Extending a Text-to-Pictograph System to French and to Arasaac

Proceedings of Recent Advances in Natural Language Processing, pages 1050–1059 Sep 1–3, 2021. https://doi.org/10.26615/978-954-452-072-4_118 1050 Extending a Text-to-Pictograph System to French and to Arasaac Magali Norr´e CENTAL, UCLouvain, Belgium FTI, UNIGE, Switzerland Vincent Vandeghinste Instituut voor de Nederlandse Taal, The Netherlands Pierrette Bouillon FTI, UNIGE, Switzerland Thomas Franc¸ois CENTAL, UCLouvain, Belgium Abstract We present an adaptation of the Text-to-Picto system, initially designed for Dutch, and ex- tended to English and Spanish. The origi- nal system, aimed at people with an intel- lectual disability, automatically translates text into pictographs (Sclera and Beta). We extend it to French and add a large set of Arasaac pic- tographs linked to WordNet 3.1. To carry out this adaptation, we automatically link the pic- tographs and their metadata to synsets of two French WordNets and leverage this informa- tion to translate words into pictographs. We au- tomatically and manually evaluate our system with different corpora corresponding to differ- ent use cases, including one for medical com- munication between doctors and patients. The system is also compared to similar systems in other languages. 1 Introduction Augmentative and Alternative Communication (AAC) is used by disabled people to help them to communicate in daily life and to be more indepen- dent in their interactions with others (Beukelman and Mirenda, 1998). As such, AAC technologies also improve the social inclusion of disabled peo- ple, including those with an Intellectual Disability (ID), which are the focus of this paper. One of the characteristics of AAC is to repre- sent the natural language in the form of pictures or pictographs, to support the communication of people with language impairment. Several sets of pictographs have been designed specifically for people with an ID to express their basic needs to their family, friends, teachers, or healthcare pro- fessionals. Pictures can be used by these persons for communication in various situations such as for social media (Vandeghinste et al., 2015), access to school (Vaschalde et al., 2018), or in the (pre- )hospital setting (Vaz, 2013; Eadie et al., 2013). Recent research focuses on applications with pictographs for medical settings and people with or without disabilities such as foreigners and al- lophone patients. This is the case for the Smart- watch prototype of Wołk et al. (2017) and the Ba- belDr project (Bouillon et al., 2017; Norr´e et al., 2021a,b). Current systems are often limited and do not always use NLP techniques, especially the applications available for the general public. My Symptoms Translator (Alvarez, 2014; Alvarez and Fortier, 2014) and MediPicto AP-HP1 are exam- ples of mobile app for medical communication with images between doctors and patients. This article focuses on the Text-to-Picto system, which automatically translates text into pictographs for people with an ID (Sevens, 2018; Vandeghin- ste et al., 2015). The system was originally de- signed for Dutch, and later extended to English and Spanish (Sevens et al., 2015). In this work, we adapt it to French. In addition, we extend the system by linking it to a third pictograph set, namely Arasaac2. …

Content truncated to 500 words