While there is often work being done in text summarization broadly across the field of NLP, few groups are working on the specific problem of summarizing screenplays. Screenplays have a different structure than most documents and provide a unique challenge in summarization. A combination of techniques used across various summarization fields may improve outcomes in making automated summaries useful and human-readable. Recent advances in attention can help models to better understand the context of scenes so that useful information can be extracted from a large body of text. Combining extractive summarization with abstractive summarization makes human readability a priority so that the results can be more useful for proposed end users. Experimentation in this paper shows that further research needs to be done on these assumptions to make any leaps in summarizing scripts.
