Sentiment Composition Lexicon via ML Classifier
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current sentiment analysis systems rely heavily on word-level sentiment and struggle with compositional sentiment, which requires knowledge of how words interact, and often require manual intervention, sentiment-labeled texts, or domain-specific data, making them incomplete and difficult to apply across languages and domains.
Innovation Solution
A method using a machine learning classifier trained on a word-level sentiment lexicon to generate and update a lexicon of unigrams and bigrams with predicted sentiments, performing statistical analysis to extract sentiment composition lexicons, which can automatically learn sentiment composition classes without manual rules or labeled texts.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If sentiment analysis systems rely on word-level sentiment, then the system is simple to implement, but the sentiment analysis precision deteriorates because compositional sentiment cannot be captured
Solution Approach 1:
The patent segments sentiment analysis into two levels: word-level sentiment (using existing lexicons) and phrase-level sentiment (using trained classifiers). This segmentation allows the system to maintain simplicity at the word level while achieving precision at the phrase level through compositional analysis of unigrams and bigrams.
Solution Approach 2:
The patent introduces an intermediary machine learning classifier that bridges word-level sentiment and phrase-level sentiment. This classifier is trained on sentiment-labeled data and uses word embeddings to predict sentiment of phrases, serving as a mediator between simple word-level analysis and complex compositional sentiment understanding.
2Measurement precision
If manual intervention and domain-specific data are used to improve sentiment composition analysis, then the sentiment analysis precision improves, but the device complexity and data requirements increase
Solution Approach 1:
The system performs self-service by automatically training the sentiment composition classifier on available sentiment-labeled data without requiring manual rule creation. The model learns sentiment composition patterns autonomously from data, reducing manual intervention while maintaining precision.
Solution Approach 2:
The patent changes parameters by using pre-trained word embeddings (e.g., Word2Vec, GloVe) as input features instead of traditional hand-crafted features. This parameter change enables the system to capture semantic relationships automatically, improving sentiment composition analysis while reducing the need for domain-specific feature engineering.
3Measurement precision
If sentiment-labeled texts and domain-specific data are required, then the sentiment composition accuracy improves, but the adaptability across languages and domains deteriorates
Solution Approach 1:
The patent implements universality by using language-model-based word embeddings that capture general semantic relationships applicable across different languages and domains. The sentiment composition classifier learns universal patterns from sentiment-labeled texts that can be transferred to different domains, reducing the need for domain-specific customization.
Solution Approach 2:
The system performs preliminary action by pre-training word embeddings on large corpora before applying them to sentiment analysis. This preliminary preparation creates a robust semantic foundation that improves adaptability across different languages and domains, allowing the sentiment composition model to generalize better without requiring extensive domain-specific data.
4Measurement precision
If compositional sentiment analysis is implemented, then the sentiment analysis precision improves, but the computational resources and processing time increase
Solution Approach 1:
The patent applies partial action by focusing compositional sentiment analysis on bigrams (two-word phrases) rather than analyzing all possible phrase lengths. This selective approach captures the most important compositional sentiment patterns while significantly reducing computational resources compared to analyzing all phrases.
Solution Approach 2:
The system uses copying by leveraging pre-trained word embeddings from large language models instead of training embeddings from scratch. This copying of existing semantic representations reduces computational resources required for sentiment composition analysis while maintaining precision.
Data Source
AI summary
A method including: generating, from a text corpus, a lexicon of unigrams and bigrams comprising an embedding for each of said unigrams and bigrams; training a machine learning classifier on a training set comprising a subset of said lexicon, wherein each of said unigrams and bigrams in said subset has a sentiment label; applying said machine learning classifier to said lexicon, to (i) predict a sentiment of each of said unigrams and bigrams, and (ii) update said lexicon with the predicted sentiments; and performing statistical analysis on said updated lexicon, to extract one or more sentiment composition lexicons, wherein each of said one or more sentiment composition lexicons is associated with a sentiment composition class.