Point feature label placement for multi-page maps on small-screen devices

Highlights

• Two-phase strategy for labeling small-screen maps with multiple pages.

• Assignment of labels to pages in off-line phase via mathematical programming.

• Retrieval of labels for selected map extent and page in interactive phase.

• The user can navigate from page to page and thus retrieve many labels without zooming.

• When the user pans the map the labeling is temporally consistent.

Abstract

Map applications on mobile devices such as smartphones and smartwatches have become ubiquitous. When visualizing spatial data on such small-screen devices, one major challenge is annotating the data with labels (e.g., small icons). The restricted space requires new visualization techniques as established strategies, such as maximizing the number of placed labels, easily lead to the omission of information. We propose an approach that distributes all labels contained in a temporarily fixed map section on multiple pages. Applying interaction techniques for navigating through the pages, a user can access all information both without any overlapping labels and without the need for zooming. We propose a method with two phases; a pre-processing phase and a query phase. We use an optimization approach to pre-compute labelings on the level of a whole city and provide the on-demand querying of individual labelings at a more local level. Our approach provides a consistent label-page assignment, meaning that labels do not appear and disappear when the user pans the map. Further, our model provides quick access to important information and a spatially balanced distribution of labels on pages. In experiments on real-world data we analyze different parameter settings and show that our model yields high-quality labelings.

Introduction

In the last decade, digital maps on mobile devices have become an integral part of our daily lives. Due to their ubiquitous availability they have become an indispensable tool, e.g., for navigation or exploration within a city. We consider a user following the information-seeking mantra by Shneiderman [1] as illustrated in Fig. 1: The user starts with an overview, zooms and filters information, and is provided with details on demand. We address the need for interfaces and methods that provide access to large and detailed geographic data sets without requiring the user to zoom to an extremely large scale. In particular, we consider the user obtaining a rough overview of a city, then zooming and panning the map to a specific area of interest. From a large set of points of interest, he or she queries for features of a certain category, e.g., restaurants. These are displayed as labels, such as small text or icons. We propose an approach that can handle high information densities without producing overlapping labels and cluttered maps. Instead of showing all labels contained in the area of interest at the same time, we distribute them over multiple pages (see Fig. 2). A user can navigate through the individual pages, each of which contains a non-overlapping subset of the labels. We call a distribution of labels over a certain number of pages a multi-page labeling.

Other labeling strategies reduce clutter by limiting the presented set of labels or shifting the labels outside the map. However, we see the following problems with these strategies. Only showing a subset of labels can lead to the omission of information relevant to the user. Hence, interactive map operations such as zooming or panning are often used for further exploration. However, repetitively zooming in and out and panning the current view of the map easily causes a user to lose orientation and the overall context of the map. Furthermore, placing the labels outside the map results in less screen space being available for displaying the map. Additionally, connecting lines are necessary for unambiguous label-feature associations, which can obscure other map features (e.g., roads) and can be difficult to follow. Instead, we propose the concept of multi-page maps which are a special type of zoomless maps. This generally refers to maps that provide specialized interaction techniques allowing a user to access all information without the need for zooming [2], [3].

Preceding research on zoomless maps focuses on the labeling of temporally fixed map sections which are considered separately. Hence, when a user pans the map, two successive map sections are not labeled consistently. This easily leads to distracting effects, such as labels repeatedly appearing and disappearing. In this paper, we tackle this problem and propose a method with two phases; a pre-processing phase and a query phase. In the pre-processing phase, we use an optimization approach to pre-compute labelings on the level of a whole city. We call such a solution a global labeling. The query phase is used for the on-demand retrieval of individual labelings at a more local level that can be accessed by a user in real time. We refer to these solutions as local labelings. Fig. 3 illustrates a global labeling and two exemplary local labelings marked with Ⓐ and Ⓑ. By querying local labelings from pre-computed global labelings, we ensure that a certain label is placed on the same page for all map sections in which it is contained. Thus, when a user pans the map, a certain label is placed on the same page in two successive map sections, helping the user to keep track while exploring information.

We emphasize that we only need to compute a global labeling once. Instead of requiring a time-consuming computation for each map section, we can rapidly extract local labelings in the query phase. Assuming a server-client communication, our strategy relieves the client in terms of computational overhead. Further, local labelings can be extracted for variable screen sizes and formats without the necessity of re-computing the underlying global labeling. Fig. 4 shows exemplary local labelings both on a smartphone and a smartwatch. To get an impression of the look-and-feel of our approach, we provide an illustrative demo video in the Appendix. Using a practical setting of a user searching for an ice cream parlor on his or her smartphone, we compare our strategy to a common labeling approach and highlight the benefits of our approach.

For the computation of global labelings in the pre-processing phase, we introduce multiple criteria that each labeling should satisfy. We set up a multi-criteria model that both aims at a small number of pages and distributes labels of important features as close to the front of the sequence of pages as possible. Thus, our model provides a user with quick access to relevant information. However, as a high label density can easily lead to an information overload [4] and a reduced map readability [5], we additionally aim at a balanced spatial distribution of labels on pages. This not only reduces information overload, but additionally avoids sparsely labeled pages. We refer to the optimization problem of finding a multi-page labeling satisfying all our criteria as MapCoverageLabeling.

We solve MapCoverageLabeling using an integer linear programming (ILP) formulation. In our experiments on real-world data, we show that large instances of MapCoverageLabeling with more than 5000 features can be solved within a few minutes. We determine a sweet spot that describes a suitable balance between our considered optimization criteria. Using different parameter settings, we compute global labelings and sample local labelings of different sizes, i.e., with different numbers of labels. We compare the extracted local labelings to reference labelings that are optimal regarding our optimization criteria. We show that the local labelings provide a comparable quality with respect to the optimization criteria. Further, we show that a majority of all labels can be placed on only few front pages.

Summarizing, this paper provides the following scientific contribution. For labeling small-screen maps, we introduce a two-phase strategy that distributes labels on multiple pages. Our approach facilitates the exploration of information without the need for zooming by swiping or clicking from page to page. Contrary to previous concepts of zoomless maps, we provide a consistent label-page assignment when a user pans the map. This requires in particular a careful design of the first phase of the method, which deals with pre-computing a data structure from which local labelings can be queried in the second phase.

Following, we provide a review of related work covering both map labeling and algorithmic research on vertex coloring (Section 2). Subsequently, we introduce our optimization criteria and propose a formal model (Section 3). Using this model, we define MapCoverageLabeling. We tackle this optimization problem by means of mathematical programming and present an ILP formulation in Section 4. Further, we explain our two-phase approach in Section 5. Using real-world data from four cities in the USA and Canada, we evaluate our approach based on experiments with different parameter settings (Section 6). Finally, we summarize our key contributions and give a brief outlook on future research (Section 7).