Semblogging - a guide to the HP Labs semantic blogging
demonstrator
Here are some simple notes for looking at the HP Labs
semantic blogging demonstrator The blog is intended to
show the use of semantic web technologies augmenting the
blogging paradigm, and applied to the domain of bibliography
management. For further details about the requirements for
this demonstrator, see the
requirements specification. We believe that the use of
semantic metadata can allow a blog to be used in new and
powerful ways.
We have chosen bibliography management because
it shows how we can use semantic web technologies to push
blogging from a communal diary browsing experience to a
rich information sharing activity
Specifically, we have divided the functionality into
semantic view,
semantic navigation and
semantic query. We have prepared a short overview of
our
vision
and
design.
This script is separated into the following sections:
Have a brief look at the blog.
It looks fairly standard. (if you don't know what a blog
'should' look like, check any of the other blogs mentioned on
the 'blogrolling' list.) Note however that all the entries
are about scientific papers. This is because we want to show
semantic blogging in a bibliographic management context.
Now, in order to enable the new functionality, we had to
provide a metadata backing for the blog entries. To check out
a blog entry's metadata we can use the rather wonderful
BrownSauce
viewer, developed by Damian
Steer while working at HP Labs. The 'BrownSauce' link
(example)will
call out to this application and allow a further view of your
blog entry's metadata. It's probably worth explaining the
metadata. The majority of the metadata is fairly standard for
blogs. For example, you will see 'creator', 'title',
'description' and so on. However because this is a
bibliographic blog, each blog entry is (or may be) about some
bibliographic item. That is what the 'contains' property
points to.
A blog entry contains a bibliographic
item
Note that you can click on the bibliographic item (in
BrownSauce) and get some more metadata about that paper
(
example).
You may wonder how we get that metadata into the blog in the
first place. That is covered below.
It is also possible to view the metadata in 'raw' form.
Check out the 'XML format' (example)
and 'N3 format' (example)
links at the bottom of each blog entry. These options display
the RDF behind each entry (note it is the same metadata in
each case, just the serialization format has changed). If you
need help deciphering the output try the RDF primer or the
N3
primer. Whatever option is chosen (XML, N3 or BrownSauce)
this actually does is to retrieve the entry's metadata from a
central RDF store which contains the metadata for the whole
blog. At the bottom of the page, there is the option to view
metadata for an entire category of blog entries, either
separately or aggregated together. [there is a fourth option,
the record card format (example),
but I shall come to that later].
One of the nice things about blogs is that you can export
your metadata. Typically we will use an
RSS feed for this, the RDF version of which is
RSS1.0. If
you click on the RDF Button 
at the bottom of a blog
page, you will get an RDF output (example).
This is meant to be machine readable, so it won't look very
pretty in a browser (click on 'View Source' in your preferred
browser to see the raw RDF, in XML format). However, when
given to an aggregator (there are some examples here) the
metadata can be used in various ways (typically, it is
currently used to display a summary of your site, or to
enable keyword based searching over a collection of blogs).
My RDF feed includes the bibliographic metadata mentioned
above, so a semantically enabled aggregator can provide
additional services. Some of these are described below.
RSS is generated and enriched using the RDF
store
How do we create new blog entries? A click on the 'Add New
Entry' link on the right hand menu leads you to our
SemBlogIT! form. This form was inspired by MovableType's
bookmarklets. The idea here is to say, while
browsing a site 'hey, this looks interesting, let's blog
it!'. So, in one click, you get an entry form populated with
a title, link to the page of interest, and any text you have
highlighted. You then add any other details you want, choose
a category and post the entry. The posting process is
password controlled. We have preserved all of this
functionality but have added two interesting semantic
capabilities.
- A way to attach items to blog entries. So, if you
wanted to blog about a bibliographic item (say, a paper),
you can enter metadata about that paper. Click on the 'add
bibliographic item' button to open a popup. Here you can
enter bibliographic fields such as author and title in the
supplied dialog boxes. Alternatively, if you have RDF
metadata to hand, you can paste it into the text area. You
can generate such metadata by using web utilities such as
Michel
Klein's excellent BibTex to RDF
converter. The blog entry will be marked as
containing the bibliographic item. Incidentally,
the popup dialog is itself driven by a
simple RDF configuration file. This is a common theme
of the demonstrator - RDF powered user interfaces -
providing another advantage of semantically enabling a
blog.
- The Category Chooser. Inspired by Matt Mower and
Paolo Valdamerin's
ENT
client, this utility is a form of assisted metadata
creation. Once you have added the text for your blog entry
and any required details you press the Next button (go on,
try it!) and you are presented with a list of suggested
categories for your blog entry. Under the hood, what this
does is to consult an ontology,
represented in
TIF format. This ontology essentially
represents a hierarchy of concepts, each with a set of
indicator terms (think of keywords) one of which is the
preferred term for that concept, The category chooser
client now goes through the text for your blog entry,
stemming words and matching them with the indicator terms.
If a term is matched, then that concept is suggested to
you. This is a form of ontology driven assisted metadata
creation.
Metadata is added to the blog using an
ontology-driven mechanism to provide user
assistance
Of course, you will not actually be able to post your
entry because you do not know the username/password for this
blog. But hopefully this is enough for you to see how
SemBlogIT! works.
Yes, we can. Click on the 'Edit Metadata' link below any
blog entry to be taken to a form (example:
I call this the semedit dialog) where you can change
metadata values, delete selected properties and add new
(possibly freeform) RDF to the blog. Again, you will not be
able to actually change any values, because this is password
control but the form should give you a flavour of the
possibilities. Interestingly, the semedit display is itself
powered by RDF. That is, you can change the look and feel of
the semedit experience by using a simple, RDF encoded
configuration file, which allows different people to have
their own preferred semedit dialog. (A more sophisticated
configuration is employed for the semantic view component,
described below).
Metadata can be edited in a schema driven
way
As a parenthetical comment, semedit is designed for
metadata access. It will not change the blog entry itself
(that is, what gets displayed to the user on the main blog).
Blog entry editing is a more conventional piece of
functionality, which in blojsom is supplied by the bledit
plugin.
This completes our metadata pipeline. We have seen how to
put metadata in and out of our blog, and to inspect or change
the metadata for any given entry. The next task is to show
how we can use this metadata to enhance our blogging
experience.
The first functionality we will examine is semantic view.
Given that we have bibliographic items in the blog, we would
like to be able to view these in a useful way. Firstly, for
any entry, click on "Record Card" underneath the blog entry
(example).
This enables a record card like view of the metadata behind
the entry. Hover over a property to discover its full
(namespaced) name. Also note that the bibliographic items are
described in a sort of indented record card. This enables you
to distinguish between the blog entry and the bibliographic
items that the entry is about. You can also view a category
of entries using record card view (click on the category
links at the end of the page -
example).
An alternative way of viewing blog entries is a summary
table (example).
You can view this by selecting the 'Semantic View - table'
option at the top of the display (next to the current
category). This will show you a list of entries contained in
this particular category, with summary information (creator,
date, concept and title) for each. You may view an individual
entry by clicking on its entry number (1st column in the
table).
Like semedit, semantic view is metadata driven. An example
of a semview preferences file (in RDF) can be found here.
Importantly, you can have many semantic views for your blog -
I have chosen two (record and table) but you could have
different views for each user. The required view is selected
at runtime through a URL parameter.
Metadata can be viewed in a schema driven,
personalised way
If you click on the semantic
navigation link on the demonstrator
blog you will find the screen transformed to a 2 panel
display. The left pane is the semantic navigation pane. There
are actually 2 forms of navigation demonstrated here:
- Tree Browsing. In the top of the
semantic navigation pane, you will see a tree
representation of the hierarchical categories. Click on
various nodes in this tree to find the blog entries filed
at this node. Try it! Please note that, since entries can
be filed at non leaf nodes, an entry filed at a child node
will not be displayed when the parent node is
selected.
- Facet browsing. The lower part of the
semantic navigation pane allows you to select entries by
author, title, description or concept. The exact format of
this display is driven by a schema.
Clicking on the navigate button displays a list of entries
that satisfy all the fields. So you might choose, for
example, entries written by stecay on the concept of
semantic-web.
In both cases, the relevant entries are retrieved from the
blog database on the basis of their metadata.
Metadata can be used to drive and to customise
the user navigation experience
If you click on the
Query By Entry on the demonstrator
blog you will find the screen transformed to a 2 panel
display. The left pane is the semantic query pane. This
contains a set of fields, corresponding to blog entry
metadata, The exact form of the display is schema
driven, but in my case I have allowed a query by creator,
subject, title, and full content. Clicking on the search
button displays a list of entries that satisfy all the
fields. So you might choose, for example, entries on the
subject of learning which have the word "introduction" in
their title. In any case, the relevant entries are retrieved
from the blog database on the basis of their metadata. The
results are piped into semantic
view - hence, you can choose your preferred format (here,
record card or table form are the alternatives).
Even better than this though, the semantic blog allows a
search on the contained items. So, say you wanted to find
blog entries about a certain paper (or about papers on a
certain topic, or by a certain person...). You can use the
Query By Item function (again, on the demonstrator
blog menu) to do this. So you might choose, for example,
to query for items with an author of "Lassila" or a title
including "semantic". The query works exactly the same as
before, except that the entries returned are those
about the items of interest.
Semantic metadata can be used to provide novel
and useful query interfaces
The semantic blogging demonstrator shows how we can use
semantic metadata to enhance the blogging experience. It also
shows how we can apply the new technology to a concrete
application domain, bibliography management. However, the
demonstrator is not meant to be the final word on semantic
blogging, it is more a proof of concept. There are many
extensions that would be possible over the existing
framework. For example:
- Streamlining the import process, making the SemBlogIT
activity truly 'one click to blog'.
- Providing extra navigation and query modalities - for
example, 'more like this'.
- Aggregating bibliographic blogs to show the potential
of the semantic web to integrate knowledge from a
decentralized community
- Providing an ontology linking facility to allow
searching over information marked up with different
vocabularies.
The demonstrator shows the feasibility of this
approach - the applications are virtually
unlimited
Steve
Cayzer 2003-09-30