Personal storytelling: Using Natural Language Generation for children with complex communication needs, in the wild…. Issue 92 (August 2016)

• Record Type: Journal Article
• Authors: Tintarev, Nava; Reiter, Ehud; Black, Rolf; Waller, Annalu; Reddington, Joe
• Is Part Of: International journal of human-computer studies. Issue 92/932016)
• Page Start: 1
• Page End: 16
• Publication Date: 2016-08
• Subjects: Assistive technology -- Augmented and alternative communication -- Natural Language Generation -- Communication aids -- User-centered design; Human-machine systems -- Periodicals; Systems engineering -- Periodicals; Human engineering -- Periodicals; Human engineering; Human-machine systems; Systems engineering; Periodicals; Electronic journals; 004.019
• DOI: 10.1016/j.ijhcs.2016.04.005
• Languages: English
• ISSNs: 1071-5819

Abstract

Abstract: This paper describes a Natural Language Generation system (NLG), How was School Today? that automatically creates a personal narrative from sensor data and other media (photos and audio). It can be used by children with complex communication needs in schools to support interactive narrative about personal experiences. The robustness of story generation to missing data was identified as a key area for improvement in a feasibility study of the system at a first special needs school. This paper therefore suggests three possible methods for generating stories from unstructured data: clustering by voice recording, by location, or by time. Clustering based on voice recordings resulted in stories that were perceived as most easy to read, and to make most sense, by parents in a quantitative evaluation. This method was implemented in the live system, which was developed and evaluated iteratively at a second special needs school with children with different usage profiles. Open challenges and possibilities forNLG in augmented and alternative communication are also discussed. Abstract : Highlights: A system that structures personal storytelling for children with limited speech. Automatically generates narrative from sensor and other domain data. Compares different ways of structuring stories from collected data. Considers a wide range of cognitive and physical abilities. Qualitative and quantitative evaluation of the system and generated stories.