Robbi Zulcharnain, Ginanjar Abdurrahman, Daryanto Daryanto

Analisis Sentimen Ulasan Duolingo dengan Metode Algoritma Multinomial Naïve Bayes

Introduction

Analisis sentimen ulasan duolingo dengan metode algoritma multinomial naïve bayes. Analisis sentimen ulasan Duolingo di Google Play Store menggunakan Multinomial Naïve Bayes dan K-Fold Cross Validation. Model mencapai akurasi 89.69% untuk sentimen pengguna.

Abstract

Duolingo menawarkan solusi efektif dengan menghadirkan tantangan, permainan, dan berbagai level menarik yang menjaga minat pengguna. Namun, seiring meningkatnya popularitas, platform ini menjadi subjek memuat terkait pendekatannya terhadap representasi LGBT ( Lesbian Gay Biseksual Transgender ) sebagai langkah penting menuju keterlibatan yang lebih besar dalam ruang pembelajaran, sementara yang lain merasa bahwa masalah ini memicu perbedaan pendapat. Penelitian ini bertujuan untuk menganalisis sentimen pengguna terhadap aplikasi Duolingo di Google Play Store dengan metode K-Fold Cross Validation dan Multinomial Naïve Bayes. Metode K-Fold Cross Validation menghasilkan nilai tertinggi dengan akurasi 89.69%, presisi 0.89, recall 0.92 dan f1-score 0.90, yang menunjukkan performa model yang konsisten. Penerapan Multinomial Naïve Bayes pada data menunjukkan hasil yang sangat baik, dengan akurasi 89%, presisi 0.86, recall 0.91, dan f1-score 0.89. Hasil ini menunjukkan bahwa metode ini efektif dalam menganalisis sentimen. Pemilihan Duolingo sebagai objek penelitian didasari oleh penerimaan sertifikat di Universitas ternama di luar negeri.

Review

This paper presents a timely analysis of user sentiment towards Duolingo, a popular language-learning application, employing the Multinomial Naïve Bayes algorithm in conjunction with K-Fold Cross Validation. The decision to focus on Duolingo, particularly in light of discussions surrounding its approach to LGBT representation, introduces an intriguing dimension to typical sentiment analysis, moving beyond general user satisfaction to explore perceptions on a more nuanced social issue. The chosen methodologies are standard and well-suited for sentiment classification tasks, and the abstract reports promising performance metrics, suggesting a robust foundation for the study. A significant strength of the research lies in its rigorous methodological approach, particularly the application of K-Fold Cross Validation. This technique ensures a more reliable and consistent evaluation of the model's performance, as evidenced by the high accuracy (89.69%), precision (0.89), recall (0.92), and F1-score (0.90) reported for the cross-validation process. The subsequent application of Multinomial Naïve Bayes also yielded excellent results, with an accuracy of 89%, precision of 0.86, recall of 0.91, and F1-score of 0.89. These metrics clearly demonstrate the effectiveness of the selected algorithm in classifying sentiment within the Duolingo review dataset, affirming the study's technical soundness. While the abstract effectively highlights the robust performance of the sentiment analysis model, a key area for improvement in the full paper would be a clearer presentation of the *actual sentiment findings*. The abstract primarily focuses on the model's accuracy and validity, rather than the distribution of sentiments or specific insights derived from the analysis concerning user opinions on Duolingo, especially regarding the LGBT representation aspect initially mentioned. Furthermore, the rationale for selecting Duolingo based on "penerimaan sertifikat di Universitas ternama di luar negeri" is somewhat ambiguous and could benefit from further clarification on its relevance to sentiment analysis. The study has strong methodological underpinnings, but the value of a sentiment analysis paper ultimately lies in the interpretability and implications of the sentiments themselves.

Full Text

Comments

You need to be logged in to post a comment.