In order to make the Semantic Web reach “real world” end-users it is important to consider Semantic Web usability. User-Centered Design from the Human-Computer Interaction community might help in this respect. First of all, the user must be defined, together with the context. Then it is possible to study user tasks. We focus our study of Semantic Web user tasks in endusers and Semantic Web online applications, trying to contribute to establishing some UCD guidelines that help the adoption of Semantic Web applications. However, we consider existing analysis for Web systems and even online information systems in general in order to avoid constraining our view to the current state of development of the Se-mantic Web. The proposed set of end-user Semantic Web tasks is Search, Browse, Annotate, Mashup, Map, Share, Communicate and Transact. They are used in order to study an existing Semantic Web platform and an application based on it. This allows putting the tasks into practice and relates them to some interaction patterns. Future work continues in this line, trying to connect the identified patterns with existing and new interaction pat-terns in order to contribute additional guidelines for UCD Semantic Web applications development.
The Semantic Web has been around for
some time and many people are asking why it
has not taken off as quickly as the World Wide
Web did [
Once Semantic Web technologies seem to
be quite mature, in order to facilitate its
adoption, it is time to focus on the face
Semantic Web applications show to users [
In the context of HCI, User-Centred Design
(UCD) proposes facing the development
process of interactive systems focusing on the
user and considering the Quality in Use,
standardised in ISO/IEC 25010:2011 [
The tasks supported by the early Web are
now neatly defined and are becoming part of
Web developers’ common practice, making it
relatively easy to develop tools adapted to these
tasks following a UCD approach. Knowledge
about tasks in the Semantic Web is much less
clear due to its novelty, but it is necessary in
order to be able to take a user-centred view on
the Semantic Web [
However, before tasks can be identified, the first step is to determine who the intended users are and their context. There might be the temptation to say that users are "everyone", but UCD recommends dividing the target audience into groups. Considered that the objective is to facilitate Semantic Web adoption among endusers, this seems the target audience to focus on. On the other hand, the context is any online application based on Semantic Web technologies, which seem the best channel for a widespread adoption of the Semantic Web.
In the context of this paper, end-user is defined as a user with no or limited knowledge about Semantic Web technologies and methodologies. We do not include in this user profile domain experts, that might be also users with limited knowledge about the Semantic Web but who have specific needs related with the development of ontologies.
Once we have characterised the kind of Semantic Web user we are interested in, it is time to determine the tasks. As we are considering the Semantic Web as a whole, and not a specific Semantic Web application, the tasks should be generic and broad enough to accommodate tools that are not yet Semantic Web enabled but that might be so in the future.
In order to make the range of tasks broad enough and avoid constraining our view to the current state of development of the Semantic Web, the focus should be broader than existing Semantic Web applications and studies of Semantic Web user tasks. We also consider user tasks in the context of the Web and even in the context of information systems. This makes it possible to check the consistency and coverage of the proposal. The range of user tasks studies under consideration is presented in Section 2.
From the analysis of these existing studies,
which range from Semantic Web to information
systems user tasks, we build our proposal of a
set of generic Semantic Web end-user tasks,
which is presented in Section 3. Then, in order
to study these tasks deeper, we have then put
them into practice in Section 4. We have
isolated them in the context of Rhizomer [
Due to the importance of a clear definition of the user tasks when developing interactive systems, especially if an UCD approach is followed, there are many studies of this kind. Usually, they refer to the user tasks for a concrete application. However, there are also studies that consider a range of applications in order to help defining guides or analysing common practices.
When analysing existing work, we have considered different studies at different scale levels, from the more general online information systems, through Web information systems, to Semantic Web applications in general and finally some specific Semantic Web scenarios. The objective is to not pass through potential tasks that might be considered in the context of the Semantic Web but that haven't been considered yet in that context.
Therefore, starting from the broader context,
Heath et al. [
Locating is about users looking for something known or expected to exists. Exploring refers to gathering information to gain understanding or background. Grazing is moving speculatively without a specific goal. Monitoring is about checking known source expected to change. Sharing refers to making something available to others. Notifying is informing others about something that happens. Asserting is about making statements of fact or opinion. Discussing refers to exchanging information on a topic with others. Evaluating is determining if some information is true or alternatives. Arranging is about coordinating with third parties. Transacting is transferring money.
Getting into a more specific context, Kellar
et al. [
Fact finding is usually a short task that stands for looking for specific pieces of information. Information Gathering involves the collection of information and Browsing is a serendipitous task where users have no specific goal in mind. In relation with the information exchange goals, there are Transacting, which stands for performing an on-line action that often involves user/password authentication, and Communicating, connected to web-based communication, e.g. e-mail or blogs.
The last goal defined by Kellar is information maintenance that includes just one user task, Maintaining. This task is about editing web resources in order to make them work properly, e.g. no broken links, and update them. Kellar et al. also consider a potential task, Monitoring as returning to a previously visited page in order to obtain updated or dynamic information. However, they do not consider a user tasks per se, but a task dimension, reoccurrence, that might be a characteristic of any of the previous user tasks, especially Browsing, Transacting, Fact Finding and Information Gathering.
If we concentrate now on Semantic Web
applications, Battle [
On the other hand, Mäkelä et al. [
Finally, Sabou et. al [
From the analysis of the existing literature,
Semantic Web applications and our experience
with the Rhizomer platform [
Though the objective is to define a set of
generic tasks, not specific to a particular
application, it is important to provide a minimal
characterisation of the user and avoid defining
them as just any user. This is also motivated by
the fact that the objective is to define the
foundations for UCD of applications to be
adopted by as many users as possible. The most
populated user profile is end-users, that from
the point of view of the Semantic Web will in
most cases stand for user with no or limited
knowledge about the Semantic Web
technologies and methodologies [
One particularity is that we do not include in this category domain experts with some knowledge to be formalised as ontologies. They might have limited knowledge about the Semantic Web particularities, but they might not be considered end-users because they have very different objectives, and consequently they will carry out quite different tasks in order to accomplish them.
The users we are considering are used to web and other Internet applications like electronic mail or instant messaging. Therefore, we should also consider a broader categorisation of tasks users carry out online. It is necessary to consider a wider range of tasks because, although the Semantic Web is not widely deployed right now, its opportunity is to underpin the whole range of online user experiences and contribute new ways to do things in a more usable and accessible way.
The next subsections present the set of generic Semantic Web end-user tasks we propose. Each task is first considered from the point of view of an end-user, i.e. without considering the particularities of the Semantic Web. Some examples of particular end-users’ tasks are then presented, together with references to the related tasks in the literature previously analysed. Finally, the tasks are analysed deeper, considering what the Semantic Web might contribute to them, what technologies and methodologies make them possible and constitute and added value for them.
All the considered tasks are basic ones. Usually, in the context of concrete Semantic Web application, user tasks will be composed of a mixture of these basic tasks. The objective is to define a basic set of user tasks, that would facilitate tasks analysis and UCD while being easily combinable in order to derive more complex and specific user tasks. In order to illustrate these features, this categorisation of tasks will be put into practice in order to analyse a generic platform for Semantic Web applications in Section ¡Error! No se encuentra el origen de la referencia., where we anticipate that basic user tasks will. In the latter, the idea is that more specific user tasks will be detected, which might be built from one or more of these basic Semantic Web user tasks. 3.1.
In general, this kind of tasks corresponds to those when a user poses a query and obtains a set of results that might be rendered in different ways. We include here when the search might be delayed or repeated in the future, like in monitoring scenarios.
Concrete examples of this task are when a user performs a simple keyword-based search using a Web search engine, an advanced query that allows constraining different search dimensions, query by example, monitoring elections results or a sports match, etc.
In the context of the Semantic Web, the user
can benefit from the implicit semantics when
performing a search and get more accurate
results, i.e. higher precision and recall.
Moreover, the knowledge captured in
ontologies can be used in order to guide the user
through the query construction process, in order
to facilitate query by example or results
presentation [
However, it is important to consider that most users are used to perform this kind of tasks by simple means like an input field where they type the keywords they are interested in. Consequently, they might be confused if a more sophisticated form or syntax is required in order to pose a query.
In some cases, and after some user testing, it might be concluded that it is preferred to hide all these subtleties from the user, to keep a simple user interface, make use of the available semantics as part of the query engine internal mechanisms and exploit the semantics from the point of view of user interaction when presenting the results and as part of the browsing user tasks, presented next.
This task includes Locating and Monitoring (Heath), is similar to Fact Finding and considers the temporal dimension of monitoring scenarios (Kellar). It also considers Information Seeking (Battle) though some of the examples Battle et al. propose may require other tasks, e.g. some sort of mash up in the case of combining news from different news sources. This task is also related to Semantic Content Consumption (Mäkelä), though it is more specific because Semantic Content Consumption also includes when users browse search results. Moreover, it is also a component of the complex user task Word Sense Disambiguation (Sabou). 3.2.
This task is performed when the user moves through the information currently displayed. In the context of Web information systems this is usually done by following the links that connect the information to related information pieces.
Concrete examples of this task are when a user gets informed about the latest news, reading blogs, entertainment, listening to music, viewing movie trailers, to follow a link received in an e-mail, etc.
In the context of the semantic web, it is
possible to build a richer browsing experience
because the underlying model is built from
component of a smaller granularity, the triples
formed by a subject, a predicate and an object.
The combination of many triples builds up a
graph. This graph might be browsed by
following the links between graph nodes
following different criteria, not but just
showing the graph structure to the user [
For instance, those triples might come from
different “documents”. All the triples from a
document, identified by a URI might be
displayed to the user, who can follow the links
to external documents or browse the current
data if it is not displayed all at once. An
example of this behaviour is followed by
Tabulator [
Another alternative is to provide a faceted
view if the metadata being browsed is
homogeneous, all the resources being browsed
have similar properties describing them. This is
possible using tools like Exhibit [
This task is related with both Exploring and Grazing (Heath), and also with Browsing (Kellar). Some of the examples of Information Seeking (Battle) also include aspects related with this task, e.g. learning more about a topic. This task is also related to Semantic Content Consumption (Mäkelä), though it is more specific because Semantic Content Consumption also includes searching. Additionally, it is also a component of the complex user task Word Sense Disambiguation (Sabou).
In this task the user describes a resource by providing properties and values that model its characteristics, its relations to other resources, etc. This task includes providing a completely new description but also complementing an existing one, modifying it or deleting some or all of the attributes currently available.
Concrete examples of this task are when a user tags a particular URL as it bookmarks it, providing the title and the description of a video, geographically locating a photo, defining a user profile that includes personal details and preferences, etc.
The main particularity of this task, in the context of the Semantic Web, is that the annotations are based on a formal model. Consequently, annotations go beyond informal and ambiguous tags into properties and values that might be constrained by the specifications captured in schemas and ontologies. This feature is not just a way to facilitate machine processing; it might be also as a way to facilitate the annotation task for the user.
The user can benefit from a domain specification defining the available kinds of resources, their properties depending on the resource type and the corresponding values. It is up to the user interface to guide the user through this knowledge space, dynamically constraining the choices to be made depending on previous user actions, the context of use and the intended goals.
An example of a tool giving support to this
task in the context of the Semantic Web is the
Semantic Forms extension2 for Semantic
MediaWiki [
2 http://www.mediawiki.org/wiki/Extension:Semantic_Forms
This task is connected with Asserting (Heath), especially if we consider that the statements made are metadata. It is also related with a broader task that considers maintaining information, Maintenance (Kellar), and also with a more specific one that concentrates on updating content, Content Update (Battle).
Considering tasks identified in the literature in the context of Semantic Web applications, this task is related with Content Indexing (Mäkelä) in the sense that by that task semantic annotations are generated, but just as long as some user intervention is required. Otherwise, it is not a user task but a system task. It can be also related with Ontology Maintenance and Publishing (Mäkelä), though from the end-user characterisation we have made this task lays outside the set of user tasks under consideration.
This task is about the user gathering different pieces of information and combining them in order to get something more than the simple aggregation of those pieces. In other word, the user tries to get something from the aggregation of the data that cannot be or is difficult to obtain from those pieces separately, without combining them into a coherent view.
Concrete examples of this task range from simple mashups such as combining a set of resources that are geographically situated in order to, for instance, which are the hotels near a venue, or resources with temporal dimension that are arranged in a calendar or timeline in order to facilitate scheduling. More complicated scenarios are also possible, which are based on combining the sources of information without a predefined output view, like combining a local list of publications with information about their impact factor in order to compute the overall impact, detect trends, highlight the more relevant publications, preparing a research activity report, etc.
In the context of the Semantic Web, this task involves combining two or more pieces of metadata about common resources in order to aggregate the available descriptions about them. It is also possible that the metadata is about different resources, but in this case, they should be similar in some sense in order to make possible the aggregation in some dimension, e.g. they all have geographical coordinates or are situated in time and can be placed together in a map or timeline respectively.
The main benefit of Semantic Web
technologies and methodologies for this task is
that as semantic metadata and ontologies are
available, it is easier to implement some sort of
assistance for the user during the aggregation
process. The assistance may range from the
ability to propagate the aggregations made to
one particular resource property to all the uses
of that property in the metadata being mashed
up, like in the Potluck mashup tool [
This task is related with, though slightly more specific than, Information Gathering (Kellar) and includes the main characteristics of both Evaluating and Arranging (Heath), which might also involve search and browse but whose added value is about combining information and extracting something more that its pure addition. It is also related with Information Synthesis (Battle) and the Semantic Content Consumption user task (Mäkelä), which is much wider and also includes searching and browsing. 3.5.
Map
This task takes place when the user defines mappings among terms from different vocabularies. It is not constrained to a particular set of resources like in the case of the Mashup task, and it does not operate at the level of particular resource descriptions. On the contrary, in this task, the user is working at the level of the vocabularies. These vocabularies might be used in descriptions for many resources, some of which the user might not be aware of it at the moment.
Results from a mapping task might be used in order to facilitate or automate a mashup, or both tasks might be carried out alternatively and co-ordinately as a process where the user is mashing up a set of resource descriptions and during that process some mappings among the vocabularies being used are defined.
It might be the case that mappings are derived from the analysis of the interaction of many users, however this is a system task, the user does not directly and consciously intervene in this case. Here we are referring to tasks initiated by the user.
Examples of this task range from simple scenarios like stating that two tags are equivalent to more complex ones like relating different product categories or stating that all the things that one repository classifies as papers are also a kind of publication as specified in a second repository vocabulary.
In the context of the Semantic Web, and also considering that we have characterised all these tasks as those for an end-user, this task corresponds to when the user defines simple mappings among classes, properties and values specified in different ontologies. It is not about exhaustive mappings among ontologies but instead about specific mappings that might be usually justified by the need of facilitating mashing up some resource descriptions, or making the mashup more systematic.
This task is a particular case of Ontology Mapping (Battle), geared towards very simple mappings and usually triggered by the system that asks users for confirmation because we focus on end-users that are not ontologists. It is also related with Ontology Maintenance and Publishing (Mäkelä) but that is also a task geared toward ontologists and domain experts, not end-users. The same applies when considering Content Indexing (Mäkelä). The same applies for Ontology Matching (Sabou) while Folksonomy Enrichment might be easier and more appropriate for end-users. 3.6.
This task considers uploading, publishing, updating and deleting pieces of content with the intention of making them available to other users, who can access the content from a place and at a time individually chosen by them. This last statement allows to clearly distinguishing this task from the Communicate task, which is presented next.
This task is also differentiated from Annotate in the sense that what is added, edited or removed is not metadata, data about data, but data itself. This data will usually correspond to different kinds of content that users want to share online, like videos, text, images, etc.
Examples of this task are posting a blog or micro-blogging, participating in a forum, sharing a photo in a social network, making a file available through a Peer-to-Peer network, etc. The participation in forums might also be seen as a communication task, described below, but the intention in most forums is to build a piece of information around a subject and to make it available for later use.
In the context of the Semantic Web, this task, as long as related with data and not with metadata, is not directly supported by Semantic Web technologies and methodologies. However, it might be enriched by triggering some sort of content indexing and automatic metadata generation. The metadata just generated can then trigger an Annotate task, which allows the user editing and managing this metadata.
However, there might be also scenarios
where the distinction between data and
metadata is somehow blurred. For instance, in
the context of Linked Data [
This task is similar to Sharing (Heath) and Information Sharing (Battle). There are not tasks related with this one in the studies of Semantic Web tasks we have analysed. This seems related with the fact that, as we have previously said, in the context of the Semantic Web sharing semantic descriptions is a task included in annotation tasks. 3.7.
This task is about sharing information directly with particular users, without the intention of making it available to other users from a place and at a time individually chosen by them. The process is in this case driven by the user participating in this task as the emitter.
Examples of this task are to participate in a chat, to send an e-mail, video-conference, etc. We have included here e-mails because they are usually kept private and not intended to make them publicly available to other users apart from the recipient. Moreover, the communication is driven from the emitter as long as the recipient has the e-mail client up and listening.
Thought communication management is fundamentally related with other tasks, like searching for a specific e-mail, browsing conversations or annotating an e-mail, there might be room for semantic technologies to play a role during this task. In any case, this is one of the areas with less results coming from the Semantic Web, and also the one that seems to provide less room for them.
For instance, Haystack [
This task is related with Notifying and Discussing (Heath). 3.8.
This is the last task, it is associated with user actions that provoke a change in the state of a real-world entity or of a resource in a system outside the scope of the system the user is interacting with.
Examples of this task are buying a book, ordering a money transfer between bank accounts, etc. The range of specific tasks included in this category might vary a lot depending on the interactive system attention is focused on. If we concentrate on the user tasks for a specific application, any task that involves interacting with other systems might be considered a transaction as a way to focus the analysis.
On the other hand, if a broader system is considered, for instance any information system, the study might be detailed further and particular tasks among the ones presented before might be identified as the goal of that interaction. In any case, actions that take place in the real world, outside interactive information systems, might be considered transactions in the context of this end-user tasks proposal.
Together with communicate, this is the task that might be less influenced by Semantic Web technologies and methodologies. This is due to the fact that by the definition of this task, they correspond to interactions of the user with systems outside the Semantic Web.
In any case, applications might take profit from these technologies and methodologies while supporting this task before and after the processing outside the Semantic Web takes 3 Rhizomer, https://rhizomer.rhizomik.net place. For instance, by facilitating form filling while the user provides the required data to complete the transaction. Another way to support the transaction might be adapting the results to user preferences and context, for instance performing currency conversions following user preferences.
This task is present in two of the tasks lists considered. There are Transacting (Heath) and Action-oriented (Battle) tasks.
Rhizomer3 is a platform based on Semantic Web technologies that facilitates publishing semantic data and building interactive Semantic Web applications on top of it. Rhizomer differs from semantic web browsers in the sense that it is not just a browser application; there is also a server part that allows defining datasets to be explored and which interacts with the SPARQL endpoints holding the datasets semantic data. However, it is also capable of browsing data not stored in SPARQL endpoints but linked from them.
For instance, if some resource from
DBPedia [
Moreover, Rhizomer also provides mechanisms that facilitate integrating external web services in a dynamic way. The external services to be integrated should be semantically described and those descriptions should specify the kind of resources (classes) the get as input. Rhizomer implements the mechanisms that allow associating at run time the resources classified as being of the input type with the corresponding service.
These associations are implemented as links that allow the user invoking the service, that will receive the resource description as input, process it and return some output. Usually, this output will be HTML content to be integrated into the interface. This way, it is easy to plug in external services that provide bridges to other services but also new ways to interact with resources.
The backend is based on a web application providing an API to interact with the defined datasets and the SPARQL endpoint that constitute them. The backend is built on top of previous API and implemented using a web framework based on HTML and JavaScript that makes the user interface highly interactive. Rhizomer gives support to most of the users’ tasks presented in the previous section, with the exception of the Communicate task: • • •
Search: pose semantic queries using HTML forms, which are dynamically generated and obtain resource descriptions rendered as HTML, as shown in Figure 1.
Browse: navigate through the graph of data retrieving fragments of manageable size and rendering them as interactive HTML, as shown in Figure 2.
Annotate: provide new semantic metadata describing a resource, or edit existing one, using HTML forms that assist the user during this process.
Mashup: mix two or more pieces of metadata about common resources, or resources similar in some sense, e.g. they all have geographical coordinates or are situated in time and can be placed together in a map or timeline respectively.
Map: define simple mappings between concepts from different ontologies.
Share: upload, update and delete pieces of content (HTML, images, videos, etc.).
Transact: generically, this task includes any user action that change the state of a real-world entity or of a resource in a system outside Rhizomer.
In order to make the Semantic Web reach “real world” end-users, special care must be placed in making Semantic Web applications more usable and accessible. It is possible to take profit from the experience accumulated by the Human-Computer Interaction community and apply User-Centred Design approaches. This kind of approaches place the user at the centre of the development process and start by defining the user, its context and the tasks to be performed by them in order to meet their needs. These tasks are specific to a particular interactive application, but it very useful to define a set of common user tasks in the context of a particular domain, e.g. Web information systems, in order to establish UCD guidelines and common interaction patterns for that domain.
This is the main aim of this work, to identify a set of common user tasks for the Semantic Web. However, it is important to concretise the user profile, it is not enough to say that tasks are for any user. In this case, as the aim is to contribute to the widespread adoption of the Semantic Web, the target user is the end-user. This is a user with no or quite limited knowledge about the Semantic Web. The context is any online application based on Semantic Web technologies.
Once the user and the context are defined, it is time to determine the user tasks. In order to consider a broad range of user tasks, it is important to avoid constraining the analysis to the current Semantic Web. The set of Semantic Web end-user tasks proposed is based on the analysis of existing tasks inventories for the Web and even for online information systems in general.
The analysis is complemented with the experience gained implementing a Semantic Web platform and a Semantic Web application based on that platform. Additionally, a list of Semantic Web capabilities has been used in order to complement the process of checking the consistency and coverage of proposed set of end-user tasks.
The set of tasks includes Search, Browse, Annotate, Mashup, Map, Share, Communicate and Transact. Each of these tasks has been described avoiding technological considerations and then presented from the point of view of the Semantic Web. They are also related to the tasks proposed in the literature under consideration.
Finally, the set of tasks has been put into practice and the Rhizomer platform in order to detect if they give support to these end-user tasks and that not additional ones are required. This analysis also allows, in the case of the Rhizomer platform, presenting how these tasks are materialised in the context of the platform as different interaction patterns.
Additionally, it is shown how, as it might be anticipated, the platform user tasks are the basic ones while for the application build on top of Rhizomer, the more complex user tasks can be decomposed into basic user tasks from the proposed set.
Future work concentrates now on the next natural step when following a UCD approach. Once the user tasks have been identified, it is really useful to have an inventory of interaction patterns that give support to these user tasks as a guideline. There are many lists of interaction patterns, though most of them focus on Web systems or other interactive systems without particularising their proposal in the context of Semantic Web applications and Semantic Web user tasks.
Our aim is to build an inventory of Semantic
Web interaction patterns starting from existing
inventories, e.g. Tidwell’s [
For instance, Welie proposes patterns for Browsing or Searching. The objective is to build on top of these pattern libraries and classify them considering the proposed Semantic Web end-user tasks. Many of them might be mapped directly from the Web domain to the Semantic Web, like Welie’s patterns for Search and Browse, two tasks near clear equivalents in the Semantic Web. In any case, they must be studied in detail, new opportunities should be detected and there is also room for novel interaction patterns that the Semantic Web might make possible. On the other hand, our analysis against related work specific for Web information systems shows that the user tasks where the contribution of Semantic Web technologies might be more important, because they a less consider or not considered at all, are Annotate, Mashup and Map.