Inspiration
It is essential for an Enterprise and Organizer to know what everyone is thinking about their product and event respectively. With the advent of Social Media, it has become easy to analyze the sentiments of people regarding anything. Keeping this in mind, we use statistical modeling to determine the sentiment of people all around the world for a given product or event. Users can use this application to stay connected, and interact with people having similar views. Suppose, you are organizing an event, and wish to know how many people are talking about it. How many people like it or dislike it. This information can help you make some changes at the correct time in order to make the event a huge success. Similarly, an Enterprise can use this application to get automatic feedback regarding its products at any point of time. We were inspired by Prompt-App to include the text message feature using the Twilio API.
What it does
The application analyses recent tweets and categorizes them as positive or negative. Using a large dataset of pre-categorized tweets, we implemented Naive Bayes classifier to identify the sentiment of new tweets. The result can either be seen in the web application, or using the SMS interface. Users can log in using their Twitter credentials, post tweets, see latest tweets based on a # tag, and form private chat groups.
How we built it
We use data from Twitter for our analysis because of the # tag feature, which enables us to fetch tweets having a particular # tag. We used PHP as the backend language because of its object oriented nature and compatibility with many APIs. We predefined our training data in a MySQL database. For analysis of a tweet, the user has to mention a # tag. We use the Twitter REST API to fetch recent 30 tweets with this # tag. For each tweet, we tokenize its text and implement a probabilistic classifier to classify it as either positive or negative. The result is presented to the user after this analysis. The result can also be seen in a very efficient fashion using SMS. This feature was implemented using the Twilio PHP library, and the free credits provided by them. The user has to send a text message to (213) 493-8650 with this syntax: @sentiment #<tag to be searched> (for example, @sentiment #hacktech). He will receive a message showing the count of positive and negative tweets within the top 30 tweets with the given # tag.
Accomplishments that we're proud of
Integration of the SMS feature is something we are proud of. The user can just send a message to get the result on the go. This eradicates the need of logging into the website to get the overview of the analysis. The web application provides a detailed list of positive and negative tweets, and other features such as posting of tweets, and chatting with other people. The text classifier was implemented from scratch using Python and trained with over 500,000 tweets collected from various sources.
What we learned
We learnt integration of Twilio REST API in PHP, classification techniques for text based analysis, and general Web development. Most importantly, while working on this idea we realized the potential of this application in day to day life. It will change the way regarding how Enterprises and Organizers take decisions, and thereby benefit from them.
What's next for Enterprise Social Network
Integration of Call feature to enable a user to get the same information via an automated call. We are also planning to increase the number of SMS commands to satisfy more specific user requirements (such as date range to search for tweets, number of tweets to be analyzed, et al). It is just the beginning of this idea, and there are endless features that can be implemented.
Log in or sign up for Devpost to join the conversation.