Earlier this week we had a masterclass on Real-Time Data Mining with Professor João Gama, which was very enriching to know a little more about the topic and what we are and will learn.
In practical classes, we decided to review the following topic.
As for the Spotify theme, we decided to abandon the idea since, with our dataset, it would be difficult to go further, especially in terms of natural language.
As for the topic of fake news, it was also no longer on the table due to another group already using a similar topic.
In the continuation of the search for the best theme we had the following ideas:
- Detect tweets that speak of real disasters – https://www.kaggle.com/c/nlp-getting-started
- Detect toxic questions in quora – https://www.kaggle.com/c/quora-insincere-questions-classification
- Question pairs in quora – https://www.kaggle.com/sambit7/first-quora-dataset
Although the 3 themes seem very interesting to us, we decided to follow the second theme (Detect toxic questions in quora), as it is, in our opinion, the most complete dataset.
Here are some of the articles we found related to the topic: