Where am I? Predicting user location semantics from engagement with smartphone notifications
- PDF / 2,282,796 Bytes
- 17 Pages / 595.276 x 790.866 pts Page_size
- 97 Downloads / 135 Views
ORIGINAL RESEARCH
Where am I? Predicting user location semantics from engagement with smartphone notifications Andreas Komninos1 · Ioulia Simou2 · Antonis‑Elton Frengkou2 · N. Gkorgkolis2 · John Garofalakis1 Received: 2 July 2020 / Accepted: 3 November 2020 © Springer-Verlag GmbH Germany, part of Springer Nature 2020
Abstract Location semantics are important for the delivery of context-aware ubiquitous services to users, such as the contextuallyrelevant handling of interruptions on mobile devices. For such purposes, user coordinates can be used to query global venue databases, to get back the likely venue (and its categories) where the user is located. This potentially compromises user privacy, allowing service providers to track users. We analyse data from a longitudinal study of 44 participants (university students and staff in Patras, Greece), including notification handling, device state and location information. Using semantic labels from the Google Places API as ground truth, we demonstrate that it is possible to semantically label a user’s location based on their notification handling behaviour, even when location coordinates are obfuscated so as not to precisely match known venue locations. On the other hand, the reliability of this ground truth is questioned through a crowdsourcing exercise. We demonstrate that Places API data can only be reliably used for some venue categories, and recommend best practices for using such data to establish ground truth in location context aware services. Keywords Interruption management · Mobile notifications · Semantic location labelling · Location Services
1 Introduction As users of mobile devices roam through urban environments, a wealth of data can be collected from their devices about their current whereabouts and activities. While it is relatively easy to obtain the location of a user, within a given accuracy estimate (e.g. through GPS, connection to Wi-Fi or 4G networks), a harder task is to assign semantics to the user’s location. The typical method of resolving this,
* Andreas Komninos [email protected] Ioulia Simou [email protected] Antonis‑Elton Frengkou [email protected] N. Gkorgkolis [email protected] John Garofalakis [email protected] 1
University of Patras, 26504 Rio, Greece
Computer Technology Institute and Press “Diophantos”, 26504 Rio, Greece
2
is by comparing the user’s coordinates against a database of known locations, and there are several commercial services that offer this type of information (e.g. Google Places API). Therefore, given a user’s location coordinates, it is relatively easy to obtain the venue and venue type that a user might currently be at, and therefore to infer their current activity (e.g., they are at Cinema X, and thus quite likely watching a movie). More formally, from positioning data (coordinates), one could infer various abstractions of the location semantics (e.g. the venue name, the venue type, the venue’s function, the purpose of visitation, etc.). Naturally, it’s not always
Data Loading...
