ISSN (Print) : 0974-6846

ISSN (Online) : 0974-5645

Indian Journal of Science and Technology, Vol 12(43), DOI: 10.17485/ijst/2019/v12i43/146432, November 2019

Core Architecture and Design Issues of Collaborative

Tagging Applications
Fouzia Jabeen* and Shah Khusro
University of Peshawar, Peshawar, Pakistan; [Link]@[Link], khusro@[Link]

Abstract
Objectives: Collaborative social tagging applications like Flickr, [Link], Pinterest, and Technorati have gained
popularity among people in a very short period. The primary reason is the sharing and provision of metadata in the form
of tags taking advantage of users’ own vocabulary. There are two major benefits of user assigned tags, firstly, it enhances
the understandability of the resources and secondly, it is used for search and retrieval. The study focuses on the analysis
of aspects such as design decisions, candidate solutions, their implications on different important aspects of collaborative
system, and available relevant technologies. Method/statistical analysis: This study gives review of the architectural
and design aspects of collaborative tagging systems. Findings: The right choice of design decisions and technologies will
greatly influence the performance of the systems and will be a step towards a well-built folksonomy system. Application/
improvement: Our study will be beneficial for researchers, designers, and developers of collaborative tagging applications.

Keywords: Collaborative Tagging Application, Collaborative Application Design, Architecture, Folksonomy, Social
Networks

1. Introduction no parent–child (hierarchical) relationships. Secondly,

folksonomies are completely uncontrolled as opposed to
Collaborative tagging applications allow its users to taxonomy where an essential part of creating it is deciding
annotate resources(s) with keywords, often called tags. what names are to be used for identifications of entities.
The resource can be a photograph, URL, computer game, In a folksonomy, there can be a variety of different words
music or video, etc. Collaborative tagging is known by for the same concept. Any relationships (co-occurrence,
several names which are social tagging, social indexing, subsumption, etc.) you see in a folksonomy can be derived
social classification, and folksonomy. Broad folksonomy mathematically (statistical clustering), utilizing external
and narrow folksonomy are alternative names used for knowledge sources (Word Net, DBpedia, Wikipedia,
collaborative and simple tagging, respectively.1 Yago). In Taxonomy, tags are added by the content creator
Comparative analysis of meaning of some of the related or by the author. On the other hand, folksonomy tags are
terms is necessary before going into depth of the design added by anyone (content creator, annotator, consumer,
issues of collaborative tagging applications. Folksonomy reader, etc.).
and Personomy-folksonomy is the term used to refer to For example, Flickr keywords (where the user add
users’ vocabulary. Personomy refers to the user’s own their own tags/keywords) for describing a photo.4
vocabulary.2 The essence of folksonomies is the combined Folksonomy, Ontology and Folktology-folksonomies and
wisdom of web users in the collection.3 Folksonomy ontologies can be placed at the two opposite ends of a
and Taxonomy-taxonomy is hierarchy of pre-defined categorisation spectrum5 in the sense that folksonomy is
categories which users can use to classify resources. relaxed and ontology is very strict. Ontology, in the field
Folksonomies, however, is opposite to taxonomy in many of information science, is basically a study and definition
aspects. Firstly, folksonomies are flat (that is, there are of the categories of things that exist or may exist in some

*Author for correspondence

 Core Architecture and Design Issues of Collaborative Tagging Applications

domain. Combining folksonomy with ontology becomes study of the available technologies. Lastly, in Section 5 we
folktology.6,7 have tabulated the summary and presented conclusions
The main contribution of this study is; it gives comparison in Section 6.
of design issues, their possible solutions, impact of these
decisions on important aspect of collaborative application
and relevant available technologies. We focused on two
2. Collaborative Tagging
core collaborative application architecture components Environment
which include the following: a) Tags, Resource, and User
b) Browsing and Searching. The reason for choosing these
2.1. Tagging Types
two aspects are first, every collaborative application will Tagging approaches can be broadly classified as simple or
have to store these three core components (Tag, Resource, collaborative in nature. Photo Gallery and Flickr when
and Tagger/user) and secondly, these tags are then used launched utilised simple tagging system. They allowed
for searching and browsing resources. Precise search annotation to the original resource only. Combined view
and retrieval result is one of the principal objectives of a of the tags for a particular resource was not available.
successful quality collaborative application.8,9 Figure 2(a) is pictorial representation of simple tagging in
The significance of the study is: it helps in selection which user annotates a resource using different tags (tagi),
of appropriate candidate solution(s) and available where i indicates the tag number.12 However, in [Link].
technologies that will have significant impact on the us and Amazon user tags pointer rather than a resource.
gathered users’ vocabulary, quality of folksonomy, In collaborative tagging, each end-user has their own
indexing depth and precision in retrieved results. In collection of tags for a resource and tags from all the users
comparison with the work of10,11 which only describes are grouped to give a combined view. That is the reason for
software components, this study covers the possible calling this model a collaborating tagging model.13 Figure
solutions for the design issues and their implications. 2(b) shows this model.12 Every resource is annotated with
The study is organised according to the framework tag tagij. Tagger (the annotating user) is denoted by j and
as shown in Figure 1. The study sections are planned i is the number of a tag. As more than one user annotates
as follows: Sections 2 and 3 outline the design issues, the same resource, a weighted collection of tags is there
comparative analysis and impact of the available solutions for each resource represented with wtagi. The alternative
related to tags, resources and user’s management, as well names for collaborative tagging and simple tagging are
as browsing and searching components of collaborative broad folksonomy and narrow folksonomy, respectively.1
tagging application. Section 4 highlights the comparative
2.2. Collaborative Tagging Architecture and
Design Styles
The architecture includes analysis of all the abstract
design choices that require attention when developing a
tagging application.

(a) (b)

Figure 1. Framework Figure 2. (a) Simple tagging. (b) Collaborative tagging.12

2 Vol 12 (43) | November 2019 | [Link] Indian Journal of Science and Technology
 Fouzia Jabeen and Shah Khusro

2.2.1. Design Issues Related to Triple and Their users watch and copy the active users’ style of tagging
Impact rendering active users act as role model.
A tagging system pyramid (also called triplet) is composed
of three parts: users, tags, and resources. In this section, [Link]. Resource Part Design Decisions
we have outlined the key decisions along with the possible For the second pillar of collaborative application that is
solutions and their impact on different aspects of social resource, we need to take decisions about contributions,
applications. original/pointer, privacy, and restrictions.
• Contribution-contribution to the system can either
[Link]. User Part Design Decisions be user-contributed or system-contributed resources.
Resources added by the users on regular basis are
The first part, user, the issues are membership and
termed as user contributed resources while system
community.
contributed resources are comprised of pre-existing
• Membership-is concerned with the skill to add databases or artifacts. Images uploaded on Flickr are
resources, to build tags, and to observe/view other an opt example for user contributed resources. On the
folks tags and resources. In Public sign-up people other hand, Amazon product catalog serves as system
become part of the system by simply signing up them- contributed resources and thus cannot be mutated by
selves. Invitation only option allows the existing mem- the end users.
bers to invite other members. External criteria/private Impact-these models are not exclusive in usage, for
membership criteria are based on subjective factors. example, Library Things combines both models (user
For example, if an organisation is using internal social + system contributed resources). If resources are being
bookmarking application, the staff members can be contributed by users, and there is a continual stream of
members. additions. How fast that stream moves totally depends
• Impact-if membership is not determined explicitly on activity of users.
then we end up in same category as Yahoo Podcasts
If resources are already part of the system or if they are
where there is no obvious relationship between tags
being added by another process, then the rate and size of
and users and one cannot even find their own submit-
change in the collection of tags and resources might not
ted tags easily.14 Through membership we can keep in
correlate with the users’ activity.
touch with turnover (The rate at which users join and
For user-contributed resources, submission of
leave), and activity (how much a user posts resources
resources and assigning tags to them is done at the same
and tags). User activity determines the dynamism and
time. In contrast, for resources already present in the
volume of the resources and tags.
system, users may not feel strong motivation to tag them.
• Community-means the way users interact. The con-
Depending on the dynamism of resource collection,
nection can be in the form of followers, contacts, or
navigation and visualisation tools are designed to be
groups. Followers, is a one-way linkage between a sub-
able to show relevant trends. In conclusion, the speed of
scriber and content producer. A user can follow any
resource collection stream influences navigation design
other user(s). [Link] is one of the most prominent
decisions.
applications supporting this concept. Contacts, a user
requests to other user to add him/her to their contact • Original/pointer-is the second design decision about
list, upon acceptance a two-way linkage occurs. Face- what is being tagged. That is, either the original
book is an example that works on this model. Groups, resource or the pointer. Original resource can be an
means one or more users make a group to share same actual video or a picture, etc. In case of multi users
set of resources related to a topic with group mem- involved in a tagging activity with respect to a resource,
bers. Permission may be required from administrator all the tags are assigned to the resource itself. In case
in order to join the group. of pointer, the tags are not attached to actual resource
• Impact-these connections act as building blocks for but to the record in the database that exists for that
social navigation: discovering information by follow- particular resource. In other words, the record for that
ing the streams of other users. In addition, passive particular resource is tagged. One resource can have

Vol 12 (43) | November 2019 | [Link] Indian Journal of Science and Technology 3
 Core Architecture and Design Issues of Collaborative Tagging Applications

multiple pointers (as many as the people annotating just a subset of resource collection or entire collection
the resource). is at its disposal. These choices are specific to tagging
Impact-the systems that allow tagging original application.
resource lack collaboration, although it is possible to Impact-it is possible to gather a lot more metadata
aggregate tags given by different users. Technically, if users are allowed to tag resources without logging
there are very few cases where one can tag the actual in, signing up, or performing some other action that
resource. confirms their identity. However, when users’ ability
• Privacy-refers to the ability to view or modify a to find resources is inhibited due to inefficacy to
resource. Privacy can be broadly categorised into four distinguish one user tags from another, users quickly
types. Open to public entails each link or resource is lose motivation to tag.
accessible to all web users, for example, [Link]. • Tag sharing-means tag added by one user is available
Configurable but public by default allow to set the pri- to other users. Precise choice can either be public, pri-
vacy settings selective for visibility of comments and vate or community/group.17
tags, e.g., Flickr. By making the default setting public, Impact-conformity theory observed in folksonomies
people are motivated to share the resources and tags. indicates that the tags that users observe from other
Configurable but private by default make default set- users will influence their own choice of tags. The proof
ting to private and able to configure according to the of this is that once a tag becomes popular it remains
needs. Lastly, in everything is private all resources and popular.18 So, a worthy tag selection by the users will
tags are private with no option to share. have significant impact on other users’ tag selection.
Impact-the sociability of tagging system depends • Selection of tags-to handle choice of the tags, we need
greatly on default privacy settings. to find out whether a user participates somehow in
• Restriction means what is not allowed. A system may the selection of tags, or should the system handle the
put restriction on resources by their types. In case of selection automatically. If the system gives opportu-
File Type as Flickr only allows image files. In case of nity of tag sharing at more than one level, then tag
Object people a notate books of library with tags for selection could be based on these levels. The two pos-
example. sible approaches for this can be either user is given a
Impact-these restrictions are usually obvious to the choice that from which level tags needs to be selected
users, for example, it is not possible that a user joins or system separates tags for the different levels, like
a photo-sharing site and starts tagging/sharing web public and private set of tags. In case of broad tagging,
bookmarks. the tags’ popularity (number of times a particular tag
is chosen) can be used to select a subset representing
the most popular tags.17
[Link]. Tag Part Design Decisions • Tag scope-scope of a tag refers to whether a tag belongs
Tags are keywords attached to the resource by the users. to an individual user or is shared by a community.
In15 classified tags as traditional tags (e.g., #tag) and There are two possible choices that are broad and nar-
non-traditional tags (geo tags, system tags). Heckner16 row tagging as discussed in Section 2.1.17
further classified the traditional tags into three categories Impact-broad tagging is collaborative in contrast to
linguistic, functional, and tag-to-text. Irrespective of narrow tagging. In narrow tagging there is restrictive
tag type, decision choices fall in permission, truth and collaboration.
control categories. • Tag support-this issue regards whether the user is pro-
vided with the help or suggestions when adding tag(s).
• Permission-deals with the issue that who can edit,
Possible choices are blind tagging, viewable tagging
remove, or add a tag. Mostly, the contributor of the
and suggestive tagging. In blind tagging, a user is not
resource(s) is allowed to add, edit, or remove tags.
able to see tags entered by other users while tagging.
However, the permission is extended to friends as well
With viewable tagging, the user can view the tags
(Flickr is an example). In case of system resources, the
added to an item by other users while tagging. The last
first permissions-related issue is whether to allow tag-
category is suggestive tagging, where the system rec-
ging or not. Secondly, should user be allowed to tag
ommends appropriate tags to the user.

4 Vol 12 (43) | November 2019 | [Link] Indian Journal of Science and Technology
 Fouzia Jabeen and Shah Khusro

Impact-thinking about tags to be associated with Access, or open source MYSQL. Another alternate choice
a resource is time consuming.19 Suggestive tagging can be a triple store.
reduces cognitive load on the users, as the users need
to choose one or more recommended tags instead of 2.3.1. [Link] Schemas
thinking and summarizing a source to find out the MySQLicious, Scuttle, and Toxis chem as are popularised
best descriptive tags. and used by [Link].20 MySQLicious is a denormalised
• Tag format and control decisions-format means the solution. In this schema, a single table is used as shown in
allowed character set for tags, and whether a tag will Figure 3.20 The most prominent influence of this schema
contain one or more words or even one or more sen- is its simple design. Usage and querying is based only on
tences. It is not possible to list all possible tag for- one table. The queries are very simple and straightforward.
mats sowing to large number of possibilities based For a very small database, this schema might be noticeable
on varying changes in character sets. However, a because of its simplicity. Although simplicity is a positive
better approach can be to give some general catego- side of this design, it suffers from duplication of data
ries, focusing on the degree of restriction. An open- due to the denormalised nature of the schema. The
minded approach is to let users make tags in any way schema is not flexible. Certain queries for obtaining the
they want, placing no constraints on the number of relationships that exist between users would be hard to
words or the type of characters permitted. This can code and slow to carry out. With a single table describing
lead to tags resembling full sentences. In contrast to every entity, a database developer would require complex
this approach, tags can be limited to single words and and time-consuming queries for retrieval of information
only alphabetical characters.17 on specific entities.
Impact-degree of restriction has direct impact on Scuttle schema shown in Figure 4 provides solution
collected tags’ quality. in two tables.20 The sc Categories stores tags and contains
• Control decision-is all about restrictions on using cer- a foreign key to the second table of sc Bookmarks. The
tain words as tags, for example, Amazon do not per- prominent pros of this technique are; it is more normalised
mit common expletives to be used as tags. as compared to MySQLicious. User can have many tags
Impact-it is very critical not to exercise too much for each bookmark. There is no restriction on repetition
strictness and control. Users might use tags that of tag textual data in the scbookmark table. To show the
express an opinion that may not be liked but is not relationship between bookmark and the bookmark’s set
otherwise objectionable. If users feel that they cannot of tags, foreign key can be implemented. If we look at the
express or are being unreasonably censored, they may bookmarks table, only single tag is stored per row. This is
leave the system. the reason the Scuttle is not, however, precisely normalised
• Truth-is the term used for storage place of tags. Truth and has limitations. A surrogate key is included in the sc
in the file refers to tags deposition in a file (resource). Categories table (id) in place of primary key tuple on bid
The other substitute is truth in the database where and category. Memory utilisation is increased because in
tags are kept separate from the resource, mostly in the tags table, Id of the tag text (Category) is not used and
database. textual tag data is used. As a result, a small number of
Impact-truth has some practical implications. If index records are being able to accommodate into a single
system is a social bookmarking and it is implemented
on intranet, the tags will be included in database
rather than the documents themselves. For the tags
used to bring precision in search results, it is required
that search engine indexes the tags in the database as
well as with the documents.

2.3. Data Models/Candidate Solutions

To store tags, users, resources, and other related data, we
can use relational databases like SQL Server, Microsoft Figure 3. MySQLicious.20

Vol 12 (43) | November 2019 | [Link] Indian Journal of Science and Technology 5
 Core Architecture and Design Issues of Collaborative Tagging Applications

index block. Which will reduce the efficiency and speed Simple tagging model as represented in Figure 6,14
of index accesses. make use of three tables (user, tags, and resources) to keep
Existing schema does not seem to be successful in users, tags, and resources. One additional table called
accurately retrieving statistical information like total resources tags is included for normalisation purpose. In
number of bookmarks for each tag. The reason is one-to- this model, each resource is unique to each user and has
many relationship is unable to represent or mis-represent multiple tags.
the true many-to-many relationship that is present Extraction of data and querying simple tagging model
between the bookmarks and tags. is easy. For example, the query shown in Figure 7 is used
The three table solution called Toxi shown in Figure for finding tags assigned to a particular resource.14 Figure
5 came up to overcome the deficiencies of above two 8 represents a query in which usage count of a tag is
schemas.20 The tagmaptable manages n-to-m relation determined.14 This metric can then be utilised as input to
between the tags and bookmarks. Word Press utilises font size, representing the frequency of occurrence in tag
this schema. The good points of the schema are first, this cloud. To conclude, we can say that this simple model can
is the most normalised solution (that is, you can go for be effectively used for personal information management
3NF). Secondly, doing so, removes the issue of duplicated applications.
information (particularly the textual tag data) that is Collaborative tagging model is represented in Figure
present in the other two designs. However, the schema is 9. There is a slight change. The table resources_tags is
14

not free from disadvantages. The cons are, when changing removed from and replaced with intermediary table
or deleting bookmarks one will get tag-orphans. The called users_resources_tags. This table is required because
other main drawback of Toxi schema (as is the case with in this model more than one user are allowed to assign
most properly normalised schema) is that there are now tag(s) to a particular resource. This table maintains the
additional tables to manage and administer. entries for ids of user(s), tag(s) and resource every time
tag is assigned by the user to a particular resource.
2.3.2. Schemas for Simple and Collaborative Two sample queries are shown in Figures 10 and 11.14
Tagging The first query extracts tags given by any particular user
and the second query retrieves hundred and fifty most
The above presented solutions are altered and improved
popular tags, which can be used to be placed on the tag
by14 and give us general model for simple and collaborative
cloud. Furthermore, resources can also be filtered by
tagging applications. These models can be adopted for a
wide range of applications by making application specific
changes and modifications.

Figure 6. Simple tagging system data model.14

Figure 4. Scuttle solution.20 Figure 7. Query for obtaining tags for a single resource.14

Figure 5. Toxi solution.20 Figure 8. Retrieving tags with number of times it occurs.14

6 Vol 12 (43) | November 2019 | [Link] Indian Journal of Science and Technology
 Fouzia Jabeen and Shah Khusro

3.1. Search via Browsing Approaches, Issues

and Their Impact
3.1.1. Facets
Browsing interface is designed in such a way that it
contains a set of meaningful labels which reflect concepts
of the domain. The name given to such design is called
the facets. The major decisions involved in facets choice
Figure 9. Collaborative tagging system data model.14 are; choice of facets, number of facets, facets chooser, and
evaluation of quality of facets.22
• Choice of facets-given the vast scope of the subjects
covered in atypical social tagging application, the
questions that arise in mind are; how to choose facets
that can be applied generally to all these subjects or
domains? Whether the choice of facets will be tested
by other parties or by the users, and how effectively
Figure 10. Retrieving individual user’s.14
these facets reflect subject domain and clients’ interest?
• Number of facets-we need answers that how many
facets will be sufficient that will cover site members’
needs? There is no agreement on the optimal number
of facets to use in social tagging applications.
Impact-multiple facets empower a more flexible,
improved and enhanced search. It supports
classification of objects of interests along several
dimensions of metadata.23,24
Figure 11. Retrieving 150 most accepted tags.14 • Facet chooser-we need to decide who will be the
chooser of the facets? The choices can be: site mem-
bers, site editors, and researchers.
using more than one tag in combination. We can say that
Impact-the diverse nature of the site members in
this model give the provision of handling an arbitrary
such applications will make reaching of consensus a
number of tag combinations.
difficult task.
Once a list of facets is chosen, the set of questions
3. Browsing and Searching could be added to the user’s interface, so that when the
site members create or re-use a tag, the questions could
Searching, browsing, and querying are prominent and
help them determine in which facet to place the tag.25
effective means used for information finding. These
methods are utilised according to the objective of • Quality of facets-we have to answer these questions:
information seeker. For example, sometimes information whether facets that are chosen reflect the needs of the
seekers require a direct “right answer”, while sometimes site members? The extent to which the facets reflect
they intend to retrieve several relevant related sources and the needs of the site members? Their ease of use, how
sometimes in-depth search is necessary and then there much they are used, and whether they are used cor-
are times when required information have to be re-found. rectly (e.g., location-type tags are placed correctly in
Tagging is effective in “exploratory seeking,” that is, when the location facet). Some site members may place tags
the goal of searching is to gain a broad perspective of in the correct facets in that case how the use of the
the information available about a particular topic.21 In facets will be monitored?
both ways of search either by browsing or by query, the Impact-if facets are not used correctly, then their
difference is the cognitive load. potential benefits may not be realised fully.

Vol 12 (43) | November 2019 | [Link] Indian Journal of Science and Technology 7
 Core Architecture and Design Issues of Collaborative Tagging Applications

3.1.2. Tag Cloud related tags into bundles. However, one tag may appear
Browsing is supported by visualisation techniques like Tag in many bundles33.
Cloud.26 Tag clouds perform best in browsing scenarios Bibsonomy tag hierarchies34; sites do not limit on the
where specific target is not clearly known. Tag cloud alone ways of use of hierarchies. However, these are mostly used
is not sufficient enough for information retrieval tasks, in finding out relationships (broader/narrower (most
but could be used to boost a user’s search activity.27 common)).
There are a number of researches conducted to find
• Shape (Layouts of tag clouds28)-which layout is a more relatedness among tags, resources, and users.35–38 Finding
appropriate circle, rectangle, a square, 2D, or 3D. Sec- of related folks helps in finding group of people with
ondly, how tags should be arranged whether alphabet- common interests which further helps in finding related
ically or according to popularity (In circle form, least resources and tags.35–38
important to most important).
Impact-semantic, folksonomy-based layouts perform
3.2. Query Based Searching
significantly better in searching as compared to random 3.2.1. Query Based/Keyword Search
layouts. However, semantic layouts should only be used Searching via tag is analogous to search by
when the quality of tag organisation can be assured.29 keyword.39 To facilitate searching via query mostly
• Number of tags-it is required to determine the num- collaborative applications provide query expansion and
ber of tags to be displayed in a tag cloud. Impact-the recommendations.
selection should be based on the fact that it puts least • Recommendations-recommendations can be Explicit
cognitive load on the user. or Implicit. In Explicit recommendations, user per-
• Coloring, visual effects and interpretation of visual sonally sends his/her recommendation(s) to other
effects-visual effects gain audience attention. Like large user. Implicit recommendation depends on the user’s
font size attracts user more towards a tag than a tag behavior that is what he/she has searched for and what
with small font size. The meaning hidden behind the articles he/he has viewed.39
font size is the frequency of use. Web Designer Wall1 Impact-quality and coverage of the recommended tags
has a small tag cloud that uses size to indicate impor- is a very important matter to look into: first, indexing
tance and frequency. terms automatically generated by search engines are
Impact-font size strongly affects recall.30 good for textual data but are not significantly useful
• Selection of tags-how to select tags to be appeared on for labelling and organizing resources which include
the tag cloud. Flickr Tag cloud displays the most popu- video, photos/pictures, and music.40,41 Second, this
lar tags of all times and hot tags of last 24 hours. Deli- facilitates the users/taggers beyond the annotation
cious displays list of top tags and also individuals have process. For instance, approaches that suggest tags
their own tag clouds.31 considering relations “is-a”, “part-of”, “tag inheritance”
Impact-tags on tag cloud can be selected keeping one are extremely useful in increasing indexing depth of a
or more goals in mind. An example of tag selection based resource.
on single-objective would be to select only popular tags. • No. of tag recommended-there is no hard and fast rule
A multi-objective algorithm, on the other hand, would regarding number of suggested tags. However, two to
keep balance among two or more goals, e.g., selecting tags twenty tags seem to be optimum.
that are popular and provide excellent coverage.32 Impact-according to the author of42 more tags of
the recommendation are regarded, better will be the
3.1.3. Grouping Related Tags, Resources and recall and worse will be the precision.
• Query modification-query can be modified either by
Folks
substitution or by expansion. In substitution original
Related tags, resources, and folks can be grouped to enrich term(s) are removed and replaced with new term(s).
browsing and to achieve required search objectives. The Query expansion aims to enrich and elaborate query
social bookmarking site [Link] allows folksto put with related classes or senses. The question arises what
1
terms/tags should be provided for query expansion,

8 Vol 12 (43) | November 2019 | [Link] Indian Journal of Science and Technology
 Fouzia Jabeen and Shah Khusro

for example, choices can be ontology terms, local or • Query types-we need to decide whether search is pos-
global terms,43 multilingual tags,44 etc. sible with compound tags? Is there is support for bool-
Global analysis is based on finding alike words for ean search?22 Personalised search is allowed or not?50
expansion of a query. These words can be obtained Impact-with facets, resource collection can be
using different techniques like from co-occurrence filtered and retrieved in a well organised way. However,
information, external data source (Wikipedia/ authors Choi & Street51 objected that the user point
DBpedia, WordNet, or web search results).16 In of view is not mirrored in facets. Personalised search
local analysis, analogous terms are retrieved from considers user taste, interests and needs. This is very
most relevant documents, which are then used for important as users’ interest changes with time.
expanding the original query.
Impact-an expansion is a much safer form of query
modification as compared to substitution.45,46 The
4. Relevant Technologies
objective of query expansion is to improve recall and At minimum, a social tagging system will have a client
precision by enriching result set with most relevant or interface; users will be able to interact with the system
at least equally relevant pages.47 Expansion is much using this interface. In addition, security measures are
more effective to precise the search results when implemented to provide required access levels to the users
query contains abstract terms. Expanding queries will and prevent the system from spammers or unauthorised
increase clarity, reduce ambiguity (disambiguation).48 users. Second, application server layer, for which Apache
Major causes of query expansion failure are: topic drift Tomcat web server can be one of the option or the tagging
and poor quality of query.49 system, can be relegated to cloud-based services such as

Table 1. Categorical overview of API support for collaborative applications development

Category API Comments Protocol Data Usage limit

format
Search and Jinni1 Movie discovery and search SOAP XML
recommendations service
Sapo tags2 Tag search service REST RSS
TagTooga3 Tag based Internet directory REST XML 1000 queries/
day, 1/sec max
Evri Recommendations and semantic REST, XML, 10,000 requests
search service JavaScript, JSON, per day.
Cocoa JSONP
[Link] Tag recommendation service REST JSON
XML
Tagyu Tag recommendation service REST XML
Content annotation GroupMe Social bookmarking REST RDF, RSS
OpenDover Semantic tagging webservice REST, SOAP XML, JSON 100 requests
within the hour
Steve In Action Museum social tagging project REST-RPC JSON
Tagatum API: Blog Blog tags repository services REST XML
tags repository services
Reflect Biology and chemical tagging REST XML
service
Wikimeta Semantic tagging and content REST XML, JSON 100 requests &
annotation service 1 MB data/day
Interface Tagul Tag cloud creation service XML-RPC SVG
Word Cloud Maker Word clouds from blocks of text HTTP POST JSON None

Vol 12 (43) | November 2019 | [Link] Indian Journal of Science and Technology 9
 Core Architecture and Design Issues of Collaborative Tagging Applications

Table 2. Frameworks support for collaborative applications

Frameworks Open Free Developed Database Folksonomy Prominent features

source in support support
Xoops with No Yes PhP MySQL – • Lightbox2 features + jquery tabs based simple
Yogurt front end
Extension
Mahara Yes Yes Run on MySQL Yes • Learning54–56 management system Moodle4
LAMP
Anahita Yes Gpl2 – – Yes5 • Ajaxed interface
Social License • Powerful profile management & messaging
Engine capabilities
• Supports OpenID & integration with Amazon
Cloud Storage to ease the management of
growing data
Elgg Yes Yes Apache 2 MySQL cross-site tagging6 • Social networking
PhP comes with a • Strong access control lists
JSON API • Multiple view support (e.g., iPhone, cell
phones)
• Templating engine
• Widget framework and more
BuddyPress Yes Yes PhP – – • WordPress theme compatibility
• bbPress discussion forums
• BuddyPress widgets
YaCS Yes PhP MySQL Free form • Feature-rich chat
Tagging7 • Multiple identities
• Guest books
• Multiple communities
Pligg Yes Yes PhP MySQL No • Five star voting mechanism similar to Digg,
Mixx and Reddit
• Vote ranking
SocialEngine Yes No PhP – Plugin available • Allow multiple social frameworks(two way
for enabling friendship, one way follower, subnetworks for
tagging8 grouping)
Hashtag pluging9 • Customisable widgets
• Membership levels for community
Ektron No No Visual – Yes • Full featured WYSIWYG editing
studio/.Net • Powerful combination of a commercially
licensed .NET CMS with a PHP-based open
source search system9
Pinax Yes No Django – Yes • Support multiple channels (e.g., technical vs
Web business)
Framework • Previewing of blog posts before publishing
• Ability to announce new posts on twitter
• Review comments per post for multi-author
workflows
• Public but secret urls for unpublished blog
posts for easier review

10 Vol 12 (43) | November 2019 | [Link] Indian Journal of Science and Technology
 Fouzia Jabeen and Shah Khusro

statusNet Yes No PhP MySQL Yes • Able to use NoSQL technologies (e.g.,
memcached) to increase performance
• StatusNet rely on open protocols like oStatus
and oAuth
KicKapps Yes No – – Yes • Its SaaS platform is powered by the first
Social Graph Engine for web publishers
• The platform includes social networking,
user-generated content, programmable video
players, drag-and-drop widget building,
WidgeAds and other applications that
are tightly integrated with Robust Media
Moderation, member organisation and
management, and reporting
Lovd By Yes Yes Ruby on – Yes • Flickr and YouTube integration
Less Rails • Activity updates and user-to-user messaging
• Insoshi a new social network platform
similar to Lovd By Less

Table 3. Social bookmarking platforms

Bookmarking Open Free Development tools/ Support tagging/ Look like/similarity

source technology folksonomy support with
Akarru Social Yes Yes PHP MySQL Yes [Link]
Bookmarking Engine
Bookmark4U Yes Yes Apache + PHP + MySQL No
Scuttle10 Yes Yes PhP MySQL Yes Delicious
PressMark Yes PhP MySQL Yes Delicious
[Link] Yes Yes PhP My SQL Yes
GetBoo Yes Yes PhPMySQL Yes Delicious
Pligg Yes Yes PhPMYSQL Yes Digg
Laicos Yes Yes [Link] and MySQL Yes Digg
[Link] Yes Perl Delicious clone
Hotaru CMS Yes PhP Yes Digg

Amazon Web Services (AWS). Third, suitable database like open source, freeware nature, and plug-in support. For
NoSQL-based database (e.g., MongoDB) or conventional example, developers prefer elggover social engine because
SQL-based database (e.g., MySQL). Lastly, for inter-layer it is free, customisable and plugging are freely available.
server-client communication, XML or JavaScript Object
Notation (JSON) as data interchange scheme.
Wikipedia provides a comparison between some of
Table 4. Summary of user-related design issues and
the available social networking software.52–58 We have
possible solutions
organised relevant technologies in tables from three
points of views. First, API support for collaborative User-related Membership Community
tagging application’s development. Table 1 is the issues
feature wise analysis of API support. Second, available Possible • Public sign up • Follower
frameworks ready to download and use. The developer’s solutions • Invitation only • Contacts
choice mostly depends on key factors which include • Private/external criteria • Groups

Vol 12 (43) | November 2019 | [Link] Indian Journal of Science and Technology 11
 Core Architecture and Design Issues of Collaborative Tagging Applications

The technology chosen has significant impact on 5. Summary

scalability of a system. Table 2 highlights the feature wise
comparison of major available frameworks. Lastly, Table In this section, we have tabulated the summary in Tables
3 gives an overview of the feature set support of available 4–10. Tables 4–6 present key decisions, their possible
platforms for social bookmarking. candidate solutions related to user, resource, and tag. In

Table 5. Summary of resource-related design issues and possible solutions

Resource-related Resource contribution Original/pointer Privacy Restriction

issues

Possible solutions • User contributed resources • Original resource • Public • File

• System contributed resources • Pointer to resource • Private • Object
• Configurable (public/private)

Table 6. Summary of tag-related design issues and possible solutions

Tag design Permission Tag sharing Selection of tags Tag scope Tag support Tag format and Truth
issues control decisions
Possible • Owner • Public • Level Wise • Narrow • Blind tagging • No restriction • File
solutions • Friend • Private • Public/private Tagging • Viewable • Single word • Database
• Tag • Group groups • Broad tagging • Restriction on
limited • Popular tags Tagging • Suggestive certain words
resources Tagging

Table 7. Design issues and their impact on different aspect of social applications

Design Portability Quality of Collaboration Turnover Activity Dynamism Volume of Motivation

issues gathered of resource collection to tag
vocabulary/ and tags
folksonomy

Membership No Yes No Yes Yes Yes Yes Yes

Community No Yes Yes Yes Yes Yes Yes
User No Yes No Yes Yes Yes Yes Yes
contributed
resources
System No Yes No Yes Yes Yes Yes Yes
contributed
resources
Original/ Yes Yes Yes No No No No No
pointer
Privacy No Yes Yes Yes Yes Yes Yes Yes
Tag sharing No Yes Yes No Yes Yes No Yes
Tag scope No Yes Yes No No No No Yes
Tag support No Yes Yes No Yes Yes Yes Yes
Tag format No Yes No Yes Yes Yes Yes Yes
and control
decisions
Truth Yes No No No No No No No

12 Vol 12 (43) | November 2019 | [Link] Indian Journal of Science and Technology
 Fouzia Jabeen and Shah Khusro

Table 7, we have analyzed impact of these design issues query-based approach is efficient if objective of search is
on various aspects of collaborative applications. Tables clear. In addition to providing facilitation in browsing,
8 and 9 are all about the issues and candidate solutions facets are helpful in organizing folksonomy.
related to browsing and query-based search. Lastly, in
Table 10, we have analyzed implications of browsing and 6. Conclusions and Developing
searching approaches. Both browsing and searching have
Trends
same objective that is to bring precise results in front
of users. However, facets or tag cloud work well when Collaborative tagging applications gained popularity
query objective is opaque and broad. In contrast, direct because of their wide spread adoptability among users. In

Table 8. Summary of browsing-related design issues and possible solutions

Facet navigation Choice of facets Number of facets Facet chooser Testing quality of
issues facets
Major possible • Reflect subject • No specific number • Site member(s) • Site editor
solutions domain • Must cove site member’s • Editor (s) • Researcher
• Reflect user needs • List of question to help
interest users
Tag cloud issues Shape Number of tags/selection Coloring, visual effects Arrangement of tags
of tags and interpretation
Major possible • Square • Most popular tags • Big Font represent • Alphabetical
solutions • Circle • Representative tags of most popular • Circular (least to
• 2D, 3D a site • Near tags represent most used tags)
• Semantic • Ontology tags related tags • Coverage
• Folksonomy-based
Browsing via tag Techniques to Techniques to discover Techniques to discover Ranking resources
bundles/related tags/ discover tags resources relatedness folks relatedness
related resources/ relatedness
related folks
Major possible • Co-occurrence Similar tags assigned to FolkRank58 FolkRank
solutions • Subsumption similar resources57
• Synonyms

Table 9. Summary of query-related design issues and possible solutions

Recommendations Recommended terms Number of tags recommended

Major possible solutions • Implicit • Depends on p-core59
• Explicit • Ratio of precision and recall
• Personalised
Query modification issues Types Terms used for substitution/expansion
Major possible solutions • Substitution • Mapping with Wiki, DBPedia, WordNet
• Expansion • Global/Local
• Ontology terms
• Multilingual equivalent
• With respect to user, resource and tag context
• Concrete rather than abstract terms
Query types Types of queries support
Major possible solutions • Support for compound tags
• Support for boolean search
• Personalised search
• Temporal

Vol 12 (43) | November 2019 | [Link] Indian Journal of Science and Technology 13
 Core Architecture and Design Issues of Collaborative Tagging Applications

Table 10. Browsing and searching implications 3. Eltemasi M, Nadar N, Niusha ZF. Communication with
folksonomy in technical libraries. In: 2011 international
Browsing Folksonomy Precision Objective Objective conference on social science and humanity; 2011. P. 479–82.
and organisation in search of a query of a 4. Difference between taxonomy and folksonomy
searching and is clear query is
wordpress news designbit. [cited 2015]. http://
approaches retrieval vague
[Link]/2010/10/10/difference-between-t
Tag cloud No Yes No Yes axonomy-and-folksonomy/.
Facets Yes Yes No Yes 5. Lachica R, Karabeg D. Metadata creation in socio-semantic
Query No Yes Yes No tagging systems : towards holistic knowledge creation and
interchange. In: International conference on topic map
research and applications; 2007. P. 160–71.
this study, we have reviewed and highlighted the design
6. Folktology. [cited 2012]. [Link]
issues, the possible solutions and their impact along with
net/(S(ghbqnlrfgoh13l3zavdexl45))/artificialmemory.
a comparative analysis of the feature set support provided aspx?ID=22294.
by the available technologies. 7. [Link]:Folktology. [cited 2011]. [Link]
This study covers key decisions that the designers’ [Link]/2011_09_01_archive.html.
needs to take when developing a collaborative tagging 8. Measuring the quality of an e-government folksonomy.
application by keenly looking at the implications of these [cited 2008]. [Link]
decisions on the performance of the system. These key papers/E-government_Folksonomy.pdf.
decisions have great impact on the major aspects of a 9. Enhancing navigability : an algorithm for constructing tag
collaborative application which include, flexibility they trees. [cited 2017 Mar 21]. [Link]
provide to the users, the emerging folk vocabulary quality view/journals/jdis/2/2/[Link].
and indexing depth, precision in search and retrieval, 10. A collaborative tagging system design framework for
supporting affective design. [cited 2015]. [Link]
success of an application measured by observing turnover
[Link]/document/7385677.
and activity. In addition, choice of appropriate technology
11. Discovering product feature and affective associations
plays a significant role in scalability and security of a through collaborative tagging. [cited 2014]. https://
collaborative system. To end with the developing trends in [Link]/document/7058789.
this area of research includes (1) augmenting collaboration 12. What are social tags? [cited 2015]. [Link]
with semantics and context. Contextual collaboration has org/2009/02/what-are-social-tags/.
the potential to improve both the working experience 13. The structure of collaborative tagging systems. [cited 2005
and the experience of working together (2) visualisation Aug 18]. [Link]
based collaborative system (3) deep learning models are 14. Tagging: people powered metadata for social web.
effectively used for dealing with unlabeled resources and/ [cited 2007]. [Link]
or recommendations. 252059367_Tagging_People-Powered_Metadata_for_the_
Our critical review presented in the study will be Social_Web_Smith_G_2008_Book_Review.
15. Social tagging & folksonomies : indexing retrieving and
beneficial in making right choices by the designers and
beyond ? [cited 2011 Jan]. [Link]
developers of tagging applications.
publication/261541547_Social_tagging_folksonomies_
Indexing_retrieving_and_beyond.
References 16. Tree, funny, to _ read, google: what are tags supposed to
achieve ? A comparative analysis of user keywords for
1. Weller K. Folksonomies and ontologies: two new players different digital resource types. [cited 2018 Jan]. https://
in indexing and knowledge representation. Apple Web [Link]/publication/29858360_Tree_
Creation. 2007;2:108–15. [Link] funny_to_read_google_What_are_Tags_Supposed_to_
general/new-themify-flow. Achieve_A_Comparative_Analysis_of_User_Keywords_
2. Understanding the user: personomy translation for for_Different_Digital_Resource_Types.
tag recommendation. [Link] 17. Social tagging of services to support end user development in
publication/268330617_Understanding_the_user_ ubiquitous collaborative environments, dissertation. [cited
Personomy_translation_for_tag_recommendation. 2007 Jun]. [Link]

14 Vol 12 (43) | November 2019 | [Link] Indian Journal of Science and Technology
 Fouzia Jabeen and Shah Khusro

18. Tagging, communities, vocabulary, evolution. [cited 2007]. [Link]

[Link] 04114-8_14.
19. Tag recommendation for folksonomies oriented towards 34. Content-based and collaborative techniques for tag
individual users. [cited 2008 Jan]. [Link] recommendation: an empirical evaluation. [cited 2013 Feb].
net/publication/228632258_Tag_recommendation_for_ [Link]
folksonomies_oriented_towards_individual_users. Content-based_and_collaborative_techniques_for_tag_
20. Faceted navigation of social tagging applications. [cited recommendation_An_empirical_evaluation.
2011]. [Link] 35. Uddin MN, Duong TH, Nguyen NT. Semantic similarity
21. Partitioning instead of clustering: an alternative approach measures for enhancing information retrieval in
for mining structured content in folksonomies. [cited folksonomies. Exp Syst Appl. 2013;40(5):1645–53.
2016]. [Link] 36. Xu G, Zong Y, Jin P. KIPTC: a kernel information
instead-of-clustering-an-alternative-approach-for-mining- propagation tag clustering algorithm. J Intell Inf Syst.
[Link]. 2015;45(1):95–112.
22. Finding cultural heritage images through a dual-perspective 37. Folksonomies. Indexing and retrieval in web 2.0. [cited
navigation framework. [cited 2016]. [Link] 2009]. [Link]
[Link]/34253/. bnegC&pg=PA289&lpg=PA289&dq=Browsing+suppor
23. Faceted search. [cited 2009]. [Link] t+in++folksonomy+system&source=bl&ots=ZJOeG02Z
com/doi/abs/10.2200/S00190ED1V01Y200904ICR005. hF&sig=QOQo8Oj-32KqVQ0R3UzGcPik2RM&hl=en-
24. Li WD, Qiu ZM. State-of-the-art technologies and &sa=X&ei=qAhOU72REYXgygP-tYHwDA&ved=0CDYQ
methodologies for collaborative product development 6AEwAQ#v=onepage&q&f=false.
systems. Int J Prod Res. 2006;44(13):2525–59. 38. Adapting K-nearest neighbor for tag recommendation in
25. Seeing things in the clouds: the effect of visual features on folksonomies. [cited 2009]. [Link]
tag cloud selections. [cited 2008]. [Link] [Link].
publications/2008/[Link]. 39. Folksonomic tag clouds as an aid to content indexing. [cited
26. Lohmann S, Ziegler J, Tetzlaff L. Comparison of tag cloud 2009 Nov 21]. [Link]
layouts: task-related performance and visual exploration. 40. Tag recommendations in folksonomies. [cited 2007]. https://
In: IFIP conference on human-computer interaction; 2009. [Link]/chapter/10.1007/978-3-540-74976-9_52.
P. 392–404. 41. Query expansion in folksonomies. [cited 2011]. https://
27. Semantically structured tag clouds : an empirical evaluation [Link]/[Link]?id=2032131.
of clustered presentation approaches. [cited 2009 Jan]. 42. Jung JJ. Cross-lingual query expansion in multilingual
[Link] folksonomies: a case study on Flickr. Knowl-Based Syst.
Semantically_Structured_Tag_Clouds_An_Empirical_ 2013;42:60–67.
Evaluation_of_Clustered_Presentation_Approaches. 43. Query reformulation using anchor text. [cited 2010].
28. Getting our head in the clouds: toward evaluation studies [Link]
of tagclouds. [cited 2007 Jan]. [Link] php?id=900.
net/publication/221517909_Getting_our_head_in_the_ 44. Analyzing web text association to disambiguate abbreviation
clouds_Toward_evaluation_studies_of_tagclouds. in queries. [cited 2008 Jan]. [Link]
29. What is a tag cloud and how do tags connect people. [cited publication/221299938_Analyzing_web_text_association_
2009 Mar 06]. [Link] to_disambiguate_abbreviation_in_queries.
blog/what-is-a-tag-cloud/. 45. Wikipedia. Query expansion – Wikipedia, the free
30. On the selection of tags for tag clouds. [cited 2011 Jan]. encyclopedia. [cited 2015]. [Link]
[Link] Query_expansion.
the_selection_of_tags_for_tag_clouds. 46. Gold-standard evaluation of a folksonomy-based ontology
31. From the crowd to communities: new interfaces for learning model. [cited 2018]. [Link]
social tagging. [cited 2008 Jan]. [Link] article/10.1088/1742-6596/971/1/012045.
net/publication/228529099_From_the_crowd_to_ 47. Studying query expansion effectiveness. [cited 2009]. http://
communities_New_interfaces_for_social_tagging. [Link]/publications/[Link].
32. Constructing folksonomies from user-specified relations on 48. He B, Ounis I. Studying query expansion effectiveness. In:
flickr. [cited 2009]. [Link] Advances in information retrieval; 2009. P. 611–9.
[Link]. 49. A personalized recommender system based on users’
33. Tag relatedness using laplacian score feature selection information in folksonomies. [cited 2013]. [Link]
and adapted jensen-shannon divergence. [cited 2014]. [Link]/companion/[Link].

Vol 12 (43) | November 2019 | [Link] Indian Journal of Science and Technology 15
 Core Architecture and Design Issues of Collaborative Tagging Applications

50. Bringing a more accurate user’s perspective into web paper/5352-semi-super vised-learning-with-deep-
navigation: facet analysis of folksonomy tags. [cited [Link].
2012 May]. [Link] 55. TRSDL : tag-aware recommender system based on deep
4 2 7 5 9 2 9 4 _ B r i n g i n g _ a _ Mo r e _ Ac c u r at e _ Us e r ’s _ learning – intelligent computing systems. [cited 2018 May].
Perspective_into_Web_Navigation_Facet_Analysis_of_ [Link]
Folksonomy_Tags. TRSDL_Tag-Aware_Recommender_System_Based_on_
51. Bringing a more accurate user’s perspective into web Deep_Learning-Intelligent_Computing_Systems.
navigation : facet analysis of folksonomy tags. [cited 2012 56. Effective retrieval of resources in folksonomies using a new
May]. [Link] tag similarity measure. [cited 2011]. [Link]
Bringing_a_More_Accurate_User’s_Perspective_into_ [Link]?id=2063657.
Web_Navigation_Facet_Analysis_of_Folksonomy_Tags. 57. Hotho A, Jäschke R, Schmitz C, Stumme G. Information
52. Comparison of social networking software. [cited 2019 Jul retrieval in folksonomies: search and ranking. In: European
08]. [Link] semantic web conference; 2006. P. 411–26.
networking_software. 58. A combined method for mitigating sparsity problem in tag
53. Collaborative deep learning for recommender systems. recommendation. [cited 2014]. [Link]
[cited 2015 Jun 18]. [Link] document/6758716.
54. Semi-supervised Learning with deep generative
models. [cited 2019]. [Link]

16 Vol 12 (43) | November 2019 | [Link] Indian Journal of Science and Technology