Tristan Teunissen

movements outside the lab

Tweets @web3.0 conference, Santa Clara

leave a comment

“The Semantic Web is great, new way of thinking about the web, ai people meet publishers”

Scott Prevost (Bing/Microsoft), Computational Linguist at stage. Founder Powerset. The dimensions of search

What is semantic search? More relevant results, but what is relevant? All searchengines today are already in someway semantic

Relevance: best result at top, completeness, freshness. Speed: page render. Ease of use: simple interface

Semantic Impact. Relevance: ranking based on meaning and concepts not keywords: direct answers

Speed: reduce time to task completion, fewer clicks. Ease of Use: intuitive queries, information aggregation&classification,simplified tasks

Query understanding: disambiguation, refinement. User sessions query1..queryn

Document &content understanding: entities,relations&concepts in text. Structured and semi structured data. Better matching of queries

User experience: reduce user investment

Queries: navigational, informational, transactional. Scope: general vs specific. Context of query. Syntax of query.

Disambiguate query,what to use? Problem: most queries are underspecified

How can semantics help? Term expansion (synonyms,acronyms), Powerset did a lot of work. Flexibel syntax: ABC <-> ACB a NLP task.

Entity detection: nesting problems [[[Carnegie] [Mellon] [University] [Robotic Club]

Document&content understanding:structured content(API’s,RDF,OWL,DBs)weather, sports, stocks, product info, freebase

semi-structured content: map site/page to ontologies, semantic tuples scraped from pages

text: keywords and proximity,entities&semantic relations. The impact: better recall and ranking of results, better organization of results

better presentation of results (captions), better user actions leading to fewer&better clicks

smarter text selections for captions, [rebate of..] vs [rebate of $600] in caption. Captions with word variation. query:sarah palin cap:she

Smart summarization without changing the meaning

whatever semantic search is, it’s already here, it won’t be a big revolution from a new startup, but there will be game changers

you need critical mass, but the ecosystem is growing, more&more publishers are helping

but our focus has to be on making systems which can automatically semantify open content

seo is all about to find the keywords that match your page,semantic tech will help out

what will semantic analytics do?Not sure how that is going to evolve,clicks only are a weak signal,but in combination with content meaning

which vars for mobile web: location, but also calling patterns as social network?

Signal from their social graph, but it’s really hard for realtime search what is really important,who is important?What is the real signal?

Freshness is important, but we’re creating hyper freshness through twitter at the moment,a lot of noise in this signal.

The role of telco’s? Voice services

Dr. Mark Greaves (Vulcan inc): The Evolving Semantic Web: From Military Technology to Venture Capital

web 2.0: the read write web, strange name, it isnt a software release. Web 3.0 “the world wide database”

what is the biggest database in the world? Social Security / Walmart database on steroids. Also problem on steroids, updating

or is it like the web? Always changing, no single view, always evolving, no central control.It’s geeky & transformative

it’s democratic,crowd based,scalable knowledge engineering. It’s the hottest area of web architecture right now

It’s the largest but also the messiest formal knowledge base on earth.

origins of the sw; symbolic logic,knowledge representations systems in AI & parallel library science was going on.Web created infrastructure

Google is making me the smartest guy in the world because I can query, but computers can’t do that

The first succes from the sw was catching bad guys: 9-11

Where is SW in 2010? academics are working on the rule parts: proof&trust

Noise is a big problem for the semantic web, there is too many information, how to find out what the best linking is? RDF/OWL is too simple

relevant == semantics,semantics are the key technology for information retrieval. Enhancing snippet presentation has business opportunities

click through rates from better snippets are 15% creating bigger revenue: business opportunity?

Best Buy:RDF markup with RDFa&goodrelations ontology: organic search engine traffic +30%.Not perfectly controlled experiment,but suggestive

they have a higher search engine placement,probably because they are more informative for searchengines.So SEO agencies have to semantify

Collect data,clean it, fuse it with quality control.Sounds boring, but Bloomberg does it and is a multi billion company.

The linked datasets are growing exponential, exciting

Network effects are starting. Link data now! Be part of the revolution

The cost of publishing semantic data is going to zero, just like normal publishing on the web

Winners will be in the mobile space again, building cleaner interfaces and can contextualize their users

Online Publishing panel: Ben Ilfeld (Sacramento Press), Mike Lee (Thoora), Mark Luckie (Journalist/Blogger), Bostjan Spetic (Zemanta)

Thoora indexes realtime news and clusters it by story, which story has the greatest interest at X moment, comments on blogs as measure

Thoora treats the web as one big database,looks at how data is linked at realtime and what the story is about,using a lot of structured data

Using semantic tech publishers get more insight in what people are talking and reading about

Strangly enough there arent many end user solutions for disambiguation

Volume is not the problem for processing, but the noisy input channel is the main challenge

Panel: Semantic Advertising

Don’t talk about semantic technology, you have to show it, and people will understand it immediately

Web of Data: Semantic Web in Marketing. SCOTT BRINKER(CTO Ion Interactive) and KRISTA THOMAS (The Calais Initiative, Thomson Reuters)

How will linked data effect the 4p’s of marketing?Data is the fuel of 21st century,but finding data is hard work,we horde it in silos

But new generation of companies is making their data open, and a wave of linked data in on the way. Marketeers love data

SEO + Data Objects = SEO ++ enrich your data for better snippets, clickrate +15%

bringing structure to unstructured txt can help: streamline SEO,improve reader engagement,unique content experience,reader analytics ..

improve ad placement,using linked data as a transport layer. Problem:structuring content is costly&time consuming.So extraction engines help

Web 3.0 Day2, Keynote: Semantic Web and the Customer Experience, Tom Gruber (Siri.com)

Big Think, Small Screen. Computer is smart when,understands your language,sense of environment,solve everyday problem,be at your service

The Cloud,the pipe,the interface,the ecosystem.Web 3.0 requires more computation (machinemuscle)=expensive,can the cloud make this possible?

The cloud is the illusion of infinite scalability and omniscience.The cloud is the new datacenter(photos,music,text); we live in the cloud

pipe:3g enables mobile inet.The second wave.What makes a smartphone smart?They’ve senses:touch,hearing,sight,proprioception,taste;brain?

Foodchain of web2.0 started with search,what will feed the ecosystem of web3.0?It’s not data but the ‘Gigantic Join’;services&api’s

Over 1600 api’s at the moment, growing at accelerated rate: non-lineair. Network effect is happening

NLP task based on Chomsky deep structure. Impressive demo. Combining services in the cloud, mobile senses as input for search

RT @kristathomas : Siri Demo Video http://bit.ly/bFMxem

thriving big think ecology:connect&combine,dont accumulate.Address human tasks.Apply intelligence at the interface contextaware&personalized

Written by Tristan

January 27th, 2010 at 6:00 pm

Leave a Reply