WordVis was developed at
NTNU university in
Norway
by
Steven Vercruysse
[1,2]
during a four month programming marathon in the summer of 2010.
WordVis lets you find words in the English dictionary (
WordNet), see synonyms grouped by meaning,
and browse them in a fluid, interactive web of words and meanings.
With WordVis you can explore the English language and perfect your language skills.
The visualiser
WordVis is a technology test and demonstration of our new interactive visualiser for large networks, "GraphVis",
which is purely written in
JavaScript
and the new emerging
HTML5 standard. This
visualiser relates to other projects like
TouchGraph,
ThinkMap,
VisuWords,
Visual Browser, etc.
GraphVis' novelty lies in two significant new aspects. First, it uses cutting-edge web technology:
it is completely written in
the web scripting language
JavaScript,
and based on the new "
canvas" HTML-element.
This makes it possible to design GraphVis as an interactive component, fully integrated inside a web page.
It avoids isolation by third-party plugins (like Java or Flash), and enables direct interaction with other webpage-components.
Second, GraphVis also surpasses existing visualisers in user-friendliness and extensive functionality:
its array of graph editing tools empowers a significantly more detailed customisation for the exploration of information networks.
The layout algorithm
GraphVis uses a
real-time, force-based layout algorithm. Nodes of the (sub)graph currently in the visualiser are modeled as objects that per animation frame undergo a capped sum of up to four different physical forces:
1. Nodes repel each other 'electrically', whereby a maximum interaction-distance limit supports computational efficiency. A node is modeled either as a charged point (think of an electron), a line-segment, or a rectangle. GraphVis automatically treats nodes representing a single line of text as linear charges, and multiline text-nodes as rectangular charges. A line-segment charge interacts like a point-charge placed on the segment where it is closest to the interactor. For example, for a point charge located left of a segment charge, the horizontal component of the distance is calculated as to the segment's left side. This prevents that text labels overlap: they look and feel like magnetic rods that repel each other over their whole length.
2. Edges are modeled as springs that attract their two nodes the further they are away than the spring's preferred length, and repel when closer. (GraphVis also supports 'edges' and spring-like connections between multiple nodes, a feature not used in WordVis).
3. 'Hidden springs' act like normal springs, but are not associated with actual edges. They can be temporarily placed around a node that 'has focus' (or 'anchored'), and invisibly connect this node with indirect neighbouring nodes, limited to some fixed path-length. In WordVis for instance this path-length is simply 2, and so they link a focused node with nodes not directly connected, but reachable by following two connected edges). Their initial spring length is proportional to the average sum of lengths of 'visible' springs that they span. Like this they arrange nodes more clearly into 'levels' (concentric rings) around a focused node.
4. Any edge can be given any preferred orientation, enforced by an 'orienting force'. This can e.g. cause 'is-a' relationships to become organised as a top-down hierarchy. An edge's orienting force is implemented efficiently, by connecting one of its two nodes via a spring-force to a precalculated target offset (the ideal rotated position) relative to the edge's moving center, and applying a mirrored force to the other node.
The visualiser in perspective
This visualiser will be used to explore biological networks (e.g. gene/protein/trait/etc. webs).
We applied it on a dictionary first, as a test-case to make the software more mature; and
hereafter GraphVis will become one module in a larger project in
Semantic Systems Biology.
About Semantic Systems Biology:
Biology is a science that generates lots of diverse of information, and
Systems Biology aims to combine information from diverse subfields in order to see the larger picture,
i.e. to get a broader understanding of biological systems.
This may e.g. lead to insights about how intricate molecular interaction systems give rise to diseases etc.
In order to deal with the diversity, complexity and extensiveness of biological information,
Semantic Systems Biology
aims to capture this knowledge into a
meaningful digital format, and apply automated reasoning and inference on large-scale integrated data.
This is a considerable challenge, one that is shared with an area of
Computer Science.
Currently, most biological information is scattered over millions of publications in natural language, a format not understood by computers.
Therefore
we are developing a web platform
that enables biologist to gather this information, and to make its meaning clear for computers.
As biology often deals with very complex and context-dependent facts, we recently designed a new and universal
text curation method
to semantically capture information of
any desired complexity. This will appear as our a next web-app.
Wordvis' visualiser will be part of this future web-application:
it will present the reward (the 'carrot') to stimulate a community of biologists to semantically clarify their research findings,
in order to view them through a WordVis-like visualiser.
We have experienced before that the ability to visualise, customise, and analyse the really complex biological network information
that resides in literature, is a powerful motivator for such a crowd-sourcing initiative.
Update: in 2011 we created a second application of GraphVis:
OLSVis.
The web-app enables WordVis-like exploration for several specialised biological dictionaries.
The dictionary
WordVis is also based on the
WordNet database, a publicly available lexicon of the English language
that groups words (synonyms) around particular meanings
[Miller, George A. "WordNet - About Us." WordNet. Princeton University. 2009. <http://wordnet.princeton.edu>].
We used the
MySQL version provided by the
WordNet SQL builder project.
About 15% of development time went into connecting
our new visualiser to WordNet and wrapping all into the actual web pages.
Acknowledgements
S.V. is thankful to :
- NTNU university for the financial support in 2008-2011 via a postdoc grant,
and the great environment.
- Martin Kuiper, group leader, for providing the fertile ground and scientific
autonomy that is required to let novel ideas grow and mature.
- Scientists of VIB/PSB/UGent
and NTNU
who were generous with feedback and feature requests on an earlier, small prototype of the visualiser (then still in Java).
- Developers of
WordNet,
WordNet SQL DB,
MySQL, PHP, WHATWG,
Google, Firefox, Firebug, etc.
- You, for spreading WordVis further in the world!
Contact
Please send all your constructive comments, feature suggestions, bug reports, etc. to this address:
.