Aardvark: A Search Engine To Show The Way?
Atul Chatterjee
May 04, 2009
I want to go on a vacation to Mombasa at $75 per day for 10 days. What are my options?
Now you don´t know anyone who can help you out with this, so you go to Ask Jeeves, Yahoo, MSN, Google, Dogpile, Clusty or others. Each returns 20 pages of links to research, that´s a load of homework.
You sniff around a few forums or go to some social networking sites to latch on to people who have been to Mombasa. A friendly soul advises you on snorkeling, a room with a view of the sea, mosquito nets, prices, tips and so on. This is the way Aardvark works.
Right now Aardvark uses folks on Facebook and other networks as its knowledge bank. Aardvark takes your query and matches it with profiles on these social networking sites. Then it transmits the question to the selected group. If anyone has the answer and the willingness (s)he replies.
There are three problems associated with this approach. Firstly, a person knowing the answer may not reply. Secondly, a person who knows the answer may be missed out by the search machine due to a profile mismatch. Thirdly, the worst scenario, conflicting answers may be returned.
If it remains a human driven knowledge machine it will not work quickly. So one can assume that ´expert´ knowledge will be stuffed into it and in due course it will be driven by some form of a knowledge machine.
Expert systems have been developed since the 1960s, one of the earliest applications being the recovery action to be taken if a bit broke during the process of boring an oil well. The system worked by interrogating the person on the site through a series of questions and then suggesting action. The knowledge of the system was created by questioning a few human experts.
These expert systems work with a narrow domain and are not general purpose search machines. If they are general purpose search machines they are unlikely to solve complex problems in a wide variety of domains. The question at the start of this piece can be classified as a complex one.
Wolfram Alpha is a gigantic general purpose project due to be launched in mid 2009, this also accepts natural language search terms, uses private databases and the Internet. Start, a project at MIT, has been offering searches using natural language since 1993.
Right now Asrdvark is under the wraps though it is reportedly being tested out by 10,000 users and has already started collecting advertisement revenue.
Ventilla the founder and an ex-employee of Google has already pulled in $7 million as venture funds for Aardvark. Obviously investors believe he is on to a good thing. While at present it depends on human inputs of knowledge from social networking sites, the future is speculation. The chances of it servicing a specific domain are quite likely.
The emergence of Aardvark is a pointer for others to create specialized user friendly search engines with some level of expert knowledge. The shortcomings of the current search engines are obvious to most of us.
Now you don´t know anyone who can help you out with this, so you go to Ask Jeeves, Yahoo, MSN, Google, Dogpile, Clusty or others. Each returns 20 pages of links to research, that´s a load of homework.
You sniff around a few forums or go to some social networking sites to latch on to people who have been to Mombasa. A friendly soul advises you on snorkeling, a room with a view of the sea, mosquito nets, prices, tips and so on. This is the way Aardvark works.
Right now Aardvark uses folks on Facebook and other networks as its knowledge bank. Aardvark takes your query and matches it with profiles on these social networking sites. Then it transmits the question to the selected group. If anyone has the answer and the willingness (s)he replies.
There are three problems associated with this approach. Firstly, a person knowing the answer may not reply. Secondly, a person who knows the answer may be missed out by the search machine due to a profile mismatch. Thirdly, the worst scenario, conflicting answers may be returned.
If it remains a human driven knowledge machine it will not work quickly. So one can assume that ´expert´ knowledge will be stuffed into it and in due course it will be driven by some form of a knowledge machine.
Expert systems have been developed since the 1960s, one of the earliest applications being the recovery action to be taken if a bit broke during the process of boring an oil well. The system worked by interrogating the person on the site through a series of questions and then suggesting action. The knowledge of the system was created by questioning a few human experts.
These expert systems work with a narrow domain and are not general purpose search machines. If they are general purpose search machines they are unlikely to solve complex problems in a wide variety of domains. The question at the start of this piece can be classified as a complex one.
Wolfram Alpha is a gigantic general purpose project due to be launched in mid 2009, this also accepts natural language search terms, uses private databases and the Internet. Start, a project at MIT, has been offering searches using natural language since 1993.
Right now Asrdvark is under the wraps though it is reportedly being tested out by 10,000 users and has already started collecting advertisement revenue.
Ventilla the founder and an ex-employee of Google has already pulled in $7 million as venture funds for Aardvark. Obviously investors believe he is on to a good thing. While at present it depends on human inputs of knowledge from social networking sites, the future is speculation. The chances of it servicing a specific domain are quite likely.
The emergence of Aardvark is a pointer for others to create specialized user friendly search engines with some level of expert knowledge. The shortcomings of the current search engines are obvious to most of us.
Print
Email
