semantic microblogging

Download Semantic Microblogging

Post on 16-Apr-2017




2 download

Embed Size (px)


Semantic Microblogging

Benjamin /

Cementic? Symantec?

Semantics = Meaning

A web where computers can (better) understand what a page is about ...

The Semantic Web

and where they can then do more without our help. Not necessarily complex stuff.

Add this event to my calendar!

Generate an album from (only) the photos in this Activity Stream!

List all #mbc09 Twitter and users who share a friend with me!

What are the most popular links about microblogging on Twitter?

Auto-block all followers with SEO or Social Media Expert in their profile!

Evolution, not Revolution!

(A semantic web could enable revolutionary new apps, though)

A little semantics ...





We're slowly increasing the content granularity on the web






moving from
documents and links

(arbitrary) things and their relations.

The How is a huge puzzle, though

Creation (NLP, annotations, wrappers, ...)

Publication (embedded, separate, API, formats, ...)

Discovery (centralized, decentralized, push, pull, ...)

Access / Privacy

Integration / Repurposing

Bringing microblogging into the SemWeb play (and the other way round).

Creating content for the semantic web is often cumbersome.

Writing microposts is easy and fun.

Microblogs mash short-term, social chatter with long-term-useful information.

Microblogging improves and accelerates product development, PR, and CRM
(but lacks reporting/controlling).

Microblog streams are the biggest productivity killer since IM and email notifications.

SemWeb technology enables self-describing content, querying, custom filtering, and automation.

The semantic web still doesn't have easy input UIs for everyday use.

Microblogging clients are becoming ubiquitous.

The semantic web uses small units to encode information.

Subject > Predicate > Object
Thing > Key > Value
ID > Verb > Object

MBC09 > location > Hamburg

Microbloggers use short posts to encode information.

Microblogs are flexible

Life / Work logging


Note taking

Announcements / CRM

Bookmarking & Tagging

Interest tracking

Microblogs are not fully Enterprise 2.0-ready yet

No long-term archives or backups

No granular privacy settings

No project-oriented interfaces

No DB-like filtering or querying

No integration of external data

A solution

Let's use microblogs to create structured content on the fly.

Get rich filtering of information streams in return.

A semantic microblog

simplifies the creation of machine-readable information

knows what posts are about

enables task-specific interaction methods

helps with repetitive processes

The structure-to-opacity ratio in Posts is already very high

@joe, the new #release is out: alec on 2009-01-24 ------------------------------------------------------------------------------------------------------------------------- RT @sue: I can #offer a #ride from #Berlin to the #mbc09 in #Hamburgby alec on 2009-01-24author, @recipient, @persons, #hashtags, date, content, URLs(even without hashtags!)

What if we could do:

@joe, the new #release is out: (#version=RC2 #fix=bug27 #_2h #_billable)by alec on 2009-01-24 ------------------------------------------------------------------------------------------------------------------------- Attended great #talk by #seesmic's @kaotisch (#topic=xmpp #event=mbc09 ****)by alec on 2009-01-24------------------------------------------------------------------------------------------------------------------------- RT @sue: I can #offer a #ride to the #mbc09 #from=Berlin #to=Hamburgby alec on 2009-01-23------------------------------------------------------------------------------------------------------------------------- Arrived at #mbc09 (#location=Hamburg)by alec on 2009-01-23

something magic happens...

and my microblog could then

tell me which countries Ive been to this year

filter posts by tag

filter by machine tag key and/or value

generate software release notes

generate project reports (or even invoices)

recommend links or experts for some topic

Provide a query API

The challenge

Finding the sweet spot between simplicity and added value

Syntax patterns should evolve slowly, and in a bottom-up way

Re-use existing patterns (#tag, @person, to:group, ...)

UIs will need slight improvements (machine tag suggestions, filters, ...)

Prototypes, Demos, Experiments


maintained by DERI Galway

source available

tag disambiguation via RDF#geo:Paris, #dbpedia:France



HD Streams

fun experiment

feed subscriptions based on seed tagstwitter,, ma.gnolia, delicious

(hash)tag extraction

entity extraction via NLP APIs (OpenCalais, Zemanta)

entity expansion via dbpedia, freebase, crunchbase

faceted entity browsing (just a crappy proof of concept)


HD Streams


semantic microblogging client with local storage

twitter / integration

structure detection / extractionhash tags, machine tags (#foo=bar), private tags (#_foo)

Re-tweets (leading RT), private posts (leading PP)

@recipient, @person

Ratings (e.g. ****)

URLs (expanded)

simple dashboard for custom lists and filters


source going to be made available at and through a article


Smesher (SPARQL API)

semantic microblogging mbc09 benjamin nowack