Key points are not available for this paper at this time.
This work introduces a set of quality attributes for chatbots. The selection is grounded on scholarly but also reputed blog references from 2016 and 2017. In addition, attributes should be amenable to be extracted (semi) automatically. On these premises, we consider four attributes: "support of a minimal set of common commands", "foresee language variations in both inputs and ouput", "human-assistance provision" and "timeliness". These attributes are worked out for the 100 most popular chatbots in Facebook Messager. The aim is to look for correlations between these attributes and chatbot popularity in terms of number of "likes". Results show that there is no significance correlation with any of the attributes. However, the experiment come up with two main insights. First, the lack of common communication paterns that would permit users to move their experiences and expectations from one chatbot to another. Second, the existence of many programming errors that reflect that bot programming is still a nascent area.
Pereira et al. (Mon,) studied this question.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: