Content uploaded by Gyuri Lajos
Author content
All content in this area was uploaded by Gyuri Lajos on Oct 23, 2018
Content may be subject to copyright.
WikiNizer™ Research: A Personal Knowledge Graph
Builder Harnessing Freebase Linked Data
András G. Benedek
Hungarian Academy of Sciences
Research Centre of the Humanities,
Úri u. 53., Budapest, 1014, Hungary
+36 20 57 88 453
benedek.andras@btk.mta.hu
Gyuri Lajos
Wikinizer LLA
2Cliffe Hathersage
S 321DE, United Kindom
+36 20 45 56 576
gyuri.lajos@wikinizer.com
ABSTRACT
WikiNizer™ Research (WikiNizeR) is a visual Wiki-like
knowledge orgaNizer which constructs Personal Knowledge
Graphs. By enabling us to visualize “meta” levels of reflection,
WikiNizeR facilitates our sense making and problem solving.
Using Freebase as a linked data source it harvests and
contextualizes semantically structured information, empowering
us to curate a private Knowledge Graph of Things, which a view
also to enhancing collaborative knowledge work. Its users
integrate open web data within the content driven graph
architectures of their personal learning environments. It enables us
to elaborate upon the emergent dynamic graphs which articulate
the typed connections that exist between ‘Things’ of interest,
facilitating the emergence of novel concepts in the associated
complexes of content which are organized into node based
structures. It supports a connectivist learning model; and as a
“self-curating” semantic knowledge management tool it can be
used in problem and project based learning to explore resources
and conceptual relationships, helping us to define learning paths
and workflows, or design meta-level didactic object structures and
activities. It is a tool which helps us track meaning construction in
a situated, intent dependent and dynamic manner. It is a holistic
solution which integrates web research, linked data, annotation,
note-taking and knowledge organization into a Lifelong Personal
Digital Archive of “born reproducible”, ab initio re-purposable,
and re-enactable, Research Objects.
Track: Open Track
1. INTRODUCTION
WikiNizer™ Research (WikiNizeR) is a Wiki like visual
orgaNizer of Personal Knowledge Work and Lifelong Learning.
The Knowledge Architectures that WikiNizeR manages are
constituted as graphs of ‘Things’, captured in atomic notes, which
can be assigned types and properties on the fly. These Things can
also be connected by typed links (links with labels also introduced
on the fly) which generate a graph of the user's growing
knowledge about Things of interest. These links, types, and
properties, because they can be viewed and modified at the meta
level, generate an emerging, situated, Meta Knowledge Graph.
This meta graph sets out what sorts of Things we are dealing
with, and how they can be related in a way that best captures our
intentions and ideas. What emerges is a Self Curated, Personal
Knowledge Graph that is private but which can be shared. Not
losing sight of the collaborative aspects of personal knowledge
work, WikiNizeR, which is the first component of the
WikiNizer™ toolbox, offers collaboration via Google Docs.
WikiNizeR is deeply integrated and compatible with
Freebase. Freebase underlies Google's community curated public
Knowledge Graph which powers semantic search. Freebase is
itself “A community-curated database of well-known people,
places, and things”. It is a semantic wiki-style graph knowledge
base of over 45 million topics, approaching 3 thousand million
facts, and as of the 4th September 2014 with tens of thousands of
meta terms, and growing exponentially. WikiNizeR can also be
viewed as a personal semantic wiki, which like Freebase can bring
the power of semantic search, turning strings to things, to a self
curated knowledge base of things. Beyond being a kind of
Semantic Wiki, WikiNizeR is also a Personal and a Visual Wiki
with meta-reflective meta-design capabilities. Being ‘Personal’
and ‘Visual’ means that it avoids the high levels of cognitive load
which characterises Freebase, but exploits its “soft” (non formal)
semantic capabilities. Like Freebase it is a web app, built initially
for the desktop but with mobile first and offline first principles
adhered to in its design.
Business Idea: We see a gap in the market of knowledge
management solutions which support personal knowledge work
end to end. That is, from the first note you make on a topic of
interest, through the detailed research and presentation of your
findings and ideas, which create re-usable, re-purposable,
reproducible, and re-enactable dynamic Research and Knowledge
Objects, which together eventually constitute a self-organizing
Lifelong Personal Digital Archive of all your work. It provides a
place where you can keep and have the tools that help you to
‘bring to mind what you have in mind’. Our holistic solution fills
the gap in personal knowledge management tools, and is not only
available for desktops, but also for mobile devices.
Personal Knowledge Management: There are hundreds if not
thousands of solutions which provide computer support for
collaborative knowledge work within predominantly enterprise
and institutional settings. In comparison, the number of solutions
for personal knowledge work are miniscule. There are a few
established and powerful solutions, typically for the desktop, and
with price tags over $100. Solutions that would work on all the
users devices are few and far between, and all can be considered
as point solutions. Trapped in their “walled gardens” users lack an
end to end workflow. Instead they create beautiful mind maps,
which are not an integral part of a holistic knowledge work flow.
Access Level: A WikiNizeR closed beta is launching in
September this year. It is being launched as a Chrome only web
app available initially on the desktop. Except for the direct
manipulation graph editor it also already works on Android. A
fully working version for Android is planned to be launched
within a month. (IOS is not supported.) It integrates with Freebase
for Linked Data and uses Google Drive for storing users data,
hence a Google account is required. Core personal capabilities
will be available for free even after the beta period, only requiring
Google Sign In. Since users of WikiNizeR will keep all their data
in their own Google Drive providing a free service does not have
much running costs. This allow us to keep the price of future
premium services lower than it would be if we had to subsidize
free services that would otherwise cost $1 per free user per year.
More advanced capabilities, including higher levels of semantic
collaboration and analysis, will be made available later, but all the
extras would be under $10 dollar per year. The guiding principle
is to get paid for services rendered. “Give away free that which
does not cost money to deliver.” The first release of WikiNizeR
helps to get off the ground the process of bootstrapping which
aims at delivering a next generation concept organization tool, as
articulated in our published papers on WikiNizeR. (Cf. [1-5])
Using existing capabilities we are well on the way to completing
within WikiNizeR a full "Intent Graphical" Model of all its
capabilities. This model is set out as a Capability Graph as an
example of a personal knowledge graph that can be accessed in
WikiNizeR. Even in its current form, without further components
of the WikiNizer™ toolbox, it is already useful as a way of
documenting and visualizing the capabilities of the live system.
As we advance the means by which we can turn these models into
constituents of the running system, we get the benefits of
intellectual manageability, and the improvements in the stability
and extensibility of the system. More importantly as system
capabilities in our Intent Graphic Modelling get themselves
elaborated as knowledge graphs, they become fully explorable,
learnable and personalizable, by the users. The entire user
interface therefore becomes but a specific dynamic visualization
of use case dependent trails within the capability graph. In the
second stage of bootstrapping the system will become user
extensible, and co-evolvable through meta-design, [5,7] turning it
into a Knowledge Augmentation Engine.
2. SYSTEM DESCRIPTION
Representing the process of the growth in our personal
knowledge, and sharing the emerging conceptual structures that
emerge within problem solving contexts, is a major task for
individual learners, and a critical coaching and tutoring problem.
Knowledge obtained from the web or other resources needs to be
personalized, reorganized, and contextualized, in order to filter
our misconceptions. Visualization of information about ‘Things’
and their conceptual organization help us with problem definition,
and consequently their solution. WikiNizeR is a Personal
knowledge Management tool which represents and organizes
“associative complexes” within computer augmented personal
Knowledge Architectures.
Our holistic solution has the following key benefits for personal
knowledge organization carried out on both desktop and mobile
devices:
- Note taking with the assurance that everything you ever “write
down to think” when problem solving could make it into its
solution.
- Your Web research leverages the semantic search capabilities of
Freebase, integrating and deepening the knowledge you find there
within your own Personal Knowledge Graph.
- Annotating and saving screenshots and entire web pages as PDF
files on your Google Drive.
- Building personal knowledge graphs of Things of interest along
with all the emergent meta knowledge needed to organize and
make sense of your learning.
- Focusing on selected parts of your graph, and generating
Propositional Trails which highlight the propositions implied by the
links you have introduced. E.g., if say an “Engelbart” node has a
creation label which makes a connection to the node “Augmentation
Research” the proposition can then be read as saying “Engelbart
created Augmentation Research”. Complementing this conceptual
graph function, as you follow the links from either the note or node
“Augmentation Research”, you see all the facts contained in your
graph organized into a propositional trail.
- Build trails across your accumulating body of personal records
by designating nodes of interest and links to follow.
- Package these trails into Personal Research Objects for working
out solutions to your problems, and then sharing them in ways that
are ab initio explorable, re-purposable and re-presentable in a
whole range of dynamically reproducible and consumable formats
to any required depth.
- Offering a single place to keep all you care about in a Personal
Digital Archive.
- All system capabilities are themselves presented as Knowledge
Objects which comprise behavioral aspects, so that they are
explorable and usable in personalized forms: “As knowledge is
indivisible, ever evolving and expanding, so should the means we
use to manage its growth.” This is critical, the needs of the user,
and the needs of the knowledge management capabilities grow
along with growth of the knowledge, which they create and
manage.
Datasets: Freebase turns strings into things. This is the essence of
Semantic Search. If the Freebase search does not turn up the thing
we are interested in we can always resort to Google Search and
following links. When the user authorizes WikiNizeR to allow
read access to all the files in their drive, they can use keyword
searches for all the documents on their own Google Drive. All
keyword searches within WikiNizeR trigger corresponding search
in Freebase, so there is an opportunity to link the user's notes on a
given topic to their equivalents in Freebase, so that semantically
relevant information could be brought in and extended into a
Personal Knowledge Graph.
How the data is used in the user interface: When you are doing
web research on a topic of interest, start with an integrated Freebase
Search. If it returns something of interest to you, a new node can be
created based on the information you get from Freebase. For
specific type of things in Freebase, like ‘persons’, or other
significant relationships, such as ‘influenced’, ‘influenced by’ (edge
direction matters and is signified with thickness), peers can be used
to find related persons automatically. When you expand a Freebase
node like these you get all the appropriately related things added
automatically to your graph. Some of them will prove to be
irrelevant so they can be removed. The process can be repeated for
some time with the new things that have been added. You can
therefore quickly integrate knowledge about Things that you find in
Freebase into your own personal knowledge graph. Once these
Things are present in this form, the user can add new links and
additional deeper knowledge found in Freebase, to follow one’s
interest. (Cf. http://alpha.wikinizer.com/LinkedUpChallenge/augment1.png)
Within WikiNizeR the user deals with 7 kinds of things: Notes,
Links, Types, Properties, Reference Nodes, Knowledge Objects,
and Research Objects . “Research Objects” describes an emerging
approach to the publication, and exchange of scholarly
information on the Web which aims to improve its re-use and
reproducibility. A Research Object within WikiNizeR is a
Knowledge Object with an executable behavioral specification.
Notes are the atomic units of the system. They have a local unique
ID, a title, icon, stub and a body. When creating a note use a
Google image search to pick a suitable icon possibly using
Awesome Screenshot Chrome extension to capture images from
parts of web pages. The granularity of a note is set by the
requirement that considered in itself can be the target of
meaningful, labeled links from other notes/nodes. Thus each note
is a kind of entity with a local unique ID (LUID). Each link is of a
named type elaborated in a separate meta level workflow.
Everything that is possible at the note level work homoiconically
at the meta level – it is turtles all the way up.
A WikiNizeR Notebook is a collection of notes. They also define
a context or neighborhood of notes/nodes. Notebooks, which have
predefined structure which map to the Google Doc semantic
markup, can also contain reference nodes that define references to
other WikiNizeR Notebooks. All the Notebooks ever created are
added as reference nodes to the system notebook, called My
Notebooks. This notebook gets created at first use, and contains
reference nodes to all the Notebooks which are created at first use.
These include a Notebook dedicated to Getting Started
Information. This document contains references to other
Notebooks. A Notebook that contains references to other
Notebooks, together with appropriate links that carry
computational interpretation between these references, is called a
Research Object or Knowledge Object, which is a Live Graph,
with executable behavioral specifications (a graph of descriptions
of system capabilities which boil down to executable nodes).
When the user “reads” Getting Started she obtains information
about the system by browsing such a graph. The simplest form of
Research Object is a Trail that contains references to nodes and
rules for obtaining a fixed traversal sequence of nodes. These
rules are a simple example of an executable behavioral
specification. This is just like the Trails envisaged by Vannevar
Bush for his MEMEX.
The “My Notebooks” system notebook creates an initial context
for the user. All work that can be accomplished within WikiNizeR
is carried out in the context of some Notebook.
Every note by constructions belongs to one specific Notebook. So
typically all workflows start with picking or creating a Notebook
as the current context.
The “My Notebooks” system notebook creates an initial context
for the user. All work that can be accomplished within WikiNizeR
is carried out in the context of some Notebook. Every note
belongs to one specific Notebook. So typically all workflows start
with picking or creating a Notebook as the context. Within the
context of a Notebook new notes can be added, and then linked
simply by giving a name for the link. At a later stage, in a separate
workflow, the user is given the opportunity to elaborate the
intended meaning of a link and its possible relations to other links
and types.
Keeping track of all the defined links and types there are
additional system Notebooks or contexts for each: My Links, My
Types, and My Properties. System Notebooks keep track of
Contexts. (In principle, every Notebook defines its use case
dependent Context, which is reflected in these system Notebooks.)
Links can be created across Notebooks. Within a given Notebook
notes from another notebook can also be transcluded. When a new
Notebook is created this fact is reflected by a new node being
added to the My Notebooks system notebook. Hence, by
definition, anything you do within WikiNizeR takes place within a
Context. Contexts form their own graph, using the same linking,
transclusion operation that are available for nodes. These graphs
visualize the Knowledge Architecture of the Things and their
relations in a given Context and every user interaction is situated
and carried out within a context.
The most significant implication of all this is that the basic work
pattern is always the same, regardless of the context, and what sort
of things with which we are dealing. Similar pattern of universally
applicable workflows can (in parts) be seen within Freebase. The
Wiki like organization in WikiNizeR works just like this: In a
search you follow links and find the right context (Notebook) for
what you have in mind. Then you explore, and navigate in the
graph. When elaborating something you create a note, and link it to
a node in the graph indicating the nature of the link. The
implications of insisting on finer granularity, and requiring
articulation of how things are linked, are profound. What you end
up with is something which reflects, in the structure of the
notes/nodes you have created, what you have in mind. Together
with the links, and their subsequent meta level elaborations, all the
semantically relevant aspects that make them conceptually
transparent become part of a personal knowledge graph.
Implementation Details: During the beta test period, our webapp
is served from alpha.wikinizer.com. In production use, the users
will “host” the web app from their own drive. When the server
becomes ultra thin, by shifting most of the load to client devices
and Google services, this supplies an architecture, responsible
only for arranging updates to be triggered, handling
authentication, and eventually the e-commerce of
Research/Knowledge Object. The app itself is package
distributed, and purchased as a Research Object itself.
We are using the latest advances in technologies. Namely, Google
drive SDK, Freebase, and other Google Javascript REST APIs,
HTML 5 working in Chrome on Android as well as on the
desktop, plus Chrome extensions, graph databases, the “good
parts” of JavaScript, (as Crockford described in his seminal book:
“Javascript the Good Parts”) all becoming available over the past
year. As the javascript Object Notation JSON created by
Crockford continues its relentless march to eclipse XML, more
and more Linked Data Sources and APIs are becoming available,
hastening the “rise of the API economy”. WikiNizeR is built
around public APIs offered by Google and will integrate more
APIs in the near future. In line with our general approach we are
developing a novel way of delivering an API for WikiNizeR. In
doing so we are aiming to maintain compatibility with the
emerging Linked Data Platform standards. With this, WikiNizeR
is poised to be part of the emerging API economy not just as a
consumer, but a producer of open cloud accessible capabilities.
Innovative Aspects: All our ideas are about 50-70 years old.
Bush MEMEX is 70 next year. Engelbart's published his
framework for bootstrapping systems of HCI augmenting the
human intellect 52 years ago. [Cf. http://www.1962paper.org/]
Ted Nelson’s hypertext with bidirectional typed links,
intertwingularity and transclusion, have been around for nearly as
long. The web scale integration of structure, collaboration,
boostrapping, and raising collective IQ, all however still seem to
be just a pipe dream. In this sense we are realizing the best ideas
‘of old’, with the nascent new technologies described above. In
the wake of the million-fold increase in memory, storage, and
CPU, coupled with the network speed that makes the computer
"hollow out", we can say that the time for these ideas has come.
What years earlier would have required hundreds of men can now
be accomplished by less than a dozen. It becomes possible
because there is no need to build complex infrastructure any more,
we can leverage the power in both our and Google's pocket.
Status Maturity: What we have today is a (minimal viable)
prototype of a system that supports Personal Knowledge Work,
focusing on research, and making sense by weaving a personal
knowledge graph that harvests Google's Knowledge Graph via
Freebase. In this sense WikiNizeR is the simplest possible
personal knowledge augmentation engine that can work in order
to bootstrap further capabilities off the ground.
Link to our submission:
http://alpha.wikinizer.com/LinkedUpChallenge
3. DISTINGUISHING FEATURES
Innovation in Education: Representing the process of the growth
in our personal knowledge, and sharing the emerging conceptual
structures that emerge within problem solving contexts, is a major
task for individual learners, and a critical coaching and tutoring
problem. Knowledge obtained from the web or other resources
needs to be personalized, reorganized, and contextualized, in
order to filter our misconceptions. A dynamic conceptual
organization of our knowledge not only helps us to define
problems, it also helps us to discover their solution.
WikiNizeR is a suitable personal knowledge management
environment for the representation and organization of
“associative complexes” in computer augmented Personal
Learning Architectures. Compared with traditional educational
approaches, it supports a connectivist learning model, and can also
be used in problem and project based conceptions for defining
learning paths, work flows, and supplying concept and resource
networks..
In a formal educational settings, it is able to replace Learning
Design and LAMS 2 type educational Meta Modeling as a tool for
designing structures of meta-level didactic objects, as its default
however, it favours connectivist web scale e-didactics, which
utilize Linked Data. Its mobile learning oriented implementation,
which is based on RESTful programming, is compatible with the
Experience/Tin Can API, and a Learning Record Store. [6]
The contextual organization of information is a basic drive in all
aspects of learning. There is a large body of products which help
learners accomplish knowledge organization tasks within learning
contexts. Many of them are of a visual nature, providing custom
made representation schemes for concepts and related data. Topic
and Concept Mapping, Circle Diagrams, SemNet, GMap,
Conceptual Graphs are all well evaluated, and research results
support the efficacy of visual knowledge acquisition and
organization. In comparison with traditional educational
technologies WikiNizeR makes the learning process more
efficient because of the following features:
- Scalable linking of educational data and resources (Each note and
node is a kind of entity with a global unique ID.)
- Refined concept mapping/graphing technique for finer distinction
of ideas (Everything that is possible at the note level works
homoiconically at the meta level - it is turtles all the way up)
- The granularity of a note and its associations is set by the
requirement that it can be the target of a link as a whole, starting from
other notes/nodes, and can itself be linked to other notes/nodes in
meaningful ways (because every note is implemented as a node).
- Fine-grained Type Hierarchies, admit and define META+META
types. (Each link is of a named type elaborated in a separate meta
level workflow).
- Freely definable link-set for Learning Process modeling (in
consequence of the above-mentioned features).
- The possibility of constructing dependency links between
concepts and skills (Types of dependencies can be defined at the
meta level).
- Articulation of Common Core subjects as personalized
knowledge graphs. (Existing Common Core subjects can be
elaborated in the form of interrelated topic and concept maps and
complemented with Freebase Linked Data and web resources).
- Knowledge discovery, and organization tailored for the human
sciences support the dynamic nature of research in these fields.
(Everything is a node connected through links on the fly, to which
the user gives meaning that captures the domain specific, context
dependent intent to organize things).
- Web capturing and preservation support recall and
comprehension. (Keyword based search function and versioning
of WN notebooks on Google Drive supplement personal
archiving).
- Filtering and visualization of Linked Data based on a Bottom-Up
Strategy (which focuses on finding relevant data sources via
manual data selection on the fly).
- Flexible, content driven semantic organization, contrasted with
the separate development and ‘post-hoc’ correlation of content
organization and ontology building. (At the heart of the
differences between formal as well as other manually created
ontologies is the fact that WikiNizeR works with attributes that
people consider relevant to entities in the context of their learning
process. While it is able to map any ontology into the context
which is exportable in a RDF form, its workflow is built on
“content driven” contextual semantic knowledge organization)
Mental map preservation: Mental map preservation has been a
topic at the forefront of dynamic graph layout. The level of layout
stability can vary between approaches balancing stability,
complexity, and quality. Our approach allows the user to adjust
the positions of nodes according to their knowledge organization
criteria, saving the layout in accordance with an inherent trade-off
between stability, complexity, and space. Our solution relies on
force directed visualization of arbor.js which affords innovative
possibilities. Arbor.js leaves entirely to the user of the library the
specifics of node and edge rendering and associated interactions.
Within rendering we make use of the following main techniques:
- Elements of the same set are identified by colored labels and
links.
- Manual layout adjustment can nudge a force directed graph to
settle into an equilibrium more closely reflecting what the graph is
about to the user and of course helps to minimize overlaps.
- Dragging a node in a force directed graph moves along a cluster
of nodes that are linked to it, giving a visual clue to the user what
is closely related to her interest.
- Similarly the ability to turn links on and off and animate them, is
a feature that supports transparency even possibly showing path
animated node-links.
We adopted the following functions supporting mental preservation:
-The user can determine the location of the coloured nodes or the path
- Edges are coloured and labeled according to their types with a
dynamically generated palette and shown in a legend that allows
controlling of what is shown.
- Anchored layouts
- User-selected multiple foci
- Transitions of bundled edges
- Stepwise transitions, change of focus and depth
- Stepwise animation for navigation based on a spring algorithm
- Stepwise animation moving (parts of) the graph together
- Force-directed layout with virtual forces
-.Simulated annealing with customizable weights for optimization
efficient algorithm and GPU implementation
- More efficient dynamic initial positions of nodes, which can be
saved, reused, and re-adjusted in a force-directed layout with
additional energy factors between time steps
- In situ integration of small visualizations
- Cluster evolution on a timeline for navigating animated node-
link diagrams
- Context sensitive, intelligent, dynamic menu widgets in the form
of a radial popup for applying operations to selected nodes.
- Two different views (Dynamic Graph with cross links and
Hypertree view) for representing associative relations and
“structural” knowledge organization hierarchies.
- The user can navigate in both the Graph view and in the
Hypertree, and edit in the Graph view. When navigating in the
emergent knowledge graph, the student is able to focus his
attention on a concept, and explore the concepts, which are
associated with it. The structural relations of the concept allow
answering to this kind of needs. The structural relations can be
predefined subclass relations, “instance of” relations, or just co-
occurrences representing the interest of the student/user
constructing an association graph of their choice in the form of a
hierarchic structure of a hypertree.
Audience: Knowledge workers, or anybody needing to do web
research. People who "write to think" and wish to break out of the
“walled gardens” of point solutions where they are not customers
but the products that are being sold. Instead we offer WikiNizeR for
those who would prefer to cultivate their own personal knowledge
graphs, for those who favor private Digital Archives where they can
record ideas and all their learning in a form that is reusable. These
potential users look for a “digital habitat” where they can harness
the power of semantic search on the Web with Freebase by
themselves and within their own WikiNizeR world without being
burdened with high levels of cognitive load. Our potential users are
not only Tech or Expert users who understand the semantic web and
other advanced technologies, or have experience in using
ontologies, but Lay-users (who span the categories of novice to
casual Users) including not only students, but also lifelong learners,
teachers, academics, postgraduates, learning designers, educational
professionals, and web researchers, etc.
Usability: We are experimenting with an approach to user
experience design pioneered by Engelbart and his team back in
the Sixties, based on explicit description of system capabilities,
and automatically managing all screen factors. The key is to build
on the core capability of WikiNizeR as an augmentation engine
[4] which is able to provide a first class representation of all the
capabilities of the system, and to provide mechanism which turns
this elaboration into user interface components, and combines
them into activity ‘screens’ which support specific activities.
Work is progressing to gradually replace hand coded and designed
user interaction elements in this design style. Even in its
incomplete state the current “Get Started Graph substitutes
descriptions, serving as user guides and documentation. The
testers have found the graph and the hypertree based navigation
extremely powerful; indeed, they have found that jiggling the
graph into desired shape is itself a thrilling experience.
Performance: We do not operate any servers. All computer
functions run on users’ devices (including smart devices) plus on
Google infrastructure. We integrate semantic collaboration and
delegate all storage, backup, synchronization, access control,
group management tasks using Google Works (previously known
as Google Apps for Business) to Google. Hence the system is as
scalable as Google itself. We therefore have response times of the
kind, which are familiar to people who use Google services.
Freebase also has a fantastic response time that is measurable in
milliseconds, and so even with increasing users, or content data,
scalability is granted. The current alpha setup is running from our
own private server, but is ready to be moved to the Google App
for Business wikinizer.com account.
Data usage and quality: We rely on Freebase, a “melting pot” of
datasets, and its ‘soft’ user developed ontology as our only linked
data source. In our tests we have found Freebase data to be of very
high quality. In many areas it lacks depth, but it does have a wide
breadth. Freebase needs applications like WikiNizer to add depth
to the data that they use. It is in our longer term plan to provide
the means to assist users to give some of their personal knowledge
graphs back to Freebase. As much of the capability of the app is
already based on an explicit intent graphic model, we have a new
and unique opportunity to accumulate information about usage
patterns over time, and use that to suggest customization
opportunities to the user, as well as dynamically drawing their
attention to other, so far undiscovered but relevant capabilities. In
the alpha test of the yet unreleased Android version we were able
to “mine” the users own browsing histories and search histories to
build a dynamic model of their own interests for themselves.
WikiNizeR is already “Tin Can Ready” (is built for a rich user
activity model which contains the Experience Api as a subset).
WikiNizeR Notebooks can be used as a clustered personal triple
store, its model can handle all that Tin Can requires, (JSON
activity stream, statements). WikiNizeR users’ devices could
communicate with external Learning Record Store or as a self
hosted distributed, soon to be peer to peer web based LMS.
Of course, we make use of Google Search and Image search. The
latter is complemented by Awesome Screenshot which also serves
our visualization approach, which applies icons and pictures to
refer to the essence of the nodes.
Mendeley has just launched its own new API for Beta testing. We
are planning to make use of it while our product is in Beta test.
Legal & Privacy: WikiNizeR is set up so that it does not handle user
data but works with data on the user’s Google Drive. There are two
levels of access. As a minimum, WikiNizeR requires storage area to
be set aside for its own use by the app. If the user wishes to extend the
capability of WikiNizeR to integrate keyword search of all of the
user’s content on his drive, this extra permission need to be added.
Freebase attribution is based on Creative Commons that allows
commercial use (CC-BY which is used by roughly 49 datasets). It is
the second most popular license utilized by Linked Datasets, and it
states “You are free to make commercial use of the work.”
The most difficult issue we foresee relates to the fair use of
images. Freebase, for example, no longer allows image uploads to
nodes. We consider our use of unattributed images to be "fair use"
while we are in restricted stealth mode pre launch beta.
All images in use will be attributes purchased or sought permission
for before public release. We are still seeking legal advice on the
particular type of licence that we would wish to use in the future since
some confusion concerning CC-BY has led to lawsuits in the past
against the companies using the data covered under CC-BY. For more
details, please check the FAQ on the Creative Commons website.
4. DISCUSSION
11 years ago, with the support of a DTI Smart award, we started
working on a feasibility study of personalized mobile computing to
support personal knowledge work. By Personal Knowledge work we
meant all the personal information management tasks that go beyond
appointment diaries, mails, social updates, and news feeds. In a
process of Seek, Make Sense, and Share, the goal was to find a way of
bootstrapping into a world of user interaction where it is possible to
go beyond the ‘sweet spot’ of the palm philosophy. On a limited
screen size mobile device according to the Zen of Palm there is a
point where further addition of features results in marked deterioration
of User Experience. The ‘sweet spot’ is the point just before that
would happen. Current UI Design guidelines hark back to this well
known figure: Fig. 1.4 (page 13).
For limited screen size mobile devices dictate to limit the feature
set to avoid degrading the user experience. The fundamental
tenets of the Palm philosophy, of instant availability of
capabilities on mobile devices still holds, but the assumption that
people would not want to spend too much time using their mobile
device at a time has clearly lost its validity.
Although it was quite quaint in the early 2000s to use a PDA, it
was clear to us that it should be possible for knowledge work to
take place on the go, bootstraping a co-evolving system which has
explicit intent graphic capabilities of the system. As the user
explores the app’s capability repertoire, using a limited set of
intuitive user interaction patterns, they can use any capability as it
is discovered, giving a pervasive user experience which requires
no new learning in the way it could be driven. Given that,
personalization becomes possible, not just the look, or
arrangement of controls, but the workflows available, enabling
them to be reached easily at any point. With this there is no longer
any need to limit the feature set just because the device screen is
limited. In doing so we have unknowingly been engaged in
building systems akin to what Bush speculated about, and which
Engelbart built, on a “milli iPhone”, and which Ted Nelson
worked towards. We did not set out to build the MEMEX etc, like
twin.gl or Artficial Memory are doing now, but built something
very much like it. At the same time we have been travelling pretty
much the same journey as Engelbart. Our goal is more limited, we
think that the road which leads to augmenting collaborative
human intellect requires first that we empower the individual
knowledge worker, and later let them collaborate. Thankfully to
date it is possible to address the individual knowledge workers
directly without the need to get funding from corporations.
Success of Brain, and many other mind mapping applications,
Evernote, Pocket, ReadItLater, Pinterest, etc and all the other
point solutions for personal knowledge management, tell us that
there is a market in this gap. If only we could avoid being trapped
in the “walled gardens” of point solutions which may deliver
“sweet spots”, but do not cater for our deep needs for reuse,
repurposing, and reproducibility. We believe that we are building
that proverbial “better mousetrap”, and will soon enough see if
people will flock to our doors. In June a new Graph Database
called Cayley, developed by a Googler, Barak Michener, has been
launched as a database that would enable developers to set up
their own Knowledge Graphs. We are investigating the possibility
for basing our future support for collaborative knowledge work on
this technology.
On the 28th of August Mendeley has launched its new Beta API.
We are excited at the prospect of integrating WikiNizeR with
Mendeley with its comparable capabilities developed for
Freebase.
WikiNizer Research is also going to adopt the Wolfram
Computable Document Format as its raw knowledge exchange
format, which will enable a clearing house of collaborative
conceptualizations to be launched more quickly than we originally
anticipated. “Launched by the Wolfram Group, the CDF standard
is a computation-powered knowledge container — as everyday as
a document, but as interactive as an app. Adopting CDF gives
ideas a broad communication pipeline accelerating research,
education, technical development, and progress.”
(http://www.wolfram.com/cdf/) CDF has the potential to bring
about the technological revolution which is a precondition of a
much needed cultural change in the way knowledge is produced,
collaborated upon, and disseminated.
5. CONCLUSIONS
In our earlier papers we described our roadmap as “From Personal
to Collaborative Concept Organization” [3] and we situated our
‘WikiNizer™ technology’ within Doug Engelbart’s Vision. We
spelled out how the first module of a Personal Knowledge
Augmentation Engine, WikiNizer™ Research augments personal
knowledge building by implementing WikiNizer™ Kernel and a
Visual Semantic Wiki-like environment. [1] This environment
enables us to organize personal Knowledge Architectures into
visual Knowledge Graphs. We compared the features of our
personal knowledge management solution with other Semantic
Wikis, [1, p. 9] and suggested that flexible content dependent
graph structures on all levels, enable us to represent and model
both symbolic (lexical and higher) and sub-lingual “cognitive
structures” and processes which belong to the domain of personal
knowledge. We conclude that personal knowledge management is
enhanced by externalizing our cognitive models and mental maps,
and that bootstrappable visual tools are able to mobilize cognitive
structures which enhance the efficacy of our problem solving.
WikiNizer™ renders this assertion empirically testable, and it
therefore can be considered a benchmark of an experimental
epistemology. WikiNizeR built Knowledge Graphs give us the
underpinnings of a more comprehensive e-didactic approach to
conceptualization and contextualization. In this sense, our
technology gives us an augmented Exploratory Epistemology
which enhances both personal knowledge building and networked
learning.
6. REFERENCES
[1] Benedek, A.G., Goodman, C.P., and Lajos, G. Augmenting
Knowing with WikiNizer™ Research. International
Conference on Innovation, Documentation and Teaching
Technologies, 8-9 May, 2014, Valencia,Spain.
https://drive.google.com/file/d/0B5Q2MasRnFZneVNWNTR4MndpaEE/edit?usp=sharing
[2] Benedek, A.G., and Lajos, G. Conceptualization and Visual
Knowledge Organization: A Survey of Ontology-based
Solutions. A. Benedek, G. Lajos - INTED2014 Proceedings,
pages: 4609-4619.
https://drive.google.com/file/d/0B5Q2MasRnFZnek5tS3BaSHBQWGc/edit?usp=sharing
[3] Benedek, A. and Lajos, G. From Personal to Collaborative
Concept Organization: Conceptipedia as a Visual Tool for
Educational Initiatives In Gómez Chova, L., et al. (Eds.)
INTED2013 Proceedings, Valencia, Spain: IATED, 2013,
pp. 1303-1316.
https://drive.google.com/file/d/0B5Q2MasRnFZnOGZ4QXRLSVZFQlk/edit?usp=sharing
[4] Lajos, G. and Benedek, A. “Building WikiNizer, a Personal
Knowledge Augmentation Engine for Next-Gen Semantic
Collaboration Platforms.” INTED2013 Proceedings,
Valencia, Spain: IATED, 2013, pp.1333-1343.
https://drive.google.com/file/d/0B5Q2MasRnFZnMThpX1RoTW1BNVU/edit?usp=sharing
[5] Benedek, A. and Lajos, G. Building Augmented Knowledge
Architectures: Requirements for Collaboration Platforms of
NextGen Concept Organization Tools. In A. López Martínez,
L. Gómez Chova, I. Candel Torres (Eds.) Proceedings of
ICERI2012 Conference 19th-21st November 2012, Madrid,
Spain, IATED, pp.1492-1506. https://docs.google.com/document/d/1
ya06AJMEwubpdSZWh1L9mzeS4RdCjI-wo7lLwYVoGCY/edit?usp=sharing
[6] Benedek, A.: Learning Design versus Learning Experience
Design: Is the Experience Api Making the Difference? in: L.
Gómez Chova, A. López Martínez, I. Candel Torres (eds.)
EDULEARN13 Proceedings, International Association of
Technology, Education and Development, IATED, 2013.
https://drive.google.com/file/d/0B5Q2MasRnFZnVURIc3dUY3BLVUE/edit?usp=sharing
[7] Fischer, G. (2009) End-User Development and Meta-design:
Foundations for Cultures of Participation. In: IS-EUD '09
Proceedings of the 2nd International Symposium on End-User
Development, Springer-Verlag Berlin, Heidelberg. pp 3 – 14.
