We now discuss the limitation of place naming method that uses location-based lookup in SNS. Most services provide a list of nearby places since direct mapping of a user’s location to a place name is challenging. We calculated the distance between the estimated location by mobile phones and the location coordinates of places in SNS. Figure 2(a) shows that only 8.6% of places are within 10 meters, and 80.4% of places are within about 100 meters. The list of nearby places should show the places within 180 meters in order to cover the actual place for 90% of the cases. The questions to be answered, then, are how many places does the list contain within error distance? and what is the rank of the actual place in the list? If the list contains many places and the rank of the actual place is high, manual selection of a place is burdensome for some users. We statistically analyzed the location information regarding the Seoul region to answer these questions. The results show that the list contained 87.8 places within 100 meters error bound (median was 44, first percentile was 20, and third percentile was 121 places); the rank of 35% of places was higher than 20, as shown in Figure 2(b). This means that manual selection of a place in the list is a serious burden for the users, due to the high density of places. A user had to enter further keywords to find his/her current place by locationbased lookup. In addition, 17.9% of places were registered as more than two places, with slightly different names, in the location database. The reason for this discrepancy is that some users failed to find a place in the list and registered a new place instead, even though the place already existed in the database. Our findings indicate that to provide a place name, we need robust and efficient features beyond simple raw coordinates of location.
We now discuss the limitation of place naming method that uses location-based lookup in SNS. Most services provide a list of nearby places since direct mapping of a user’s location to a place name is challenging. We calculated the distance between the estimated location by mobile phones and the location coordinates of places in SNS. Figure 2(a) shows that only 8.6% of places are within 10 meters, and 80.4% of places are within about 100 meters. The list of nearby places should show the places within 180 meters in order to cover the actual place for 90% of the cases. The questions to be answered, then, are how many places does the list contain within error distance? and what is the rank of the actual place in the list? If the list contains many places and the rank of the actual place is high, manual selection of a place is burdensome for some users. We statistically analyzed the location information regarding the Seoul region to answer these questions. The results show that the list contained 87.8 places within 100 meters error bound (median was 44, first percentile was 20, and third percentile was 121 places); the rank of 35% of places was higher than 20, as shown in Figure 2(b). This means that manual selection of a place in the list is a serious burden for the users, due to the high density of places. A user had to enter further keywords to find his/her current place by locationbased lookup. In addition, 17.9% of places were registered as more than two places, with slightly different names, in the location database. The reason for this discrepancy is that some users failed to find a place in the list and registered a new place instead, even though the place already existed in the database. Our findings indicate that to provide a place name, we need robust and efficient features beyond simple raw coordinates of location.
การแปล กรุณารอสักครู่..
