Almost every human has a rich spatial
experience that is constantly developed while they live in a certain region.
This experience summarizes the human spatial orientation skills, human
perception and associations with particular places, and even human opinions
about these places, including personal feelings, assessment of the quality of
life, etc. Extracting and evaluating this experience is a promising yet
challenging task within Digital Humanities research. To solve this task,
different representations of so-called mental maps are used.
It is worth noting that the terms “representation
of a mental map” and “mental map” are not well-established yet, so other
authors may use different terms to denote these concepts, for example “mental
map” (to refer to what is here called a “mental map representation”) [1, 2], “cognitive
map” [3], “sketch map” [4]. As a rule, it depends on the specifics of the
scientific field and the preferences of the researchers.
For the sake of clarity and disambiguation,
let us define the terminology to use hereafter.
Mental map
– a
representation of a certain space in someone’s mind, possibly in conjunction
with some subjective representations of a non-spatial nature (if a person’s map
is being considered).
Mental map representation
– a reflection of a
mental map on some material storage (for example, paper or computer memory).
So, the
digital mental map representation (DMMR)
is a mental map
representation, created and stored in a computer or another digital device.
Generalized
mental map representation (or just generalized map)
– a data structure that
stores information about the objects in mental maps representations from a
certain set.
Generalized map visualization
– an interactive graphical
representation of the generalized map in the form of a two-dimensional chart.
In this paper, DMMR is used to compose a generalized map.
The methods of studying mental maps are
strongly affected by the development of digital technologies. It became clear
that the study of mental maps using digital technologies has a lot of
advantages [5–8]. Some digital systems for working with representations of
mental maps have appeared [6, 9, 10], including the Creative Maps Studio
application (https://creativemaps.studio),
which has been developed to allow creating digital
representations of mental maps. More than 900 maps from 12 regions of Russia
were collected by using this application. In order to analyze the general
perception of the country by the residents of each region, an analytical module
was developed within Creative Maps Studio. One of the operators of this analytical
module is a generalized map visualization operator (GMV). Generalized map
consists of a set of maps drawn by the residents of a certain region. GMV is
the subject of this paper.
In addition to describing the operation of
the GMV itself, its input and output data as well as some theoretical issues of
map data processing are discussed. Several interpretations of the output data
are proposed along with the discussion of pros and cons of using the GMV
rendering results. Assumptions are made about possible future works.
New analytical features are introduced to
the previously developed Creative Maps Studio vector graphics editor:
1.
The ability to automatically visualize a set of
mental maps representations (generalized map rendering).
2.
The ability to view the coordinates dispersion
of objects within the generalized map and compare the frequencies of occurrence
of any objects on maps (with a possibility to choose which objects will be
displayed).
Hypotheses about the existence of mental
maps are suggested not only in relation to people, but also in relation to
animals. Since there are no known cases of animals drawing representations of
mental maps, as a rule, their mental maps are studied indirectly. Only
assumptions can be made about how mental maps are arranged (if their existence
is taken for granted) based on how animals act in space. A pioneering work in
this field is [3] followed by [11, 12]. In these works, statistics are usually
collected on the actions of animals and the frequency of actions that more or
less contribute to the achievement of the estimated goal (usually, the
foraging).
As for people, here a wider range of
research tools is available. A person can be asked to draw something or fill
out a survey. In the present work, we concentrate on the map drawings as a
natural way to represent mental maps.
Informants can draw on ready-made maps,
marking features, which are important for them [7]. They can draw maps on their
own from scratch, both on paper [4, 13, 14] and on digital devices [8].
Accordingly, generalized analysis can also be carried out in traditional form
[13, 15] or in digital form [6, 8]. In the latter case, as a rule, the
researcher manually draws up a generalized map visualization. In other words, a
qualitative assessment takes place [13, 14]. Quantitative assessment is also
used, but it usually involves some expert manual steps [6].
At the same time, it is important to note
that mental map representations are sustained reflections of information from
the real world, albeit this information is modified by human perception [4,
16]. So, the study of representations of mental maps is considered relevant to
Digital Humanities research.
The mental map representations used in this
work as a data source are collected in digital format using the Creative Maps
Studio application developed earlier [10]. The analytical module (the GMV is a
part of) is built using the ontology engineering principles and data flow
programming paradigm. This approach was first introduced in the SciVi
ontology-driven visual analytics platform and proved its efficiency in solving
real-world visual analytics tasks from different application domains [17]. The
analytical pipeline is being described as a data flow diagram – a chain of
operators linked by data. Each operator has its certain typed inputs, outputs,
settings, and implementation described by a lightweight ontology using
predefined concepts like "Operator", "Input",
"Output", "Setting", "Worker", etc., and
paradigmal relations "is_a", "a_part_of", "has",
and "is_instance". This allows defining the analytical capabilities
in a knowledge base, and thereby easily extending the module’s palette of
operators without manual changes of its source code.
The idea of generalized visualization of
mental map representations was already discussed in [18]. In that work, maps
were collected in the traditional paper form, and the visualization of the
generalized map was built manually. To do this, the researcher looked at maps,
selected the most frequently drawn settlements, calculated the dispersion and
median values, and drew these data in the form of a diagram (an example of a
diagram from [18] is shown in Figure 1). However, with the advent of digital
technology, it became clear that the process of calculating the generalized map
can be automated. Moreover, more information can be extracted from digital
maps. Therefore, we decided to develop an analytical module in the Creative
Maps Studio, in which it would be possible, among other things, to visualize
generalized map.
Fig. 1. Generalized map visualization diagram from [18]
People’s ideas about space are very vague,
so the first issue that a researcher may face is to understand what tools to
use for the most complete and accurate reflection of the mental map. The usage
of DMMR makes it possible to create some ready-made generalized images of objects
that informants may want to draw. Appleyard [13] proposed a classification of
the types of mental map representations on paper. Looking at these types, it
can be figured out what kind of graphical primitives the informants may need.
Our study proposes three types of objects that an informant can place on a map:
a contour object, a point object, and a filled closed curve. In this case, the
digital graphics editor offers the informant a set of specific objects of each
of these three types, representing some generalized images of real objects.
When informants draw maps, they usually use
more or less primitive labels and colorings for their objects. This can be seen
by looking at examples of maps studied by various researchers [1, 2, 4, 8, 13,
14].
Accordingly, the interpretation of such
objects is usually carried out mainly due to the presence of explanatory labels
near the objects. The DMMR allows us to improve images of objects themselves to
make them more specific. Of course, it may turn out that the desired object
does not exist, then the informant can use something more abstract to represent
the object.
The specific set of objects depends on what
kind of space the informants draw. In our study, the informants draw a map of
Russia; accordingly, the set of objects represents the main objects, which can
be marked on national maps. The set of objects can be viewed here (the panel at
the top of the application page):
https://creativemaps.studio/.
Note that
for maps of a different scale, it may be necessary not only to use other
generalized images of objects, but also, possibly, fundamentally new types of
such images. For example, to depict roads or rivers on maps of urban space, it
may be convenient to use a closed polyline or curve (which can be filled with
specified color).
When informants draw maps, they may
represent the same object in different ways. Accordingly, the object will be
represented on different maps in different forms and with different names.
These differences can be divided into several groups.
The first group includes differences in the
graphical representation of the object. Informants can choose different default
objects from those available for the image of the same real object, or, if the
object is large, its shape can be very diverse for different informants. For
example, for the image of the city, the object “Town”, the object “City” or the
contour denoting the bounds of the city can be used (an example is in Figure
2). Therefore, when automatically processing the DMMR, it should be kept in
mind that diverse default objects can be used to display the same real object.
Fig. 2. An example of default objects set that
can be used to depict a city
The second group of differences is varying
names for the same object. The informant may not know the correct spelling of
the name, or may use an abbreviation or an informal but familiar name. For
example: the “Åêàòåðèíáóðã” (Yekaterinburg) city can be called “Åêàòèðèíáóðã”
(accidental spelling error) or “Åêá” (common unofficial abbreviation). In this
case, it may be useful to prepare a dictionary in advance, which will record
which real objects with which names can be found on maps. In this study, such a
dictionary is compiled manually. Automation of this dictionary composing is a
non-trivial task that requires advanced artificial intelligence. This may be a
subject of future works.
The third group of differences is the
possible ambiguity in the definition of an object by its name. This case can be
divided into two parts. Firstly, there are objects, which have the same name
(or an option of the name), while being in different places. For example, small
settlements, such as villages, often have typical names and can occur dozens of
times in different regions of the country. Secondly, objects can be different
in nature but have the same name. For example, there are a lot of cities, which
share their names with rivers. The first case is difficult to process
automatically, so our study uses manual processing for that. The second case
can be automatically handled by dividing all objects by type. That is, objects
of the “river” type cannot be checked when searching for a specific city, and
vice versa.
The fourth group of differences is
represented by unknown objects. These are objects, which were not named by the
informant. And looking at those objects, it is impossible to unequivocally
conclude whether they are reflections of any real objects or not. For example,
a city is marked on the map, but it is not described in any way. Such an object
can be interpreted in different ways. Maybe this is a specific city that the
informant noted but did not name. But also, it may be that the informant meant
that there is probably a city in this place, but he did not know which one.
Based on this, it is obvious that fully
automatic processing of DMMR is not yet possible. In some cases, expert
participation in map processing is required. However, the achieved level of automatization
still removes a lot of routine work from the researcher.
Each DMMR contains a set of objects as
imagined by the informant. Each generalized object of a generalized map can be
assembled from the corresponding objects of each of the DMMR in different ways,
as described below.
Firstly, it is possible to assemble a
generalized object using the most frequent parameter values that are found on
specific DMMRs. This will give us an approximation of the “typical” map drawn
by the “average” informant. On the one hand, this is convenient from the point
of view of identifying a typical DMMR. But on the other hand, this approach
does not consider the less frequent values of the object parameters, so the
information on the generalized map may be incomplete.
Secondly, generalized objects can be
assembled using the weighted averages of the parameters. This approach is
better than the first one because it considers all parameter values, but at the
same time it gives a slightly different result. The generalized map will show
the average values and they may not look like the most popular ones. This
approach, like the previous one, reflects well the values for those parameters,
which are represented by single numbers. For example, the size of a point
object. However, this approach works worse for those parameters, which are
represented by vectors, for example, for coordinates or color (decomposed into
channels). This is due to the fact that in weighted averaging, the parameters
are considered separately. And the resulting joined value may not actually
appear on the source maps at all.
A value that is not found on the DMMRs can
also be obtained for a parameter represented by a single number. This situation
can arise if the distribution of the parameter value has more than one local
frequency maximum. However, the occurrence of several distribution peaks in
different places rather indicates that the informants are incorrectly grouped
and there is no meaningful generalized map for them. Perhaps the group is too
large and it should be further divided according to some attribute. Therefore,
this shortcoming was rated as having little effect.
For some parameters, the best way to
generalize them is to display all possible values simultaneously (in a form of
a grid or a diagram), but this can negatively affect the cognitive clarity of
the generalized map cluttering up its visual representation.
The visual representation of the
generalized DMMR is an issue that deserves attention. In many ways, this issue
is related to the previous one, since the main difficulty of visualization is
that, on the one hand, it is necessary to display as much information as possible,
and on the other hand, the clarity of image can sink due to excessive
information. Therefore, it is necessary to find the best option for each
visualization element.
It makes sense to visualize the parameters
of the DMMR objects in different ways, since they represent fundamentally
different properties of the objects.
If two-dimensional space is used to create
a visualization, then for coordinates, it is convenient to visualize not only a
generalized value, but also the entire range.
If we consider the color of objects, then
it is advisable to schematically represent the object itself and its label,
coloring them in a generalized color. Regarding the object type, different
informants can choose different pictures of the same real geographical object, so
it is relevant to choose something neutral to depict a generalized object (for
example, use just a circle to represent a settlement, and not an icon, as
Creative Maps Studio allows using different icons for settlements, see Figure
2). As for the color, it can be generalized by averaging the RGB components. It
may be not the best solution, as RGB space is non-linear by its nature, but
accruing to our experiments the averaging gives satisfactory results. Improving
the color generalization may be a matter of further studies.
Differences in size are also conveniently
reflected in the sizes of generic objects. However, the sketchiness of the
generic object also affects the size of the generalized object. Since, for
example, it is difficult to draw a generalized curve, we can simply bound the
area, where the curves are usually found, with a rectangle.
Informants who participated in this study
rarely used the transparency parameter for objects, so transparency for
generalized objects can be used to visualize another parameter. It is quite
intuitive to reflect the importance of the object by its transparency.
Therefore, it is convenient to use transparency to reflect the frequency of the
presence of an object on the DMMR (the more often an object is present, the less
transparent it is).
Informants draw objects at different times.
One possible way to show sequence on a generalized map is to place objects on
different planes (using so-called z-order). This is a logical solution, but it
has a drawback: the z-order of arbitrary objects is difficult to compare if
they do not intersect on a two-dimensional projection. Therefore, the sequence
of drawing objects is better reflected by numbers indicating the minute from
the beginning of the map drawing, in which the object was drawn on average
(assuming the drawing process is sampled at the minute scale).
Summarizing the assumptions above, we can
conclude that the various parameters of objects have their own peculiarities
and the right approach should be found to visualize each of them.
Since the DMMR is based on vector graphics,
its internal format is textual. Listing 1 shows a format for the DMMR. It has
two parts: the final state and the history of actions. The action history is
used to load the map into the application and to analyze the map drawing
sequence. The final state is cached for analytical purposes to avoid excessive
calculations related to traversing the history and dynamically inferring the
final state. An example of a loaded map is shown in Figure 3.
Listing 1. Suggested format of the DMMR
{
"figures":{
"c0405a2b-584b-4af8-b0f4-fb12fee2b2da":{"x": 234,
"y": 83, …, "path": "M 61,76 Q 37,78 237,0 Q 300.5,61
214,178 Q 0,206 50,190 Q 43.5,186 39,190 L 61,76", …},
…
},
"actionHistory":[
{"type":"addPolygon", "x":131, "y":93,
"points":"M 61,76 Q 37,78 237,0 Q 300.5,61 214,178 Q 0,206
50,190 Q 43.5,186 39,190 L 61,76", "time":11635, …},
{"uuid":"c0405a2b-584b-4af8-b0f4-fb12fee2b2da",
"type":"moveFigure", "time":27660,
"x":234, "y":83},
…
],
"mapName":"Íèæíÿÿ
êàìåíêà",
…
}
Fig. 3. An example of DMMR loaded in Creative Maps Studio
To build a generalized map, an analytical
module is used, described in the next section. A generalized map format is
shown in Listing 2. In fact, the generalized map contains a list of corresponding
map objects split by category. Each object stores the frequency of its presence
on the maps and a list of corresponding parameters. Each parameter contains a
frequency distribution of its values.
Listing 2. Suggested format of a
generalized map
{
"Mountains": {
"Ural mountains": {
"frequency": 0.44,
"params_distribution": {
"colorB": {
"0": 1,
"40": 2,
"60": 1,
…
},
…
},
…
},
},
…
}
At the moment, an analytical module is
being implemented in Creative Maps Studio, based on the visual data flow
programming concept. Generally speaking, this analytical module contains not
only the visualization algorithm presented below, but also other algorithms,
however they are not considered here laying beyond the scope of the present
work.
The point of creating a visual programming
system directly in the Creative Maps Studio is that it allows you not to send
large input data over the network to other software (for example, to the SciVi
platform [19]), and also makes it possible to use algorithms and software
operators, which are implemented using the React framework. React is a frontend
framework that uses so-called components, which, unlike regular JavaScript,
allow easy incorporating of HTML elements and provide state management routines
to simplify the building of graphical user interfaces.
Figure 4 shows a data flow diagram that
describes a step-by-step transformation of a set of source maps into a
generalized map and a visualization of this generalized map. Each operator
performs some data transformation algorithm. The inputs and outputs indicate
the types of data, which must be submitted to the input, and which will be
returned as a result of the algorithm. The “Data Source” operator has no inputs
and is designed to download data from a server or local computer. As can be
seen in the diagram, the processing is as follows. First, the list of maps is
loaded (the data format
List[m]
denotes a list of maps) parallel to the loading of the objects’
names dictionary (the data format
mpng-usn
denotes a mapping of the users’
names of objects to the official ones). After that, the data is loaded and
processed by the “Map Generalizer” operator. The operator removes all
unnecessary data, combines all the necessary information into one data
structure and returns it in the
gm
format (Generalized Map, see Listing 2).
After that, data are passed to the "Generalized Map Visualizer" that
renders a visualization result.
Fig. 4. Data flow diagram for generalized map visualization
Fig. 5. Scheme of the proposed interaction, based on the API of the
Creative Maps Studio and SciVi
As future work, we plan to set up the
interaction of the SciVi and Creative Maps Studio systems in terms of data
exchange via API (remote procedure call). This will allow using the data
processing and visualization algorithms of both systems together without
duplicating the code (see the diagram in Figure 5). The generalized map
visualization algorithm is implemented as one of the software operators inside
the analytical module and is executed on the Creative Maps Studio client.
An example of a generalized map is shown in
Figure 6. The generalized map displays the average values of various parameters
of objects that were found on the maps this generalized map was built upon. The
circles represent the settlements. The little squares represent the mountains
and mountain-like objects (such as volcanoes). The big rectangular areas
represent the seas, lakes, islands, and other big objects. The size of each
figure reflects the weighted average size of the corresponding object on the
maps. The color of each figure (except for the alpha channel) reflects the
weighted average color of the corresponding object. The color and the size of
the label near each figure reflects the weighted average of the corresponding
properties of the corresponding map labels. The number after the label in
parenthesis shows the time in which the corresponding object is drawn on the
map. That is, for example, in the illustration below, the informants draw the
settlement “Ìóðìàíñê” (Murmansk) on average earlier than the sea “Ìîðå Ëàïòåâûõ”
(Laptev Sea), but later than the peninsula “Êàì÷àòêà” (Kamchatka). The alpha
channel reflects the frequency of occurrence of the corresponding object on the
maps (in addition, the quantitative value of the frequency is indicated as a
percentage placed in or near the corresponding figure). The greater the transparency,
the rarer the object occurs, and vice versa (at a minimum frequency,
transparency takes on a value of 20%). The frequency of the opaquest object at
the moment is displayed in the upper right corner (in Figure 6 it is 48%).
It is also possible to display only one
object category. Example is shown in Figure 7 (settlements only are shown).
There is a possibility to display coordinate spreads for figures. However, it
is more convenient to do this when there are few figures on the generalized map
visualization. Figure 8 shows an example of manual filtering of objects on the
map with display of coordinate spreads (the “Show dispersion of coordinates”
option is set). It can be noticed that the maximum frequency among the
displayed objects is 52%, while the most frequent object displayed is
“ßêóòñê”
(Yakutsk). And it can be seen how precisely the informants
represent the positions of the objects (the most precise result here is for the
settlement
“Êàëèíèíãðàä”
(Kaliningrad), because it has
the smallest spread).
Fig. 6. Example of a generalized map visualization
Fig. 7. Example of particular visualization of a generalized map
(settlements only are shown)
Fig. 8. Example of particular visualization of a generalized map
(manual filtering is applied and “Show coordinate dispersion” option is
switched on)
The visualization of the generalized map
makes it possible to obtain some aggregated information about the informants’
collective ideas about the object of study. In particular, firstly, looking at
the map, we can find out which objects are in the center of attention of the
informants and how they approximately appear (although the visualization of the
generalized map provides much more scarce information regarding the image of
the object (color, size, etc.) than if the text layer of the maps were
analyzed). Secondly, we can compare the frequency of drawing an arbitrary group
of objects on the maps (marking them in the column on the right of the
graphical user interface shown in Figures 6–8). Thirdly, it is possible to
obtain information about the positioning of objects: how uniform is the opinion
of informants about the location of a certain object. Fourthly, information
about the approximate order in which objects are drawn on maps can be obtained.
Also, in some cases, it is possible to
determine the region of residence of the majority of informants from the
sample. This is possible due to the fact that informants more often mark the
capital and other objects from their own region, than from the foreign regions.
The exceptions are the cities of
“Ìîñêâà”
(Moscow) and
“Ñàíêò
-
Ïåòåðáóðã”
(St. Petersburg) and some other
popular objects, such as lake
“Áàéêàë”
(Baikal), which
are often marked on the map, regardless of the region of residence of the
informants, since they are very famous in Russia. However, according to our
experience, this exception does not significantly affect the definition of the
informant’ region of residence.
In conclusion, it can be said that
automating the processing of map data (in this case, generalized map rendering)
has some advantages. First, graphical display of a generalized map can help to
generate or test hypotheses faster than having only a set of maps, which can be
viewed separately from each other, or some textual data, which describe this
generalized map. Secondly, it allows us not to waste time on some routine
operations with maps. For example, some quantitative calculations and
comparisons of maps can be done automatically, since maps have a well-defined
digital format.
Of course, the proposed approach also has
its drawbacks. First, expert input is needed in the early stages of preparing
maps. The researcher needs to determine how to recognize whether an object is a
populated area or not. To do this, it is necessary to compile a dictionary of
objects’ names, including in it, in addition to the official names of objects,
also informal names, which informants may use to designate these objects. Of
course, the more maps are processed, the less such participation is required in
the future (new maps will contain less and less new objects’ names), since for
any object the set of its possible names is quite limited. Secondly, as
mentioned earlier, there is an ambiguity problem for many rivers and small
settlements, such as villages, and a few large ones, for example, the cities of
“Ðîñòîâ” (Rostov) and “Ðîñòîâ-íà-Äîíó” (Rostov-on-Don) (both can be called “Ðîñòîâ”),
“Âåëèêèé Íîâãîðîä” (Veliky Novgorod) and “Íèæíèé Íîâãîðîä” (Nizhny Novgorod)
(both can be called “Íîâãîðîä”), etc. Nevertheless, the visualization of the
generalized map seems to be appropriate and quite informative.
There are several directions for further
research development. First, we plan to compare the visualized map with the
real geographic objects (in different projections) to get information about the
accuracy of the positioning of the objects relative to their real position. It
is likely that for many objects there will be a regular shift. Saarinen
mentioned this phenomenon in [14]. Moreover, there are quite a lot of works in
the field of comparing representations of mental maps with GIS (for example,
Curtis et al. composed a review of similar works in [5]).
Secondly, it is possible to improve the
color representation of objects. The weighted average of the color can be
replaced with a pie chart because it can give more relevant information about
the colors used by informants.
Perhaps generalized map visualization can
be useful not only for the analysis of mental map representations. Despite the
fact that it is not yet known whether there are data of a similar format in
other scientific fields, it is considered appropriate to organize API
interaction with the SciVi visualization system so that, if necessary, it would
be possible to reuse the created visualization operator in other application
domains.
This work was supported by the Russian
Science Foundation (grant number 20-18-00336).
1.
Gieseking, J.J. Where We Go From Here: The Mental Sketch Mapping
Method and Its Analytic Components // Qualitative Inquiry – 2013. – Vol. 19. –
PP. 712–724.
DOI: 10.1177/1077800413500926.
2.
Castellar, S.M.V., Juliasz, P.C.S. Mental Map
and Spatial Thinking // Proceedings of the ICA – 2018. – Vol. 1. – PP. 1–6.
DOI: 10.5194/ica-proc-1-18-2018.
3.
Tolman, E.C. Cognitive Maps in Rats and Men //
Psychological Review – 1948. – Vol. 55. – PP. 189–208. DOI: 10.1037/h0061626.
4.
Blades, M. The Reliability of Data Collected
from Sketch Maps // Journal of Environmental Psychology – 1990. – Vol. 10. –
PP. 327–339. DOI: 10.1016/S0272-4944(05)80032-5.
5.
Curtis, J.W., Shiau, E., Lowery, B., Sloane, D.,
Hennigan, K., Curtis, A. The Prospects and Problems of Integrating Sketch Maps
with Geographic Information Systems to Understand Environmental Perception: a Case
Study of Mapping Youth Fear in Los Angeles Gang Neighborhoods // Environment
and Planning B: Planning and Design – 2014. – Vol. 41. – PP. 251–271. DOI:
10.1068/b38151.
6.
Aram, F., Solgi, E., García, E.H., Mohammadzadeh S., D.,
Mosavi, A., Shamshirband, S. Design and Validation of a Computational Program
for Analysing Mental Maps: Aram Mental Map Analyzer // Sustainability – 2019. –
Vol. 11.
DOI: 10.3390/su11143790.
7.
Pánek, J. Emotional Maps: Participatory Crowdsourcing
of Citizens’ Perceptions of Their Urban Environment // Cartographic
Perspectives – 2018. DOI: 10.14714/CP91.1419.
8.
Huynh, N.T., Doherty, S.T. Digital Sketch-Map Drawing as an
Instrument to Collect Data about Spatial Cognition // Cartographica – 2007. –
Vol. 42. – PP. 285–296.
DOI: 10.3138/carto.42.4.285.
9.
Gardony, A.L., Taylor, H.A., Brunyé, T.T.
Gardony Map Drawing Analyzer: Software for Quantitative Analysis of Sketch Maps
// Behav Res – 2015. – Vol. 48. – PP. 151–177. DOI: 10.3758/s13428-014-0556-x.
10.
Chumakov, R.V., Ryabinin, K.V., Belousov, K.I., Duan, J. Creative Map
Studio: a Platform for Visual Analytics of Mental Maps // Scientific
Visualization – 2021. – Vol. 13. – PP. 79–93. DOI: 10.26583/sv.13.2.06.
11.
Menzel,
E.W. Chimpanzee Spatial Memory Organization // Science – 1973. – Vol. 182. –
PP. 943–945.
DOI: 10.1126/science.182.4115.943.
12.
Boesch, C.,
Boesch, H. Mental Map in Wild Chimpanzees: An Analysis of Hammer Transports for
Nut Cracking // Primates – 1984. – Vol. 25. – PP. 160–170.
DOI:
10.1007/BF02382388.
13.
Appleyard,
D. Styles and Methods of Structuring a City // Environment and Behavior – 1970.
– Vol. 2. – PP. 100–117.
DOI: 10.1177/001391657000200106.
14.
Saarinen, T.F. Centering of Mental Maps of the World // National
Geographic Research – 1988. – Vol. 4. – PP. 112–127.
15.
Kitchin, R.M., Fotheringham, A.S. Aggregation Issues in Cognitive
Mapping // Professional Geographer – 1997. – Vol. 49. – PP. 269–280. DOI:
10.1111/0033-0124.00076.
16.
Wang, J., Schwering, A. Invariant Spatial Information in Sketch Maps
– a Study of Survey Sketch Maps of Urban Areas // Journal of Spatial
Information Science – 2015. – Vol. 2015. – PP. 31–52. DOI:
10.5311/JOSIS.2015.11.225.
17.
Chuprina,
S.I., Ryabinin, K.V., Koznov, D.V., Matkin, K.A. Ontology-Driven Visual
Analytics Software Development // Programming and Computer Software – 2022. –
Vol. 48. – PP. 208–214.
DOI: 10.1134/S0361768822030033.
18.
Zelianskaia, N.L., Baranov, D.A., Belousov, K.I. Naive Geography and
Topology of Geomental Maps // Socio- and Psycholinguistic Research – 2016. –
Vol. 4. – PP. 126–136.
19.
Ryabinin, K., Chuprina, S. Development of Ontology-Based
Multiplatform Adaptive Scientific Visualization System // Journal of
Computational Science. – 2015. – Vol. 10. – PP. 370–381. DOI:
10.1016/j.jocs.2015.03.003.
