Ride Hailing Service Assessment

Scenarios

Twitter posts about sentiment and transit performance

Model and data analysis on Tweets from the New York's transit

Automatically categorize the tweets

Multi-kernel Convolution Neural Network

Model uses multiple kernels for convolution to capture local context among neighboring words in texts and is simplified by summarizing parameters in traditional models using a kernel function. Pre-trained model will soon become available for sentiment analysis / transportation evaluation.
Data Collection

The titles of popular ride-share companies like Uber and Lyft were used as keyword for filtering as the scope of this study is social media data analysis. Total of 1,925,952 Uber/Lyft relevant tweets were collected between January 23 and February 1 in 2019. Due to terms and conditions from Twitter, we cannot directly share out dataset. We are currently communicating with Twitter to make our annotated data public.
Transit Performance Category and Annotation

Traditional performance categories were reviewed and evaluated based on their applicability to the ride hailing service environment. Nine main categories were created including availability, travel time, cost, human interaction, reliability, technology, safety, vehicle quality, and community outcomes.
Visualize
Transit Tweets
New York !