Aspect-Dependent Sentiment Lexicon Construction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current sentiment analysis systems face challenges in creating a universally optimal sentiment lexicon, as word polarity is sensitive to the topic domain, and there is no principled method to combine different sources of sentiment information, including dictionaries, language clues, and document-level overall sentiment ratings, especially in aspect-level sentiment analysis.
Innovation Solution
A system and method for learning a domain-specific and aspect-dependent sentiment lexicon using an optimization framework that combines multiple sources of information through an objective function, allowing for the construction of a unified and principled way to assign sentiment polarity scores, addressing the issue of contradictory sentiment signals and sparse information from single sources.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If a universally optimal sentiment lexicon is used, then the system can handle multiple domains, but word polarity becomes inaccurate because it is sensitive to topic domain
Solution Approach 1:
The patent segments the sentiment lexicon into domain-specific and aspect-dependent components. Instead of using a single universal lexicon, the system creates domain-specific lexicons for different topics (e.g., electronics, healthcare) and further divides them into aspect-level lexicons (e.g., battery life, screen size). This segmentation allows the system to maintain high measurement precision for word polarity within each specific domain while preserving broad adaptability across multiple domains through the hierarchical structure.
Solution Approach 2:
The patent implements dynamic sentiment lexicon selection based on the input text's domain and aspect. The system dynamically adjusts which lexicon to use depending on the context, transitioning from a static universal lexicon to a dynamic, context-aware lexicon selection mechanism. This allows the system to adapt to different domains and aspects in real-time, maintaining both versatility and precision.
2Measurement precision
If multiple sources of sentiment information are combined, then the quality of sentiment analysis improves, but there is no principled method to combine them leading to contradictory signals
Solution Approach 1:
The patent merges multiple sources of sentiment information (dictionaries, language clues, document-level ratings) into a unified framework. The system combines these diverse sources by mapping them to a common aspect-level structure and using a principled aggregation method that resolves contradictions. This merging process maintains high measurement precision by leveraging multiple evidence sources while managing complexity through a structured, hierarchical approach.
Solution Approach 2:
The patent introduces aspect-level representations as intermediaries between different sentiment sources and the final sentiment analysis. These aspect-level lexicons act as mediators that standardize and reconcile different sentiment formats and sources, transforming them into a unified representation. This intermediary layer simplifies the combination process and resolves contradictions by providing a common reference framework.
3Measurement precision
If aspect-level sentiment analysis is performed, then the detail and precision of sentiment identification improves, but the complexity of processing and lexicon construction increases
Solution Approach 1:
The patent segments the sentiment analysis process into distinct hierarchical levels: domain-level, aspect-level, and sentence-level. This segmentation allows the system to manage complexity by processing and analyzing sentiments at different granularities independently. The aspect-level lexicons are pre-computed and stored, reducing the complexity of real-time processing while maintaining high precision for aspect-level sentiment identification.
Solution Approach 2:
The patent performs preliminary action by pre-computing and storing aspect-level sentiment lexicons during an offline phase. This preliminary construction of domain-specific and aspect-dependent lexicons eliminates the need for complex real-time lexicon building, significantly reducing processing complexity during online sentiment analysis while maintaining high measurement precision through the pre-computed aspect-level representations.
Data Source
AI summary
A system and article are disclosed for objective-function based sentiment. In one example, the system includes a set of domain information, and a computer programmed with executable instructions which operate a set of modules. The modules include a sentiment polarization module for identifying a domain-aspect opinion-word pair within a set of domain data, and assigning a sentiment polarity score to the domain-aspect opinion-word pair based on an objective function which includes sentiment data from the domain information.