Is Natural Language Search Better?

Recently I read an interesting VentureBeat article on a new natural language search engine called Powerset and I became immediately intrigued. I really wanted to try it but the search engine was only available by invitation and I learned it would only be released to the public sometime later this year in the fall. Since I was disappointed and wasn’t able to test Powerset I decided to investigate and see if there were any other natural language search engines available that I could try. I was able to identify several and they included Hakia, Lexxe, and Brainboost. So I decided to learn more about what natural language search entailed.

In really simple terms I came to understand that a natural language search engine allows you to use a plain regular spoken language such as everyday English to enter your search query. From my initial cursory understanding using this type of search, you can ask a question or type in a sentence to describe what you are looking for and the search engine will process the sentence through programmed logic to find the keywords by their position in the sentence. With such a basic understanding I decided to try out some simple queries on each of these three search engines and see how they compared with each other and with Google.

The first query I tried was a simple one: How old is Hillary Clinton?

Brainboost was able to return the year Hillary Clinton was born to me which was 1947 as the first link while Hakia was able to return her birth year on the fourth link. With both these search engines I would have had to calculate her age myself. Lexxe on the other hand did a much better job and returned me the answer: Will Be Sixty-one Years Old in 2008 as the first link. Google returned me her age on the first link as 59 years old and her birth year as 1947.

So I decided to try a geometry question next. The query I posed was: How many sides in a pentagon?

Lexxe returned: There are five.
Brainboost returned: 5 = S in a P – Answer >>> Sides in a Pentagon.
Hakia returned: A polygon with five sides is a pentagon.
Google returned: A pentagon has 5 sides and 5 angles.

I then posed this query: How many programming languages?

Only Brainboost returned something that seemed relevant as an answer: Thousands of different programming languages. The others including Google disappointed.

Another query I made was: What is contributing to climate change?

All the search engines returned results in somewhat different order but for the most part found links that showed humans may be contributing to climate change in some fashion or other.

The last natural language query I thought would be interesting to look at was: Will natural language processing provide better search?

Lexxe’s response to this question was indeed amusing as it answered “no” but it did provide some very useful links on natural language processing. Brainboost faired the worst on this search as it only returned one link on natural language processing. Hakia and Google also both did a good job at returning useful links on the subject matter. Hakia listed this rather useful link on Natural Language Processing that I thought was especially noteworthy to mention for those wanting to learn more about the subject.

Obviously many other search queries can be compared but that I will leave as an exercise for yourself. From my superficial analysis all these natural search engines appear to perform adequately and probably as good as the traditional Google search. However, I am unqualified to determine whether they are indeed better or worse. The issue is quite open to debate. Here are some links for you to read to get a more in-depth perspective on this debate:

On Natural Language Search
We Are All Natural Language Searchers…
Hello Natural Language Search, My Old Over-Hyped Search Friend
Jeff Bezos’ Risky Bet
In a Search Refinement, a Chance to Rival Google

I suppose only time will tell which type of search is better but at the moment these natural language search engines do provide an interesting alternative.

Below are some links if you want to read further about natural language processing:

Wikipedia: Natural Language Processing
Natural Language Processing Frequently Asked Questions
Parc Research: Natural Language Processing