Twitter communities are short-lived and primarily formed around the trending hashtags. Once those hashtags are no longer trending, the communities die out and it’s difficult to stay connected with the people you once tweeted, retweeted, and had meaningful conversations with. Thus, having mental health in mind to support the users by comparison program.
What it does
When a user searches for a hashtag, they can access tweets, photos, and videos around that topic. They can also see the users who are a part of that community. Currently, the user needs to have the community name in their bio/name to be a part of that community. However, not everyone wants this information made public. Also, having the community name in their name limits them to the number of communities they can be a part of.
A new feature that Twitter recently launched is called “Moments”. Users need to create their own Moments with tweets that they like. We plan to use this feature and make it more automated. Users can be part of communities by tweeting, retweeting, and saving tweets with a particular hashtag. These communities that users choose to be a part of can be automatically added to their Moments page. They can continue to post new tweets and interact with the members even after the hashtag is no longer trending, allowing them to set the visibility of their tweets to only the members of that community.
How we built it
Thanks to the Twitter APIs' capability to extract out tweets, users' handles, and its location. Which was then uploaded to the local website. Thus, building up documents of tweets to vectors in order to convert them into tokens. This brought us to term frequency, after inverse document frequency, and then through CountVectorizer, we can count the number of words that occur the most, but yet excluding stopwords. Filtering them to words that occur the most, then we went for Sentiment Comparison using the average occurrence of sentiments, and getting help from TextBlob library.
Challenges I ran into
Extracting was a challenge, thanks to Twitter API for its access to us. The challenge was to remove most occurring words that were quite common those were stopwords. Sentiment Comparison was a challenge as there many sentiments parts of speech in tweets.
Accomplishments that I'm proud of
Proud to be able to access to Twitter API and use it for our research. Able to find the frequent words that are mostly occurring and that makes sense. Sentiment Comparison is what makes us proud to find the words that are most occurring and understand user sentiments based on it and their preference.
What I learned
Use of API, Learned about Framework django, nltk, transformers, sklearn, textblob, statistics and made use of them that can bring best results.
What's next for softWEare
Based on the categorization of tweets we can filter out trending words and associate them with tweets they are related to and then depending upon the sentiments of the user we will be displaying back the trending hashtags and words to the user for his use to bring change and make use Twitter not only for a better world but also for awareness.