FPT Sentiment Analysis Report

---

Table of Contents

1. Overview
2. Data Overview
3. User Analysis
4. Sentiment Analysis
   • Sentiment Analysis Overview
   • Model Used
   • Key Findings

---

Overview

This report provides a detailed sentiment analysis of user interactions on the FPTPlay platform. The analysis focuses on understanding user behavior, device preferences, and emotional responses through text-based comments. The insights aim to guide improvements in user experience and platform engagement.

---

Data Overview

• Peak Interaction Period: The highest number of comments occurred between mid-2018 and 2019.
• Active User Trend: From 2015 to mid-2019, there was consistent user activity (comments) on the platform. However, after mid-2019, interactions significantly declined, with almost no recorded activity post this period.

---

User Analysis

Device Usage

• Android Dominance: Most users interacted with the platform using Android devices (smartphones, smart TVs).
• iOS Usage: In the early months of 2015, iOS users were the most active commenters. However, Android users surpassed iOS users shortly after and remained dominant until Q1 2020.

Content Tags

• Tag Popularity:
  • The tag 'vod' had the highest number of comments.
  • The tag 'highlight' ranked second in terms of comment volume.
• Top Videos:
  • 9 out of 10 videos with the highest interaction were tagged with 'vod'.

---

Sentiment Analysis

Sentiment Analysis Overview

• Definition: Sentiment Analysis is the process of processing text data to extract human emotions (e.g., positive/negative or agree/disagree).
• Methodology: Performed using Machine Learning or Deep Learning techniques applied to Natural Language Processing (NLP).

---

Model Used

• PhoBERT: A Vietnamese language analysis model developed by VinAI.
  • Applications include word type identification, sentence completion, and more.
• Pre-trained Model: For sentiment analysis, a pre-built model trained on the PhoBERT framework was used. The model was trained on a dataset of 30,000 review comments in Vietnamese from an e-commerce site.

---

Key Findings

1. Sentiment Distribution:

   • The majority of comments were negative, as evident in the stacked bar chart analysis.
   • Negative comments dominated across most videos.

2. Tag-Based Sentiment Trends:

   • 'Event' Tag: Videos with the 'event' tag had the highest proportion of negative comments, surpassing positive comments.
   • 'VOD' Tag: Conversely, videos tagged with 'vod' had slightly more positive comments compared to negative ones.

---

Conclusion

The sentiment analysis of the FPTPlay platform highlights key trends in user interactions and emotional responses. While Android users dominate the platform, the sentiment analysis reveals that negative feedback is prevalent, particularly for videos tagged with 'event'. These insights can help the platform improve content quality, user experience, and overall engagement.

For further inquiries or contributions, feel free to open an issue or contact the author.

---