WG 3 – Dialogue, interaction and conversational natural language generation

The global success stories, such as IBM Watson (Ferruci et al. 2010), Apple Siri, Microsoft Cortana, Amazon Alexa and IPsoft Amelia, have resulted in a new wave of research and development in the field of Human-Computer Interaction (HCI). Virtual assistants helping with simple tasks is a reality for mobile users, and are becoming more and more widespread in business applications. Even though there are virtual agent applications available in the market, these technologies are still being actively researched and are considered to be innovative near-future technologies. According to Gartner (Gartner, 2014; Gartner 2016) this smart and intelligent machine era will blossom, and it will be “the most disruptive in the history of IT”.

The dominant technique in recently created virtual assistants is the application of deep learning models that learns from texts, examples and other relevant data (e.g. Vinyls & Le, 2015; Serban et al., 2015, Li et al., 2016). However, most current solutions are offered in English, since data in other languages for training conversational agents is sparse. Moreover, models struggle with long-range dependencies, complicated dialogue structures, and only a little research exists on combining deep learning techniques with external background knowledge. This working group aims to foster collaboration between researchers interested in low-resource languages, external background knowledge and neural methods for natural language generation.

If you’d like to join WP3, please email Dr Dimitra Gkatzia at: d.gkatzia(@)

Skip to content