Using a web scraping tool, I obtained the text form of 50,000+ Straits Times articles. Then, I adapted the word based multi-layer LSTM model from this source code and trained a generative model. A backend script takes the first sentences of 10 latest Straits Times articles from the business section and generates 100 words based on the first sentence. I copied the style of Straits Times online articles and output html files that resembles ST articles.
Generated news articles are not actual news. It is exciting to see that the style of the articles closely resembles newspaper (perhaps even with a particular Straits Times style) reporting!