DeepPress: guided press release topic-aware text generation using ensemble transformers.

Guided text generation is one of the key issues when it comes to creating human-like artificial intelligence writing machines. Humans can use their writing skills depending on the topic of the text and the pieces of information they want to include. The context and style also play an important role in mediating the engagement level of the press release. However, current research does focus on conditional text continuation rather than a specific writing task. To address this problem, we propose DeepPress, a topic-aware approach that generates effective press release content when the keywords are situated in the context. We used a variety of public press datasets on specific topics to build and test our models. We show the proposed deep model achieves fine-grained control of attributes such as topics and sentiment while retaining fluency for the generated press release articles. [ABSTRACT FROM AUTHOR]