Skip to content

Latest commit

 

History

History
57 lines (34 loc) · 3.81 KB

README.md

File metadata and controls

57 lines (34 loc) · 3.81 KB

Sentiment analysis performed on the transcripts and titles of youtube videos related to the game Cyberpunk 2077

License: MIT

Cyberpunk 2077

The purpose of this project is to apply Natural Languague Techniques, concretly, sentiment analysis to transcripts and description of several youtube videos.

The chosen topic is the videogame Cyberpunk 20177 and the data was scrapped from Youtube 6 days after the game was released.

Technical details about thje project:

📍 Programming language: Python

📍 Libraries: moviepy, speech_recognition, youtubesearchpython, librosa and nltk.sentiment.vader

Data scrapping process:

  1. Firstly several searches were performed in YouTube, such as: cyberpunk 2077 review, cyberpunk 2077 recommended, cyberpunk 2077 problem... There were made several of them as the package SearchVideos has a limit of videos for each search
  2. The videos were downloaded as mp4 format
  3. The audio of each video was extracted as wav format
  4. Speech recognition was performed for the first 3 minutes of each video
  • Result: 136 videos with the following features: url, description, duration, views, transcript and rating

🔗 Data extraction

Next steps:

🔗 Data description 🔗 Sentiment analysis on Transcript 🔗 Sentiment analysis on Description

Some figures:

Audio representation:

alt text

Density plot of rating feature:

alt text

Wodcloud on Description, not liked group:

alt text

Wodcloud on Description, liked group:

alt text

Results and conclusions:

- Sentiment analysis applied to the youtube platform could be an attractive analysis for some companies. Many people before buying the products review related videos on this plattform

- This kind of analysis could be also usefull for content creators, because they will be able to extract information about what the people like the most

- In some analysis such as the presented, it is enough to have the title of the video, rather than a partial transcript, however this would depend of the analysis

- The current analysis showed that after the release of the game Cyberpunk 2077, the youtube users didn't like videos related to technical aspects of the game, about the plattforms or criticism of the game. However, it should be analysed deeper in the video comments if the people didn't like the videos because they agree with the content of the video (the didn't like the game) or just because they didn't agree with the video (they like the game)

- In my opinion the people may have been curious about the game, for this reason the group of liked videos are mostly related to gameplays, about features of the game (for example cars) or speaking well about the game. This was expected, as the videos were extracted just 6 days after the release of the game, and probably a lot of possible buyers were looking for more information about the game