A Large-Scale Analysis of Mixed Initiative in Information-Seeking Dialogues for Conversational Search

Conversational search is a relatively young area of research that aims at\nautomating an information-seeking dialogue. In this paper we help to position\nit with respect to other research areas within conversational Artificial\nIntelligence (AI) by analysing the structural properties of an\ninformation-seeking dialogue. To this end, we perform a large-scale dialogue\nanalysis of more than 150K transcripts from 16 publicly available dialogue\ndatasets. These datasets were collected to inform different dialogue-based\ntasks including conversational search. We extract different patterns of mixed\ninitiative from these dialogue transcripts and use them to compare dialogues of\ndifferent types. Moreover, we contrast the patterns found in\ninformation-seeking dialogues that are being used for research purposes with\nthe patterns found in virtual reference interviews that were conducted by\nprofessional librarians. The insights we provide (1) establish close relations\nbetween conversational search and other conversational AI tasks; and (2)\nuncover limitations of existing conversational datasets to inform future data\ncollection tasks.\n