the local importance of each token, allowing meaningful tokens to have higher weights while noisy tokens receive lower weights, thereby improving overall matching accuracy.
2Adaptability or versatility
If the search engine processes long queries and document summaries with arbitrary lengths, then it can handle diverse content, but the computational resources required increase and processing efficiency decreases
Solution Approach 1:
The patent implements partial action by processing only the most relevant tokens identified through the neural network's probability scoring. Rather than uniformly processing all tokens in long linguistic items, the system focuses computational resources on tokens with higher probability scores, reducing overall processing requirements while maintaining accuracy for diverse length inputs.
3Measurement precision
If the search engine uses detailed token-level analysis to improve relevance, then it can provide more accurate results, but the complexity of the processing mechanism increases
Solution Approach 1:
The patent merges multiple processing functions into a unified deep-structured neural network architecture. The network combines word embedding generation, state vector computation, and probability scoring in an integrated framework, reducing the need for separate processing modules while achieving detailed token-level analysis for improved relevance determination.
Data Source
AI summary
A computer-implemented technique is described herein for identifying one or more content items that are relevant to an input linguistic item (e.g., an input query) using a deep-structured neural network, trained based on a corpus of click-through data. The input linguistic item has a collection of input tokens. The deep-structured neural network includes a first part that produces word embeddings associated with the respective input tokens, a second part that generates state vectors that capture context information associated with the input tokens, and a third part which distinguishes important parts of the input linguistic item from less important parts. The second part of the deep-structured neural network can be implemented as a recurrent neural network, such as a bi-directional neural network. The third part of the deep-structured neural network can generate a concept vector by forming a weighted sum of the state vectors.