Unsupervised News Topic Modelling
In the sea of information nowadays, it is difficult to find customer behavior pattern that can be exploited for industrial benefit. In a joint effort with Eureka Analytics, we are attempting to address this problem by developing an unsupervised topic modelling for news. This algorithm generates topics from news that is read by telco customers. Afterwards, Telco company can use the generated topics to understand their customers behavior. The understanding can lead the company to a better targeted marketing.
In parallel, we are also trying to discover novel algorithm to automatically construct knowledge graph from the embedding of generated topics. This approach is intriguing in the eye of science, as the concept can integrate deep learning into creating knowledge graph.