A Natural Language Interface for an Energy System Model

Conversational Agents for Energy Awareness and Efficiency: A Survey

“…Description of the agent [24,41,[44][45][46]48,[52][53][54][55]] Data-to-text architecture description [42,43] Behavioral study [51] Framework definition [47]…”

“…Consumption monitoring [24,[42][43][44][45][46]48,49,[53][54][55]] Energy-saving advice [42][43][44]48,51,54] User-tailored [24,42,43] Smart-home automation [24,44,53] Other [41,47,52] As for the analysis of the main usage scenarios in which these systems are developed or designed, a relevant aspect that emerges is that, in the vast majority of the cases considered, the main application pertains to the domestic sphere. This seems to imply that such systems are conceived primarily to meet the needs of household users.…”

“…A significant example is the project presented in Gamage et al [45] (and in their follow-up work [46]), where the chatbot they developed was integrated within the Latrobe Energy Analytics Platform (LEAP), a platform built for the energy management and optimization of net zero carbon emissions at La Trobe University, Australia. Similarly, in the research conducted by Hulsmann et al [52] the data used to generate the feedback was retrieved from a model that simulates the operation of Germany's energy system, enabling users without technical knowledge of energy system modeling to effortlessly retrieve simulation output information, modify model settings, and then launch new simulations.…”

“…Own implementation [42,43] With ext. libraries [45,46,48] RASA [41,52,55] Google services [44,54] Other frameworks [24,49,53] The investigation into the preferred modes of communication adopted by these agents aims to complement the analysis of the development approaches, in that it draws particular attention to the favored interaction mechanisms. We point out that this aspect refers only to the modes used to interact with the conversational agents, thus excluding possible additional applications, such as web interfaces or gaming functionalities, integrated into the whole system, as in MySmartE [24] or Leafy [48].…”

“…Text [42,43,45,46,49,52,55] Voice [24,41,44,51] Multimodal [48,53,54] Finally, the role of language in interactions with agents of this type is another aspect that we believe is crucial in our analysis. Indeed, we observe that the vast majority of them adopted English as the main language, despite the fact that some systems were developed and tested in non-English-speaking countries, such as Greece [41] and Germany [49,52]. The only few exceptions are made by the GreenMoCa assistant [44] and the proof of concept presented in Ketsmur et al [53], which were developed for Spanish and European Portuguese, respectively.…”