Cross Lingual Sentiment Analysis: A Clustering-Based Bee Colony Instance Selection and Target-Based Feature Weighting...

Area of Science:

• Natural Language Processing
• Machine Learning
• Computational Linguistics

Background:

• Sentiment analysis faces challenges in low-resource languages due to a lack of data.
• Existing cross-lingual and semi-supervised methods suffer from poor translation quality, data sparseness, and language divergence.

Purpose of the Study:

• To propose an integrated learning model to overcome challenges in cross-lingual sentiment analysis for low-resource languages.
• To introduce a novel clustering-based bee-colony-sample selection method to mitigate translation errors and improve feature selection.

Main Methods:

• An integrated learning model combining semi-supervised and ensemble techniques was developed.
• A clustering-based bee-colony-sample selection method was proposed for optimal training data selection.
• Experiments were conducted using an English-Arabic cross-lingual dataset.

Main Results:

• The proposed model significantly outperformed baseline approaches in sentiment classification performance.
• Statistical analysis confirmed the effectiveness of the proposed training data sampling and feature selection methods.
• The approach demonstrated a performance close to that of in-language supervised models.

Conclusions:

• The integrated learning model effectively tackles language divergence and translation errors in cross-lingual sentiment analysis.
• The novel sampling and feature selection method enhances robustness against translation inaccuracies.
• This research offers a promising solution for sentiment analysis in resource-scarce linguistic contexts.