study,author,year,venue,approach,key_finding,gap_addressed_by_this_work
Sentiment Analysis of Code-Mixed Social Media Text (Hinglish),Gaurav Singh,2021,arXiv,early Hinglish sentiment baselines,accuracy hurt by transliteration inconsistency and token errors,this work uses dedicated preprocessing + multiple embedding/hybrid models for code-mixed Hinglish
Current State of Hinglish,Varsha Thakur et al.,2020,SSRN,survey of traditional models,limited contextual depth and poor scalability,this work systematically benchmarks hybrid deep learning across language strategies
Improving Sentiment Analysis,Neha Agarwal et al.,2024,Educational Admin: Theory and Practice,hybrid deep learning,hybrid DL improved accuracy over classical ML,confirms and extends with GloVe/Word2Vec/FastText + BiLSTM and multistage training
A self-Attention hybrid model for code-mixed language,Gadde Satya Sai et al.,2022,Social Network Analysis and Mining,contextual embeddings,contextual embeddings outperform CNN/RNN,this work isolates embedding choice within a BiLSTM hybrid for hate-speech
This work (Biswas contribution),Pankaj Biswas,2026,B.Tech Project (RSET),GloVe/Word2Vec/FastText + BiLSTM and many-to-one LSTM with multistage language training,GloVe+BiLSTM multistage combined reached F1 0.8041 / AUC 0.9139 on PRISM,n/a (this is the contribution)