Wolfram Alpha is a new project by Steven Wolfram, british mathematician, physician and creator of Mathematica - technical computing application. This week Wolfram annouced that his company is going to launch new search engine on May. it is going to be named Wolfram Alpha. Known as innovative and good in writing computing algorhytms Wolfram may creata something really big. From what we could have read in the media buzz on Wolfram Alpha is going to be combination of semantic search engine with question gathering one.

The main innovation is fact that WA is going to generate answers in real time, without gathering them previously. It is something tottaly opposite to what we could have seen , for example, in Ask.com. Answers are “computed” from unstructurized data processed by engine’s algorhytms and than answers are generated. These algorhytms are said to be based on natural language, which is fully understood by the engine. Answers are going to be given in plain language and contain extract from indexed data.

Interesting thing is that index for this search engine is not like traditional one - created by some crawlers from data of web-origin and stats. Some parts of WA’s index are large databeses from various fields, huge amounts of information about physical world is gathered in them - WA is going to offer more formal sort of knowledge than for example Google, that bases on different types of media information, which often is informal.

Right now the project is in private beta phase.  You can ask for invitation or subscribe for newsletter.

Questions we can aks at this moment are concerned on the nature of the results given by WA, under some assumptions this project might look like step back. Is it going to be improved Wikipedia, with formal and credible knowledge? What about network business model - if users are not going to be contributors at the same time, would they like to use it, if the knowledge is not going to be democratized?

-mw

Aardvark is a search engine founded by ex-Googlers, that uses social network power. It is not based on information created by users and than indexed. It is distributing the query among your afiliates and their afiliates, who find themselves good at different fields - if you’re lucky you might get the answer. Idea is not stupid, or rather decent. User can ’search’ by using IM, e-mail, SMS or Twitter.

To increase probability of being answered you should take care and gather as much friends in the service as possible, which is a little bit hard especially at the begging after releasing public beta version. But there is an option to grab the contact list from Facebook connect.

After typing in the query the algorhytm searches for person that is most probable to have knowledge about what user is aksing. Than user just have to wait for the answer. Basis for poining on specific user as the one who has appropriate knowledge is the fact that he introduced himself as having knowledge on particular field, coming from specific region or interested in something user asked about.

It looks like this sort of human powered search is one with completely new approach. It is interesting how many degrees of affiliation can be potentially used. However it can be a great source of help and really practicl information in some cases. Except for that Aardvark looks like it is going to be not only the source of information but also sort of place to make friends and have fun.

-mw

Swagbucks is a search engine that offers its users a sort of a game. By searching you can win prizes. All you have to do is to register - to enable Swagbucks’ administrator localize you and send you prizes. It works very simply - the more you search, the more valuable prizes you get. This service is also connected with e-shopping - when you shop you earn some points that you can later change on prizes from offer.

Swagbucks is also kind of Argentinian piramid loyalty programme. You can invite you friends and when they are getting points you get part of it. These who register as really first ones will get more of goods. So hurry up if you want to get a piece of pie.

Service as entity looks like someone connected few different marketing tools to camouflage e-store as search engine. If this will work and gather nice community around it might bring a proper income.

-mw

Libris is a swedish search engine geared towards searching for books, gathered in swedish universities libraries. Since now it added an option of adding semantic data to records from catalogues. System bases on existing ontologies like DC, SKOS, FOAF and Bibliontology, but also offers references to other web resources like DBpedia.

Librarian resources data published as Linked Data enables to integrate data from catalogues with Web. Simillar attempt have been taken recently by authors of LCSH project, who were trying to include Library of Congress resources into semantic network, but due to unofficiality of whole process project was closed.

-mw

Google has introduced new feature to its search engine, semantic analysis of query with providing the answer for our question. It is possible to formulate a question and get answer for it at the top of list of results. At the moment number of answerable questions is seriously limited, but still impressive. Semantic analysis can deal just with recognizing geographical names, can distinuish famous people and their relatives, provide dates of birth and death of famous people, but also answers for questions such as “what is XY famous for?”, “who is the author of book X?” or “who is the president of X?”. Data extraction processed from all search results seems to be quite advanced. In addition with each answer we get the refference to its source.

Although it seems to work efficiently, all names have to be given in full versions - like Frederic Chopin insted of Chopin. There are also some problems with names of companies - algrithm does not deal with abbreviations.

So far Google didn’t annouce any official statement or explanation how does it work. It looks like algorithm was based on extracting structural data from RDF and HTML as well. Main improvement is introducing mechanism that analyze the query in order to define relations between objects and if these relations are clear the answer is displayed separately from other results.

It can be a breakthrough in semantic web developement, ability of deriving the meaning from websites content and extracting the data into answer enable to save time by removing the need of crawling through search results. Except for that, these feature seems to be a foundation for further improvements - if we can get answers for simple questions why not try to get answers for logical questions and than why not ask questions in more informal language, not with plain words?

-mw