Topic Recommendation and Voice Conversation Search Based on Peripheral Information in a Long-Distance Vehicle
Taiki Funakoshi (AY 2020)
In recent years, with the proliferation of smartphones and other Internet-enabled mobile devices, more and more people are searching for information while traveling. Traveling involves transportation, and many people choose to use passenger cars, including rental cars, as their means of transportation. However, when traveling to an unfamiliar place, even if there is a Point of Interest (POI) that would be of interest to the user in a passing place, the information needs for the POI are difficult to satisfy because there are few cues for searching if the user does not have a thorough understanding of the surrounding area.
Therefore, this study aims to realize an in-vehicle information access system that recommends topics based on geographical information in the vicinity and enables users to search for detailed information by voice conversation with the user starting from the recommended information in the car during long distance travel. The proposed system was developed as an application for the Amazon Echo smart speaker. The proposed system was developed as an application for Amazon Echo, a smart speaker, and was designed to provide constant access to the latest information by using a cloud API. The system was developed by designing a system that reads location information to search for POIs in the vicinity, goes through a system to get an overview of the Wikipedia page about the searched POI, and then the smart speaker application outputs the results by voice. Users can also use the smart speaker's conversational voice search function to perform detailed searches.
After the system was developed, an evaluation experiment was conducted with five experimental participants. In the experiment, we first conducted a survey on the user's driving experience. Then, the information recommendation system was operated using the location information corresponding to the driving route, while the user watched a video of the scenery ahead taken from inside the car while driving along a pre-selected route. The users were asked to evaluate the four POIs recommended by the system in terms of prior knowledge, unexpectedness, timeliness, and satisfaction of information needs during a 30-minute driving scenario. In addition, the participants were asked to evaluate the overall usability of the system, including the notification lamps used when recommending information and the voice search function.
As a result of the experiment, it was confirmed that users do not feel uncomfortable when notifications are interspersed with information recommendations, and that the voice conversation search is useful for searching for information while driving. It was also found that users tend to feel surprised when information about unknown POIs is recommended. Future directions include the recommendation of information other than the current location, such as speed and direction, and the evaluation of extended usage scenarios. It is also necessary to investigate the effects of the proposed system on driving.
(Translated by DeepL)