Abstract:

Topic modeling of short text is an area of utmost importance these days because great amount of short text is generated every second through social media and other platforms. This short text information can be very useful if handled properly. Existing models for topic modeling use probabilistic approaches and consider frequency and coherence to discover topics from huge amount of data. These models produce problematic outputs for short text where text is not coherent and does not discuss single topic. Aim of this research is to compare the performance of different baseline models that are commonly used for topic modeling of short text and propose a context embedded topic model to produce relevant and meaningful topics from huge amount of short text data. We have used two datasets having 4,00,000 customers reviews and 2,00,000 tweets for topic modeling and compare the results of our model with already existing models through various evaluation metrics and shown that our proposed model gives more promising results when compared to existing models.

Committee: 

Dr. Asim Karim (Advisor)

Dr. Agha Ali Raza

Zoom link: https://lums-edu-pk.zoom.us/j/93746342865?pwd=ZEZmc2FvRkhyVEdpa3Y0aFFveXJZZz09

Meeting ID: 937 4634 2865

Passcode: 367262