Foundations of Map-based Web Applications
A Survey of the Use, Limits and Opportunities Offered by Digital Maps
Alessio Antonini
1
, Guido Boella
1
, Stefania Buccoliero
1
, Lucia Lupi
2
and Claudio Schifanella
1
1
Dep. of Computer Science, University of Turin, Corso Svizzera 185, Turin, Italy
2
Interuniversity Department of Regional and Urban Studies and Planning,
Polytechnic of Turin and University of Turin, Turin, Italy
Keywords: Urban Informatics, Digital Maps, Web Applications, Data Visualisation, Urban Entities, Urban Computing.
Abstract: Traditional maps are one of the oldest way to express relevant information on a locality base, as synthetic
representations of reality. The traditional visualization theory of maps and the related principles used to
structure spatial information can inspire the modelling of new solutions in the field of information
management in web application. But, the fast and generalized spreading of digital maps, and the related
production of geo-localized social media is not followed by a deep integration of map in web applications,
preventing the effectiveness of digital maps in solving pressing issues like aggregation, retrieval,
recommendation and presentation of spatial media. Through the analysis of key concepts of maps, this
contribution addresses the foundations of map-based applications, discussing the limits of current approaches
and introducing new opportunities based on deep integration between maps and applications.
1 INTRODUCTION
Traditional maps are one of the oldest way to express
relevant information on a locality base (Shahaf,
2013), combining and defining different semantics
using visual signs (MacEachren, 1995), in order to
present a synthetic representation of reality at
different scales and organizing data in a structure
defined by the purpose of the map (Keates, 2014)
(Okada, 2014). The traditional visualization theory of
maps and the related principles used to structure
geographical information can inspire the modelling of
new recommender systems for online tools, based on
the integration between data structure and interaction
of applications and digital maps.
Nowadays, the fast and generalized spreading of
digital maps, and the related production of geo-
localized social media, is due to the intuitiveness of
maps in representing reality. Digital maps often
complement web applications, even thoes
applications not specifically related to geographical
purposes such as social networks, public platforms or
thematic portals. Digital maps provide a spatial
context, an alternative query system and a
visualisation tool to spatial media.
Nevertheless apart from web geographical
information systems (Web GIS), digital maps are not
“fully” integrated in web applications, in terms of
features, interaction and information management.
The decoupling of maps and applications generates
ambiguities and counter intuitive behaviuor in web
applications, leading to segregating digital maps to be
uneffective “add-ons”.
Maps are used as “canvas” to present spatial
media, and to help users in selecting an area of
intereset – a bounding box is a section of the canvas-
and therefore a subset of application entries. The main
use of digital maps is to let users interact with the
system to navigate a surface (a world map), changing
scale and area of interest, and visualizing at the same
time extra sources of information as points of
interests (POIs) or other geometrical features.
A light-weighted use of digital maps does not
require any understanding of maps as knowledge
theory, because limited to the visualization of spatial
features mostly collected automatically (implicitilly)
through mobile devices. Current web applications
including maps cannot be considered “map-based”
since not exploiting map features in the information
management: representation, retrieval,
recommendation and visualisation.
This contribution enquiryies how the theory about
of maps can be reflected in web applications, how a
map-based application should integrate maps
92
Antonini A., Boella G., Buccoliero S., Lupi L. and Schifanella C.
Foundations of Map-based Web Applications - A Survey of the Use, Limits and Opportunities Offered by Digital Maps.
DOI: 10.5220/0006136600920099
In Proceedings of the 12th International Joint Conference on Computer Vision, Imaging and Computer Graphics Theory and Applications (VISIGRAPP 2017), pages 92-99
ISBN: 978-989-758-229-5
Copyright
c
2017 by SCITEPRESS Science and Technology Publications, Lda. All rights reserved
extensivelly, which are the limits of common
approaches and new opportunities. Specifically, maps
are analysed from multiple perspectives: data
visualisation, filtering, aggregation and
recommendation:
How to extend visualization mechanisms used
to create maps from data sets of geographical
features to “points of interest” (POIs)
How to extend map scale theory to the
management of POIs
How to use geographical entities in data
aggregation
How to take in account users’ interaction with
a map in data collection
How to build a geographical recommender
The theoretical concepts behind the definition of
maps are demonstrated to be relevant to provide a
meaningful and coherent context for spatial media.
The relevance theory within the implementation of
scales in cartography is presented as solution of the
issues about ambiguities of spatial feature of social
media and filtering, leading to the revision of the
current mataphor behind the interaction and
information management with digital maps.
The solution to the ambiguities in POIs collection
open the path to overcoming the current solutions in
data aggregation and filtering, defining suitable
alterntive to popularity ranking, and spatial
clustering. Moreover, the strong connection between
POIs and geographical entities a identification of the
connection between POIs and geographical entitie
open the possibility to new multiple solutions.
The rest of the paper will discuss the theoretical
and technical steps needed after a state of the art and
followed by a discussion of open issues, further
research topics and conclusions.
2 DIGITAL MAPS
In web applications, digital maps provide the
geographical context of spatial media: the
visualisation as POIs of their position (spatial feature)
on a cartography. Also, digital maps support
exploring POIs with dialogue popups - clicking on
markers – and interacting with the map viewer
changing viewer bounding box and scale to change
the retrieved data.
Digital maps can be used to entry geographical
features such as positions, in alternative to using
addresses or GPS. But, the decoupling between map
status and the data leads to ambiguous entries, hard to
process automatically and to present to users (Figure
1).
The integration of digital maps in web
applications opens issues related to the heterogeneity
of crowd-data and design tools able to support
unexperienced users without a common ground. The
web domain forces toward intuitive and pragmatic
tools that can be used by everyone without any
training or guidance, and at the same time archiving
an overall good quality of the result.
Following, we are going to analyse current uses,
limitation and constraints of digital maps, outlining
alternative solutions and new approaches.
2.1 Information Management Systems
Nowadays, the importance of user-generated contents
(UGC) is continuously arising in many fields, and
crowdsourced unstructured and/or incomplete
dynamic data with spatial attributes need to be
organized and analysed in order to leverage their
immense value.
Web-GIS (web geographical information
systems) are a class of applications focused on
working with geographical information, providing
tools to define geometries (paths, polygons, lines,
multi polygons, points). Web-GIS are often supported
by volunteering and crowd-based geography
movements, and used to build big datasets of
geographical information where there is a lack of
official sources. The major project in this regards is
OpenStreetMap (OSM). Web-GIS specific focus are
geographical data, thus they cannot be bended to
social media where the geographical aspects are
important but marginal and with they do not share the
same efforts in terms of correctness of entries,
methodology, features, required training, etc.
Social media such as Facebook and Twitter, make
use of spatial features as fundamental part of their
strategies to handle crowdsourced big data. A crucial
part of the success of those applications is their effort
on usability: intuitive interactions, smart defaults,
action feedbacks, personalised interfaces, etc. The
result is archived combining any means like user
profiling, reasoners and domain specific
recommenders. Geographical profiling is the natural
consequences of adding geographical features to
social media.
2.1.1 Limitations
The concept of scale remains implicit in digital maps
and in location based social media. In general, what
you see on a map the application is not aware of and
therefore is not what you get interacting with the
application. Without considering the scale, the
Foundations of Map-based Web Applications - A Survey of the Use, Limits and Opportunities Offered by Digital Maps
93
organization of information does not reflect the user
experience, mostly oriented to search for different
sets of information at a neighbourhood, urban scale or
territorial scale.
Scale is a broad concept investing technical
schemas, urban plans, public strategies programmes
and many other fields, and it is referred to the need of
visualising different cluster of data in different scales
for specific purposes. This concept translated in
digital maps leads to do not aggregate information
considering the geometrical proximity of POIs based
on their coordinates, but their relevance at each scale.
Another aspect to consider in current digital maps
is that the spatial features of social media are mostly
ambiguous and under specified. For instance, what is
the meaning of a media position? A user is visiting
something or it is just passing by? Is the posted
content related to where a user is or from somewhere
else? In case of complex environment like shopping
malls, where the user actually is? There is a huge
effort in disambiguating this information but a major
contribution can come directly on less ambiguous
interfaces enabling users to specify where and what
they are doing and the scale of reference. In Figure 1
for instance, the same input (position) may have
different meaning according to the view scale,
problem that does not rise using address as reference
system.
Figure 1: A position changes meaning at different scale: the
marker is referring to Campidoglio square (a) and to the city
of Rome (b).
Furthermore, a disconnection between the
visualization state of the map and the application
status heightens the ambiguity of scale in user
interactions with digital maps. In other words, the
change of scale associated to a change of zoom level
is limited to the cartographic base, but it does not
change the status of the application in terms of
retrieval and visualization mechanisms.
2.1.2 Alternatives
Which aspects of the concept of scale should be
considered for web applications? In general, it is
possible to consider three kind of scales: cartographic
scale, geographic scale, and operational scale. The
cartographic scale address the relationship between a
distance on a map and a distance on the ground. The
geographic scale reflects the geographical structure of
social interactions and size, level and relations define
it. Operational scale corresponds to the level at which
relevant processes operates and include the traces of
actions belonging to the same chain of activities
(Lam, 1992) (Smith, 1992) (Howitt, 1998).
OpenStreetMap (OSM) is currently the biggest open
geographical data source available, and OSM’s
definitions of spatial entities are mostly focused on
space functions and uses (Marston, 2000). Therefore,
applications features relying on geographical data
should consider what is available beforehand
(operational and geographical scales).
In digital maps:
The size is controlled by users through zoom
controls
The level of focus - local, national, regional
is bounded to what is considered relevant at
each scale, from buildings to the city level, and
usually is related to a bounding box, defined by
two points, the South-West North-East
extremes of the map window
OSM entities are rendered as geometries
according with their size and vertical ordering
Names are rendered as labels over geometries
The connection between the state of the map and
the state of the application can be archived reflecting
the change of scale within the application. Changing
the zoom level changes the cartography but not the
data on the map. There is not an explicit mapping
between the zoom levels and scales which can be used
to create a dynamic visualisation system.
The status of a map is not limited to the bounding
box but also to geographical units currently visible,
ranging from administrative entities as for instance
cities or districts, to informal boundaries as
neighbourhoods, to spatial units as building blocks,
public spaces or buildings (Figure 2). The concept of
geographical unit is connected again to the scale, as
the scale change the relevant units change. This
mechanism unveils the existence of a “focus layer” of
entities which are significant - and highlighted in the
cartography at - changing as the map status changes.
Connecting POIs and geographical unites is therefore
possible to filter the information as the users interact
with the map, in a more comprehensive way.
Furthermore, key concepts of a map implemented
with the map theming, symbols and legends can be
extended to information management and vice versa.
A map legend can be effectively reflected in the data
organisation and for instance, a category system can
be used to generate map theming dynamically.
HUCAPP 2017 - International Conference on Human Computer Interaction Theory and Applications
94
(a) (b)
Figure 2: Example of geographical units considered for the
indexing system: building (a) and square section (b).
In digital maps, theming and scales are
implemented into styling rules defining what should
be rendered and how - colours, labels, hatches, lines,
symbols, vertical order - during the map generation
process. Following an example of styling rule about
how to fill the buildings polygons from CartoCSS in
Mapbox (https://github.com/mapbox/carto/):
[
class='street’][zoom>=9]{
line-pattern: @small;
[zoom>=13]{
line-pattern: @medium;
}
[zoom>=16]{
line-pattern: @big;
}
}
In this example, the CSS rules {…} are valid for
the interval between 9 and 22 (max zoom possible),
from the city to the indoor level. Following the
previous example, streets do not have a style for
levels higher than 9 and consequently not rendered
(even if the information is still there). The inner rules
override the styling of streets accordingly to the zoom
level. At 13 and 16 there is an increment of the size
of streets. For instance, the vertical ordering - a street
overlapping a river – is also encoded by styling rules.
Rules are applied to labelled spatial entities at
rendering time, but with current vector maps it is
possible to change them dynamically. The logic
within styling rules can effectively extracted and
generated, enabling a two-way connection between
cartography and application.
Summarising, as the map changes POIs can be
filtered, and as users exclude/include contents with
other tools the map theming can also change.
Moreover, the theming can be connected to
application organisation system such as categories
and vice versa.
2.2 Data Viewers
Technically speaking, digital maps are web viewers
for pyramid images (cartography). The cartographic
bases are images (raster, svg) or data sources
organised as “tiles” in layers. A tile is a bounding box
of fixed dimension, identified and retrieved by a
triple: z for the layer, and x and y for the area. A map
viewer is a tool to identify and retrieve tiles,
converting the bounding box and zoom level in triples
and to present tiles and layers of geographical
features on a spatial canvas. POIs are represented by
their geographical feature (point) in overlay layers
above the tile layer.
2.2.1 Limitations
One of the most common limit of digital maps is the
lack of control over the visualized information. The
visualisation of data is done pushing data in overlay
layers, which are implemented as arrays. The map
viewers usually provide control over the visualisation
of layers but not about the contents of layers beside
the bounding box. For instance, dynamic information
can make maps overcrowded and impossible to be
used, which is in contrast with the very purpose of
maps in general: synthetic representations based on
hierarchical representation of information and
selection.
Figure 3: Aggregation approaches: (left) marker clustering,
(centre) heat map, (right) grid-based aggregation.
The problem of overcrowded maps is addressed
directly by applications through filtering,
recommendations, and ranking (ordering solution) in
conjunction with specific solutions for digital maps,
like spatial clustering (Figure 3). Spatial clustering
creates dynamic clusters of markers accordingly to
their distance, ignoring geography of their location
like physical barriers and administrative boundaries
(Figure 4), and other distances like street routes.
Ordering solutions are used in combination with
limits on the results, e.g. top 100. The limit that can
be global (in a bounding box) or local to each section
of the map (tiles). It can be implemented hiding
marker or using marker of different dimensions.
Foundations of Map-based Web Applications - A Survey of the Use, Limits and Opportunities Offered by Digital Maps
95
Figure 4: Marker clusters take in account the distance (left)
despite any physical obstacles like a river (right).
2.2.2 Alternatives
Again, the limits of the mentioned approaches are due
to not considering maps as knowledge systems.
Considering data structures of maps, the projection of
a subset of geographical features should be done
coherently with a synthetic theory implemented in
theming and scales. In other word, the data
visualisation can be based on accordingly to the
hierarchy implemented in the current scale and on
what is currently visible in the cartography. To bridge
maps and applications views it is required to use the
same rules in generating maps within the application.
Alternative solutions to overcrowded maps can be
built on top of the visualisation of geographical
entities on maps, in terms of type like buildings,
administrative areas, parks. The connection between
POIs and geographical entities can be used to transfer
properties from geographical entities to POIs and vice
versa. For instance, the hierarchical organisation and
vertical alignment of geographical entities can be
reflected on data overlay layers in alternative to
highlighting popular contents.
2.3 Recommender Systems
Digital maps are currently used as location-based
filtering systems, instead recommendation is mostly
based on users’ spatial profile or their current
location. But current web technologies allow to
generate maps almost runtime, applying dynamic
themes and even manipulating map data sources. Can
dynamic maps support recommendation?
POIs recommendation is still a major issue in
location-based social networks. POIs have spatial and
temporal dimensions, making this domain hard for
common recommender (Griesner, 2015). The
recommendation of POIs should also be user-based,
location-aware, and context dependent (Liu, 2013).
There are several possible approaches to build a
geographical recommender system:
to combine temporal and spatial dimension of
POIs (Griesner, 2015) (Lu, 2015) and to use
time/location data of users;
to exploit POIs content, considering associated
textual and context information (Liu, 2013);
to collect data about the interactions of users
with digital maps and update their profile with
category preferences and locations (Oku,
2010).
to extend the friendship-based approach to
locations (Ye, 2010), considering the emerging
correlations between friendship and favourite
places.
2.3.1 Limitations
Users’ geographical profile is not always a reliable
source of information, in particular where users can
access the same platform for very different purposes,
like in case of public portals and civic platforms. The
users’ current goal is unknown and it cannot be
inferred especially in multi-thematic and
multipurpose platforms. For instance, the reasons
behind selecting a place, planning an activity,
enquiring for a service or trading a good is commonly
implicit and overlooked, and not reflected on map
views.
In this context, users have a passive role – without
control - in determining which contents are displayed,
and recommended contents based on users’
geographic profile is just the result of an aggregation
of preferences associated to spatial features.
2.3.2 Alternatives
Map theory again can support the definition of new
geographical recommender system, which do not
require users’ geographical profile, but can exploit
the geographical types in combination with users’
activities, giving control to users on what they want
to see case by case.
Traditional maps are “a priory” recommender
systems, where readers could see only a specific set
of information, accordingly to what they need,
selecting the right kind of map, in terms of scope,
scale, purpose/theming.
Currently, digital maps open a wide range of
approaches, providing the possibility to create
multiple maps from the same data source in term of
theming and scales. For instance, it is possible to
settle a general goal “traveling” and different
approach at each scale “walk”, “bike & car”, “bus &
train”, “airplane & boat”, generating very different
vires.
The introduction of an indexing system for POIs,
using geographical units connected to the zooming
system results in: 1) providing a tool to users to
explicitly control the results and 2) extending the
recommendation aspect from maps to geographical
units and then to contents.
HUCAPP 2017 - International Conference on Human Computer Interaction Theory and Applications
96
Interactivity is the main features of digital maps,
enhancing the value of users’ inputs on maps enhance
the effectiveness of the map as tool without raising
the overall complexity of a web application.
Users’ needs play a central role to understand
what connects their favourite locations and their
activities (Guy, 2015). In this perspective, we propose
to start considering users’ actions, in alternative to
systems based on users’ spatial profiles. Allowing
users to control the outcomes and introducing a
concept of users’ contingent goal/task consent to
provide a tailored support to their current activities. A
goal is something that can be handled by users
directly and explicitly reducing the premises to how
best support specific user activities rather than
guessing users’ interests.
As result, maps can be shaped dynamically
accordingly to users’ tasks, for instance removing
useless entities and highlighting a work area like a
district or neighbourhood.
3 MAP-BASED APPLICATIONS
What are the features of a map-based application? An
application can be considered map-based if it exploits
the core mechanisms of maps in such a way that
interacting with a map is interacting with the
applications’ entries and as the map status changes
the application status changes and vice versa. Thus,
the fully integration between map and application is a
requisite to provide an intuitive tool, improving the
overall capability to manage complex scenarios such
as social networks, or any big data, relying on the
geographical features of data.
A map-based application should implement the
principle “what you see is what you get”
(WYSIWYG), solving the ambiguity of data input,
and the decoupling between map view and
application data. In this regards, not all web
applications using maps should be map-based. For
instance, hotel web applications do not need to
provide any other information aside hotels, therefore
there is no need to consider scales or implementing
dynamic theming.
Summarizing, a map-based application should
support the following mechanisms:
The connection between application contents
and map entities, that each spatial media is
linked to a geographical entity;
The connection between map status and
application status, or rather actions in the
application should be aware of the current
status of the map in term of scale, area and
available entities for users.
3.1 Information Management
What does it mean to connect map and contents? As
general principle, connecting a map to its content
means to show, hide and highlight accordingly to the
map goal. For instance, a political atlas shows nations
despite their size, population or political influence,
and nations are usually coloured to highlights their
boundaries, but there are no orographic references or
any other element describing their morphology. A
political atlas is not meant to describe orography nor
morphology of countries, but to show the physical
relations among them. Conversely, in physical atlas
countries with similar territory can be easily confused
– there is not a neat distinction of national borders -
because the aim of this kind of representation is to
highlight natural element characterizing the territory.
Maps produced for different goals contain and show
different elements.
Considering another perspective, road guides are
books for traveling by car, usually organized in local
and national tables. National tables show main roads
such as highways and main dorsal railways. Local
tables contained also secondary and local streets.
Even if local and national tables are intended to show
similar entities (streets, railways, paths) and they
looked very similar, their content is different. Even in
this case, maps select information for the user,
considering different goals: travel by car to reach a
place or to explore and find a specific address of a
place. The representation theory about scales is
directly linked to theories of relevance about map
contents.
Digital maps allow users to change the map zoom
to observe different portions of space, but this kind of
interaction does not necessarily affect the map scale.
In fact, there is not an explicitly connection between
zooming and scale because scales are usually outside
the scope of applications, encoded in “tile servers”
(map providers). Even if a map switches from a zoom
level to another does not mean that map contents
change accordingly, this leads to overcrowded maps.
Making explicit the connection between map and
content management will reproduce the same use of
physical maps, the choice of the right map for the
current use but seamlessly. This requires to build a
data structure based on the data source of maps,
which can be possible only with crowd-based and
volunteering geographical data sources.
Foundations of Map-based Web Applications - A Survey of the Use, Limits and Opportunities Offered by Digital Maps
97
3.2 Connecting Maps with Applications
Why are POIs cannot be connected to geographical
entities in current applications? Digital maps are
considered “flat” even those they are not. The data
layers and POIs are “pinpointed” on maps, without
any regard to what lies behind. The current metaphor
is using maps as pincushions, without any connection
between map and markers (pinpoints).
Pinpointing a marker on a map is maybe the
easiest method to add the spatial dimension to
contents, and the practice demonstrates that so far is
a solid metaphor when comes data collection. In fact,
very few non-geographic systems support anything
more than markers.
But in terms of other tasks we may find a more
suitable metaphor considering again the data structure
behind maps and how users interact with them. Rather
than pinpointing the action of placing POIs is
“dropping” them on geographical entities. What is the
difference between pinpointing and dropping? Even
if technically raster maps are flat, we can assume
users placing POIs considering a structure of ordered
elements they saw: streets are above rivers, buildings
are above parks, etc.
Spatial entities work as castle of “buckets” one
inside another, in respect to their size and vertical
ordering. POIs are falls inside these buckets because
users can “see” where they place them. Applications
not aware of what users are interacting with are
disregarding partially the meaning of users’ actions
with the consequence of losing the essential
information and failing in representing correctly
users’ contributions.
Considering what users interact with saves a lot
of issues and computation, and it enables to build a
reliable connection among users’ contents and
geographical data by design.
3.3 Applicative Solutions
The requisite is the connection between POIs and
geographical entities instance. It can be archived
introducing a specific geographical index recording
the relations as users generate contents (Antonini,
2017). The indexing system requires the definition of
focus layers, a coverage made of geographical units,
for each defined level of the index.
As users change scale the application can load the
relative focus layer, to build features based on the
understanding of what users can see and can interact
with. It enables to:
Dynamically change the map theming
Filter POIs which are not related to the
geographical unites in the current focus layer
Enhance the entry of data on maps, from points
to “deep” points referring to a specific scale
Aggregating information accordingly to the
geographical entities they belong to (Figure 5)
Rank POIs accordingly to the focus layer
The WYSIWYG principles can be archived. For
instance, a POI about a building is retrieved if the
building is retrieved, not only in terms of bounding
box but in terms scale.
Figure 5: POIs connected to geographical units can be used
to generate area-based aggregates, preserving the coherence
with the geographical data as users see them on the map.
The system provides support to synthetic
representations, saving computational costs and
relative approximations. It can have great impact on
performances, readability and intuitiveness, even
with critical issues in web applications related to
crowd sourced big data.
In terms of recommendation, it is possible to
exploit the “type: of geographical entities to build
users’ “focus area” within the bounding box.
In Figure 6 the results of a first experiment. Using
a simple flooding algorithm, the users’ position is
expanded on a weighted visibility graph connecting
the visible geographical units. At users is given the
control of a “task” selector, where to express their
current activity. Tasks correspond to weights array
indicating the cost of expanding the area accordingly
to the “relevance” of the type of geographical units.
For instance, during “sport” crossing a street “costs”
much more than traversing a long park. The
expansion of the focus area was bounded to two
criteria, a maximum size of 50% of the bounding box
or the inclusion of 50% of POIs within the focus area.
In Figure 6 the shape outline of three areas
generated for the three tasks, starting from the same
dataset of geographical entities and POIs.
Figure 6: The outline of three focus area corresponding to:
(left) shopping, (center) education and (right) sport,
generated from the same setup: user position, POIs and
geographical data.
HUCAPP 2017 - International Conference on Human Computer Interaction Theory and Applications
98
The use of focus area to select geographical
entities and therefore POIs, is an alternative solution
to other recommenders not requiring user profiling.
4 CONCLUSIONS
This contribution addressed how the theory about
maps can be applied to web applications to archive
deep connection between data and interactions from
maps to applications and vice versa.
The revision of current approaches and their
limits were presented to introduce new opportunities
offered by the foundation of maps from multiple
perspectives: information management, data
visualisation, data filtering and recommendation.
It was addressed the issue of connecting
application data and geographical entities, discussing
a geographical indexing system able to provide to
applications the insight about what users “see”
interacting with a map. Based on the geographical
index, it was introduced the concept of focus layer a
coverage of geographical units relevant within a
given scale. Combining the connection between POIs
and geographical entities in focus layers it was
introduced the solution of area-based aggregation to
overcome the limitation of current clustering-based
approaches. Finally, it was presented the preliminary
result of a geographical recommender, which select
contents accordingly to users’ “focus area”.
Currently the presented ideas are being developed
in solutions to support civic platforms such as
FirstLife (http://firstlife.org/) and WeGovNow
(http://wegovnow.org/). The necessity to address
challenges issues such as aggregation, quality of
crowdsourced data, spatial filtering in a public
platform without relying on user profiling, was the
stimulus to pursue alternative solutions to main
stream approaches of social networks and thematic
web portals.
ACKNOWLEDGMENTS
A part of the research leading to these results has
received funding from the European Union’s Horizon
2020 research and innovation programme under grant
agreement n° 693514 ("WeGovNow"). The article
reflects only the authors' view and the European
Commission is not responsible for any use that may
be made of the information it contains.
REFERENCES
Shahaf, Dafna, et al. "Information cartography: creating
zoomable, large-scale maps of information."
Proceedings of the 19th ACM SIGKDD international
conference on Knowledge discovery and data mining.
ACM, 2013.
MacEachren, Alan M. How maps work: representation,
visualization, and design. Guilford Press, 1995.
Keates, John Stanley. Understanding maps. Routledge,
2014.
Okada, Alexandra, Simon Buckingham Shum, and Tony
Sherborne, eds. Knowledge Cartography: software
tools and mapping techniques. Springer, 2014.
Lam, Nina SiuNgan, and Dale A. Quattrochi. "On the
issues of scale, resolution, and fractal analysis in the
mapping sciences*." The Professional Geographer 44.1
(1992): 88-98.
Smith, Neil. "Contours of a spatialized politics: homeless
vehicles and the production of geographical scale."
Social text 33 (1992): 55-81.
Howitt, Richard. "Scale as relation: musical metaphors of
geographical scale." Area 30.1 (1998): 49-58.
Marston, Sallie A. "The social construction of scale."
Progress in human geography 24.2 (2000): 219-242.
Griesner, Jean-Benoît, Talel Abdessalem, and Hubert
Naacke. "POI Recommendation: Towards Fused
Matrix Factorization with Geographical and Temporal
Influences." Proceedings of the 9th ACM Conference
on Recommender Systems. ACM, 2015.
Liu, Bin, and Hui Xiong. "Point-of-Interest
Recommendation in Location Based Social Networks
with Topic and Location Awareness." SDM. Vol. 13.
2013.
Lu, Haokai, and James Caverlee. "Exploiting Geo-Spatial
Preference for Personalized Expert Recommendation."
Proceedings of the 9th ACM Conference on
Recommender Systems. ACM, 2015.
Oku, Kenta, Rika Kotera, and Kazutoshi Sumiya.
"Geographical recommender system based on
interaction between map operation and category
selection." Proceedings of the 1st International
Workshop on Information Heterogeneity and Fusion in
Recommender Systems. ACM, 2010.
Ye, Mao, Peifeng Yin, and Wang-Chien Lee. "Location
recommendation for location-based social networks."
Proceedings of the 18th SIGSPATIAL International
Conference on Advances in Geographic Information
Systems. ACM, 2010.
Guy, Ido. "The Role of User Location in Personalized
Search and Recommendation." Proceedings of the 9th
ACM Conference on Recommender Systems. ACM,
2015.
Antonini, A., et al. “Topology-aware indexing system for
Urban Knowledge”. IEEE Technically Sponsored
Computing Conference 2017.
Foundations of Map-based Web Applications - A Survey of the Use, Limits and Opportunities Offered by Digital Maps
99