Perhaps the biggest problem with natural language search is that it's incredibly difficult to try and automate machine-assigned ontologies. Essentially, machines just don't get it.
This is precisely the reason why Canadian technologist Bruce Johnson switched his focus from semantic tagging to a new style of search. Says Johnson, "Machines don't really deconstruct language well. They miss so many of the ambiguities and they often don't pick up on synonyms." As a result, Johnson's Semanti was built in the belief that humans are best at determining search relevancy. ReadWriteWeb spoke to Johnson, about how his start up, differs from some of the semantic web's more-recognized players like Hakia and Powerset.
Most semantic search services are natural language search engines; however, Semanti employs a system of personal bookmarks, a drop-down menu with multiple definitions, and search recommendations pulled from Facebook friends. Semanti actually increases relevancy by introducing human eyes and opinions into the search process.
Semanti's search experience is incredibly different than natural language search engines. Firstly, Semanti is not a separate engine, but a Firefox plug-in to be used with Google, Yahoo or Bing. This means that users are not being asked to change the habit of visiting their preferred search engine. Secondly, every time you type terms into your preferred engine, Semanti Suggest automatically provides a drop down menu of possible definitions. The search term "apple" offers the fruit, the company, the car, the record label, and Gwyneth Paltrow's uniquely named baby. From here, users choose the relevant definition and results from alternate definitions are omitted. This portion of the search is particularly useful if you are looking for one of the seemingly obscure definitions.
Additionally, Semanti uses a series of bookmarks to further categorize results. Users install the Semanti browser plug-in and bookmark the pages they are likely to revisit. However, unlike regular web-based bookmarks, Semanti's MyWeb stores the entire page's text and the original search terms. Later, when the same page is needed, Semanti scans the bookmarked pages and offers them at the top of search results. This means that the more you search, the more relevant your results are likely to be. Once you've saved a few pages to MyWeb, you can also increase search relevancy through Semanti Social Search.
Semanti Social Search is a web-based community where you can invite your Facebook friends to share their MyWeb bookmarks. Once a friend has accepted your invitation, you've got access to their search results in addition to your own. Now every time you search, Semanti offers your bookmarks and your friends' bookmarks at the top of search results. Below your friends' bookmarks, you can also view bookmarks from the Semanti community. Or if you'd like to turn off these suggestions, Semanti also offers you that option as well.
Semanti is clearly a new approach to finding information. The drop down definitions are very likely to speed research efforts; however, because the social search component is dependent on our friends, it will be interesting to see how results emerge over time. It's always a celebration of human achievement to see a community effectively "harness the wisdom of the crowds"; however, there's always a chance that a good thing can be destroyed by the ignorance of the masses. The community will just have to be discerning with their invitations.
Semanti is currently available in French and English and the team of nine plans to roll out additional language versions in the coming months. To install Semanti and test drive the product yourself, visit Semanti.com.