Master's Thesis Alina Dats

Characteristics, Applications, and Architectures of Conversational Search Systems: A Systematic Literature Review

Abstract

Humans constantly seek information using digital devices to help them in their daily lives, and research shows that they prefer to interact naturally and conversationally. Therefore, conversational search is emerging as one of the key technologies in fulfilling that need. In conversational search, a system and a user interact over several semantically coherent turns on a search task through a natural language dialog. Such interactions enable the system to understand users’ information goals or help users to clarify their needs by asking the appropriate questions directly. Conversational search thus aims to maximize the user’s information gain by finding search results with maximum utility. 

An increase in natural language dialog between users and computer systems could even lead to conversational systems replacing the prevailing interaction model of one-time keyword queries due to their effectiveness and ease of use. Understanding the design and engineering of conversational search systems is an ongoing task, and academic researchers and developers have more work to do on the theory and practice of conversational search. This research gap opens the opportunity for researchers to explore this paradigm. However, there is a lack of publications that summarize existing work and consolidate findings in this area. 

In this master’s thesis, we conduct a systematic literature review that covers 50 publications to investigate how conversational search systems can be  conceptualized and designed based on observations from the academic literature. We approach the overall problem in four directions, formulated as research questions, and discuss the characteristic properties, the suitable application scenarios, the architectures for conversational search systems, and the dependency level between scenarios and architectures. Based on the findings, we provide a conceptualization of conversational search containing feasible characteristic properties for conversational search systems. We determine suitable modalities, application scenarios, and domains for conversational search. Finally, we present a reference architecture for conversational search based on six fundamental layers. We discuss the functionalities and techniques used to implement these layers to enable the possibility of practical integrations with the help of our reference architecture.

Research Questions

1. Which characteristics of conversational search systems are defined in the academic literature?

2. What application scenarios have been investigated for conversational search systems and why?

3. What architectures have been proposed for conversational search systems?

4. To what extent do the system architectures depend on the scenarios?

Sources

• C. Khatri, A. Venkatesh, B. Hedayatnia, R. Gabriel, A. Ram, and R. Prasad. “Alexa prize—state of the art in conversational ai”. In: AI Magazine 39.3 (2018), pp. 40–55.
• M. A. Hearst. “’Natural’ Search User Interfaces”. In: Commun. ACM 54.11 (Nov. 2011), pp. 60–67.
• J. Trippas, D. Spina, P. Thomas, M. Sanderson, H. Joho, and L. Cavedon. “Towards a model for spoken conversational search”. In: 57.2 (2020). doi: 10.1016 j.ipm.2019. 102162.
• J. Trippas, D. Spina, L. Cavedon, and M. Sanderson. “How do people interact in conversational speech-only search tasks: A preliminary analysis”. In: Association for Computing Machinery, Inc, 2017, pp. 325–328. doi: 10.1145/3020165.3022144.
• F. Radlinski and N. Craswell. “A theoretical framework for conversational search”. In: Proceedings of the 2017 conference on conference human information interaction and retrieval. 2017, pp. 117–126.
• M. McTear. “Conversational AI: Dialogue Systems, Conversational Agents, and Chat- bots”. In: Synthesis Lectures on Human Language Technologies 13.3 (2020), pp. 1–251. 
• A. Anand, L. Cavedon, H. Joho, M. Sanderson, and B. Stein. “Conversational search (dagstuhl seminar 19461)”. In: Dagstuhl Reports. Vol. 9. 11. Schloss Dagstuhl-Leibniz-Zentrum für Informatik. 2020.