This item is available under a Creative Commons License for non-commercial use only
Towards the last two-decade Artificial Intelligence (AI) proved its use on tasks such as image recognition, natural language processing, automated driving. As discussed in the Moore’s law the computational power increased rapidly over the few decades (Moore, 1965) and made it possible to use the techniques which were computationally expensive. These techniques include Deep Learning (DL) changed the field of AI and outperformed other models in a lot of fields some of which mentioned above. However, in natural language generation especially for creative tasks that needs the artificial intelligent models to have not only a precise understanding of the given input, but an ability to be creative, fluent and, coherent within a content. One of these tasks is automated story generation which has been an open research area from the early days of artificial intelligence. This study investigates whether the transformer network can outperform state-of-the-art model for automated story generation. A large dataset gathered from Reddit’s WRITING PROMPTS sub forum and processed by the transformer network in order to compare the perplexity and two human evaluation metrics on transformer network and the state-of-the-art model. It was found that the transformer network cannot outperform the state-of-art model and even though it generated viable and novel stories it didn’t pay much attention to the prompts of the generated stories. Also, the results implied that there should be a better automated evaluation metric in order to assess the performance of story generation models.
Araz, K. (2020). Transformer neural networks for automated story generation. Masters dissertation. Technological University Dublin. DOI:10.21427/m9wy-px38