Cross-Lingual Text Categorization Using Social Media Analysis

This article delves into the issues regarding cross-lingual methods for text categorization and sentiment analysis tasks using multilingual social media data. Sentiment analysis ascertains the subjective thoughts or sentiments stated, whereas text categorization automatically applies predefined categories to text. Effective cross-lingual strategies that may transmit knowledge across languages are essential given the popularity of user-generated content on social media platforms that are produced in a variety of languages. We assess the effectiveness of various transfer learning techniques, such as machine translation, cross-lingual word embeddings, and recent developments in multilingual language models, using benchmark datasets spanning multiple language corpus pairs like English-Hindi, English-Telegu) for sentiment analysis and text categorization. Our experiments reveal insights into the strengths and limitations of different cross-lingual transfer approaches. Our results showed the advantages and disadvantages of various cross-lingual transfer strategies. We investigated data augmentation by translation to enhance efficiency for languages with limited resources. The findings have consequences for commercial and research activities in this domain and lead to the development of strong cross-lingual content analysis algorithms for the multilingual social media landscape.