Mediaglobe - the Digital Archive

Der neue Teaser-Trailer für unser Projekt 'Mediaglobe - The Digital Archive' ist online. Weitere Infos über das semantische Videosuchmaschinenprojekt unter der Mediaglobe Projekt-Webseite oder über unsere Semantic Technologies Webseite am HPI.

XInnovations 2008, Berlin, Day 02 - Sept. 23, 2008

Today, I'm going to visit the 'Corporate Semantic Web' workshop at XInnovations 2008 in Berlin. At least it seams that semantic technology has reached industry and corporations. "There is no market for semantic technology", as Christoph Tempich from Detecom Int. quotes a former oracle statement in his talk "Analytics drive the Corporate Semantic Web". Therefore, you just have to provide another label, which is 'Enterprise Information Management' with semantic web technology as underlying technology.

During the coffee break I followed a discussion on the planning of an 'Asocial Semantic Web Workshop' for the next WWW conference or ESWC conference. The goal o fthe workshop should be to show in which way the semantic web is vulnerble by SPAM or other offensive techniques, as e.g. denial of service by providing a deadly RDF-sequence that causes temporary data to grow exponentially.....sounds rather intriguing. I'm looking forward wo contribute ;-)

The 2nd session this morning starts with a presentation from Markus Luczak-Rösch from FU Berlin on 'Corporate Ontology Engineering'. Next, Holger Seubert from IBM is presenting 'Enriched Content Browsing', i.e. during page load in the traditional web, the web page is enriched with additional content. The text of the web page is analysed and terms of interest (info spots) are selected and linked with additional contextual information (from the web, from corporate data bases, etc. in another frame of the same window) without leaving the current context.

We had lunch in a small cafe underneath the nearby public railway with russian dishes (Cafe Chagall, Georgenstr. 4, 10117 Berlin). The pelmeni was really delicious!

The afternoon session starts with Thomas Hoppe from Ontonym with a presentation on 'Corporate Semantic Web'. According to his interpretation, the general 'Semantic Web' concept of Tim Berners Lee cannot be simply transported into the corporation as it is. Inside the corporation, it's a different world compared to the outside. All users are employees, vocabulary is (most times) strictly controlled, there are strict access restrictions, services have to be integrated in portals and corporations have to support corporate processes. The session continues with a presentation by Ralf Heese from FU Berlin on 'Corporate Semantic Collaboration'. He introduces the simple text-annotation tool loomp which has the purpose to enable nonexpert users to provide semantic annotations.

[to be continued tomorrow, W3C-Day, XInnovations 2008, Berlin, Day 03...]

A short note on Semantic Search Engines

Lars featured an article on Hakia, a 'semantic search engine' in his blog today.
Contrarywise to Google, Hakia uses natural language processing (NLP) to 'understand' search queries given in natural language (and not as plain keywords). Ok, this is now new.
Just remember AskJeeves, a.k.a. Ask.com. But, in difference to that search engine, Hakia claims to perform a 'semantic search' (and not a keyword based search). If you read a little bit further in their technical description at Hakia labs you will find thet they are using a parser called 'OntoSem', which as they claim is able to perform a 'deep semantic analysis' of sentences. This parser is used for query string analysis and also for the generation of the search index. But, their search index is different to that we are used from 'traditional' search engines.
Just remember, traditional keyword based search engines extract so called 'descriptors' from web pages that are used to describe the content of those pages. These descriptors are managed within an inverted index. Thus, by accessing the inverted index with a descriptor (=keyword in query string), a list of web pages will be returned together with some weights indicating the relevance of the descriptor for the page.
So, how does the index work in Hakia? They clain, that their parser is analyzing all web pages to be indexed sentence by sentence. In effect, all possible questions that can be posed for each sentence are generated, forming what they call the 'QDEX data'. All possible questions for all sentences in all pages have to be stored in an index-like data structure (simply for fast and efficient access). If now a query string contains a question, this question is mapped against the index, resulting in a large number of 'relevant' (questions, sentences and in the end...) web pages. Now they apply a 'smart' algorithm called 'Semantic Rank' which orders the resulting list of documents according to their relevance wrt. the question given in the query string. More details about the technique is not published (at least not to my knowledge).

Thus, the only way to find out about the quality of their approach is to try out their search engine. A nice feature is that they have included small example applications where you may try out OntoSem or QDEX interactively by yourself. I have only tried OntoSem (because for QDEX you have to sign up a request) and the result was not really different from any standard NLP-Parser (and thus not really convincing. I will sign up and give QDEX a try..and of course I will post the result).

If I compare their approach to Google's, the problem is that Google is also claiming to incorporate semantic technology altogether. So, again I can only compare the results of my queries. I have tried several queries and have come to the following results:

• in general the results from hakia do look very promising!


• in comparison to Google results, they are not really that different this may come because of my 'queries' and thus, my results are probably not really objective).
  Let me give you an example: I was asking both (Google and Hakia) the question 'When did the Semantic Web start?'. Ok, it's not so easy to understand at all. First, it didn't really start (as an being implementation), but the topic itself startet several years ago...and so I was curious about the results:
  
  
  
  • Hakias result list presented an ecrim article (being co-authored by Tim Berners-Lee) entitled 'The semantic web lifts off' as top result.
  
  
  • Google's result list on the other hand presented (as second result) an artice entitled 'When did the Semantic Web enter our lexicons?', which at least answered one part of my original question.
  
  

Ok, as stated before, this is not representative at all, but I will keep an eye on Hakia anyway. As soon as I have more representative experiences (and hopefully some comments or other interesting comparisons), I will write about.

One thing in the end. After all, what I have read about Hakia and on their web site, they seem to have a completely different notion of 'Semantic Search'. What they do is to apply (enhanced) information retrieval techniques based on natural language processing. They do NOT evaluate any additional given semantic annotation (RDF, OWL, etc...), which might be addid into the web pages or connected to them. They claim (same as Peter Norwig) that the average user is much too stupid to supply semantic annotation (because for doing that he needs to be a linguist as they say).
Of course, if you are following the PLAIN trail of the W3C and encode everything by hand (HTML, XML, RDF, OWL, SWRL, etc...) then you have to be some real expert. But today, you (at least most of you) don't encode HTML by your own. Remember, there are lots of real nice WYSIWYG editors for doing that AND (even more) there are several simplifications (just think of writing your blog). In the same way there will be smart user interfaces and editors providing help in generating semantic annotations for your texts. First and most simple step is providing labels and keywords (just 'tags') for your blog posts. Most people do that...just because (1) they want to put some order into the set of their postings and (2) they want their posting to be found.

Semantic Search ... confusion ahead

When I was attending a talk of Thilo Götz on UIMA, the word came to 'Semantic Search'. Up to that point in time, I was quite sure about the meaning of this term. But, I had to realize that several people think in different ways about it.
As far as I have understood the meaning of that term, 'Semantic Search' refers to all techniques and activities that deploy semantic web technology on any stage of the search process. Thilo Götz (and he's not alone with that) refered to 'Semantic Search' as a 'traditional' search engine that is using a semantically enriched search index (i.e. a search index that incorporates ontologies or information/relationships infered from ontologies).

From my point of view the later definition refers only to a part of the whole process. Let's take a brief look at search engine technology: You have to consider the index generation (including the crawling processes, information retrieval for creating descriptors, inverse indexing, overall ranking, clustering, ...) as well as the user interaction (including query string evaluation, query string refinement, visualization of search results, navigation inside the search domain), not to forget personalization (concerning a personalized ranking of the search results including some kind of 'pre-selection' according to the personal information needs of the user, a personalized visualization, etc.) -- which will become of much more importance in the nearby future.

But, to generate a semantic search index there are several possibilities to consider:

• Using unstructured web-data (html, etc. ...) in combination with information retrieval techniques to map the information represented in the web-data to (commonly agreed) ontologies of well defined semantic.


• Using semi-structured web-data that already include links to well defined ontologies (being commonly agreed upon or at least being mapped to som standard ontologies).

For both steps, the generation of a semantic index requires more than just compilation of the retrieved data. Although the index might contain unstructured web-data including ontologies of well defined semantics, the main purpose of the index is to provide fast access to the information being represented in it. To generate the answer for a query, the search engine simply does not have enough time for performing logical inferences to deduce knowledge (a.k.a. answers) online. Of course, this (logical inference) has to be deployed beforehand (i.e. offline), just in a similar way as the computation of today's pageRank.

So, what is the use of machine processible semantics in a search engine's index data structure? The following possibilities can be considered (the list is open for additional suggestions...):

• to add new cross-references between single index entries (associations),


• to find similarities between index entries ( = web data) w.r.t. their content, and


• to discover dependencies between single index entries to enable
  
  • better visualization of the retrieved domain of information, and also
  
  
  • efficient navigation to better fulfill the users information needs.
  
  
  
  
• of course also to disambiguate and to cluster web-data for better precision and recall (this is already done with current IR techniques).
  

To compile a semantic index, also the crawling process has to be considered. While the primary goal of a web crawler is to gather as much web-data as possible as fast as possible (and of course to maintain its level of consistency), a 'semantic' web crawler besides unstructured web-data also has to look for semantic data, as e.g., RDF and OWL files, and also for possible connections between unstructured web-data and semantic data. For crawling RDF or OWL, a traditional crawler has to be modified. While the traditional crawler just analyzes the HTML data for link tags, RDF and OWL don't contain link tags but they often include several namespaces that determine new input for the crawler. Besides mere data gathering, the crawler should also preprocess data for being included within the index. This task often is implemented as a separate step (and denoted as 'information retrieval'). But, it influences the crawlers direction and crawling strategy and thus, also has to be considered here.
Web crawlers often are implemented in a distributed way to increase their efficiency while working in parallel. New URLs found in the web pages being crawled can be arranged according to the location of their domain (geographically). In this way, an instance of the distributed crawler receives only new URLs to be crawled that are located in the same (or a nearby) domain. The same argument holds for the resources that are to be crawled by semantic web crawlers. But, for semantic crawlers, also the (semantic) domain of the crawled data might be of importance, e.g., an instance of the distributed crawler might be responsible for crawling a specific domain (=topic) or only domains that are (semantically) closely related to the domain of interest.


For the semantic search engine the compilation of an index from the web pages being delivered by the crawler differs from the compilation process of the traditional search engine. Let us first recall the traditional index compilation process (for text related data, i.e. this does not hold for multimedia data such as images or video clips):

1. resource normalization, i.e. all resources that contain explicit textual content have to be transformed into text files


2. word stemming, i.e. transform all terms of a retrieved and normalized web-document to their word stems


3. stop word removal, i.e. cut out all terms that are not well suited for identifying the processed text file (i.e. that are not suitable as descriptors). Often only nouns are taken as descriptors (this can partly be achieved by applying pos-stemmers (=part-of-speech stemmers).


4. black list processing, i.e. terms that for some reason do not qualify as descriptors are cut out.

This process results in a list of descriptors that describe the web-document being processed. For each descriptor a weight according to its descriptive value for the text file has to be calculated (e.g., by term frequency - inverse document frequency (tf-idf) or other weigth function). The table resulting from combining the weighted descriptors with their underlying web-documents constitutes the index. By inverting this index a descriptor delivers all related web-documents in combination with their pre-calculated weight function (that determines how well a given descriptor is suited to describe the content of the according web-document). To increase precision and recall, the general relevance of the web-documents can be computed beforehand (i.e. nothing else but the Google PageRank).

For a 'semantic index', metadata (such as, e.g., ontologies) have to be taken into account and be combined with the traditional index...

...to be continued