Using Social Media Posts as Knowledge Resource for Generating Life Stories

Using Social Media Posts as Knowledge Resource for Generating Life Stories Robee Khyra Mae J. Te De La Salle University Manila, Philippines robee te@dlsu.edu.ph Janica Mae M. Lam De La Salle University Manila, Philippines janica lam@dlsu.edu.ph Ethel Ong De La Salle University Manila, Philippines ethel.ong@delasalle.ph Abstract People thrive in storytelling as a means of sharing their experiences with one another and learning about the world through narra- tives. With the advancement of technology, social media platforms have become a com- mon medium for people to recount and to share stories about their life events anytime and anywhere as they happen. Automated story generation systems can be designed to utilize this vast knowledge resource. In this paper, we describe the approach we used in sourcing storytelling knowledge from social media posts, specifically Facebook, to extract events that can be used to form a life story. We also describe briefly our strategies in clas- sifying the posts based on their textual con- tent to find snippets of life events. These posts are represented in an event model that is then used by our story generator to produce a per- sonal life story that chronicles travelling, din- ing and celebrating activities. Results from user evaluation showed that Facebook data can provide enough information to generate a person’s life story which introduces him/her to others through the narration of some life events that happened in his/her life. 1 Introduction People are natural storytellers. We use stories to re- count events in our lives, and to share them with our families and friends. The advent of mobile and on- line technologies has moved this practice to social media platforms, enabling one to post their adven- tures and discoveries anytime and anywhere as they are taking place. The past decade has also seen the development of software agents designed to mimic human abili- ties in story generation. While bulk of the challenge in story generation is in planning the content and sequencing the events that will comprise the story (Solis et al., 2009; Ang et al., 2011), the amount of information available for a computer to source its knowledge from is also a major consideration on the variances of stories that computers can generate. As Meehan (1977) described it, “the lack of knowledge, often of the most mundane variety” gave rise to the production of “wrong” stories. This led researchers in story generation to employ varying strategies to provide story generators with the knowledge base they need, as described in Cua et al. (2010), Yu and Ong (2012), and Ong et al. (2018). Various studies have identified the different types of knowledge needed for story generation, which include domain-based knowledge about our world, operational knowledge on story structures and how to write stories (Rishes et al., 2013), and linguis- tic knowledge on how to express the abstract story representation into surface text in a form that hu- mans can understand. Domain-based knowledge comprises attributes and behaviors that story char- acters may embody, descriptions of objects that …

Content truncated to 500 words