Entity Sentiment Analysis via Signal Filtering
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing techniques for document analysis often assign sentiment polarity to entire documents or sentences, failing to accurately capture the varying sentiment polarities of individual entities within them.
Innovation Solution
A computer-implemented process that applies a filtering operation to the sequence of sentiment values surrounding identified entities in a document to determine their specific sentiment values, spreading sentiment values among nearby tokens and combining contributions from adjacent tokens to produce a sentiment value for each entity.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If sentiment polarity is assigned to entire documents or sentences, then the sentiment analysis covers the whole text, but it fails to capture the varying sentiment polarities of individual entities within them
Solution Approach 1:
The patent segments the document into individual entities and applies sentiment analysis to each entity separately rather than to the entire document. This is achieved by identifying entities within the document and computing sentiment values specifically for each entity based on surrounding tokens, thereby resolving the contradiction between measurement precision and analysis complexity
Solution Approach 2:
The patent applies different sentiment analysis treatments to different entities within the same document. Each entity receives a customized sentiment value computed from its local context (surrounding tokens), allowing the system to capture varying sentiment polarities for different entities while maintaining a unified analysis framework
2Measurement precision
If a filtering operation is applied to spread sentiment values among nearby tokens, then precise sentiment values can be assigned to entities, but the computational processing complexity increases
Solution Approach 1:
The patent introduces a filtering operation as an intermediary step between raw sentiment detection and final entity sentiment assignment. This filter spreads sentiment values from sentiment-bearing tokens to surrounding tokens, creating a distributed sentiment representation that improves precision while keeping the computational process manageable through a clear intermediate transformation step
Data Source
AI summary
A document can be processed to provide sentiment values for phrases in the document. The sequence of sentiment values associated with the sequence of phrases in a document can be handled as if they were a sampled discrete time signal. For phrases which have been identified as entities, a filtering operation can be applied to the sequence of sentiment values around each entity to determine a sentiment value for the entity.