Content search engine UpTake started its work

According to last year’s comScore survey, planning a trip and buying tickets via the Internet has become an indescribable nightmare for those who are preparing to relax. And the reason is not that it is difficult to conduct operations with some ticket services, but in a huge amount of unnecessary information that you have to deal with when planning a trip. According to a comScore survey, on average, in order to make a trip plan, you need to use 12 travel-oriented search engines to visit 22 websites: this usually takes about 29 days. The semantic search project UpTake (formerly Kango) is designed to make it easier for users.

UpTake is a search engine with a vertical search engine, which has the largest database of more than 400,000 hotels and entertainment establishments in America, mentioned on more than 1,000 sites. Using such a top-down vertical approach, UpTake managed to collect more than 20 million reviews, opinions and descriptions of US hotels and entertainment venues and provide this information by meaning. This is something like Metacritic just for travelers. The main criterion for the search is not only the popularity rating, UpTake also tries to characterize the hotel or entertainment establishment on the basis of the information that the search engine draws from the database. For example, is a hotel suitable for a family holiday, is a good place for which you are requesting information for a romantic trip, is the price of a holiday acceptable, etc.



“UpTake also provides the user with the most useful descriptions, photos, etc. for the most suitable hotels and entertainment establishments, the choice of which is based on attributive and relative analysis of reviews and other textual data. Analysis is intended to select the most important meta tags based on our travel ontology, ”says Yen Lee, President of the company, one of the founders of the CitySearch office in San Francisco, and the former General Manager of Yahoo! in the field of travel.
What UpTake can't be compared to is with ticket services, such as Expedia, with search services whose search is based on the cost of rest (Kayak) or the community of travelers. UpTake is only engaged in sorting descriptions and semantic analysis; using this service it is impossible to order tickets. According to the company, only 14% of users form a request for booking tickets, which symbolizes the following: users are more interested in exploring places of rest, the cost of rest for them is secondary. Many listings on the site have the opportunity to check the rating of the hotel, the data is taken from third-party sites. So in UpTake plan to earn.
')
UpTake works with the help of a specially created travel ontology, which consists of objects, links between these objects and the rules that generate the result, because the database contains 20,000 reviews and descriptions. This ontology allows UpTake to extract meaning from the structured or partially structured information of the form “pool, the level of hotel comfort, and children love pools”. Thus, the hotel will receive additional points if its description meets the phrase "suitable for families with children." This ontology also includes the fact that the pool in which they swim naked is unacceptable for children.
A simplified ontology model is presented below.


In addition to everything, the system takes into account the way you plan a holiday - a romantic trip, a family holiday or outdoor recreation - the site takes into account the state of mind to determine whether the user will like something from what will be presented in the results search. The search engine is looking for words with an emotional coloring, such as “like”, “love”, “hate”, “downtrodden” or “with a good view”, and based on these words it concludes that people like the hotel or not, and that they think about this place. All these collected data are reflected in the rating, which exists for each place of rest.

Conclusion

Yesterday, we quickly tested the Powerset search engine, based on a regular language search, and found that the results are almost the same as Google’s. We wrote that “Google does not give up its position and remains a“ pretty good ”solution, it will take a lot of time for the Powerset to lure users to it.” But, despite the fact that the semantic search may seem clumsy and cumbersome for the modern global network, it is still able to make a large semantic contribution to various thematic verticals. An ontology is more limited, and such a site does not attempt to answer specific questions, but it has the ability to semanticly define a concept, such as romance or, more importantly, a definition of quality.

I asked En Li what UpTake thinks of a vertical top-down approach compared to the traditional bottom-up approach. Lee replied that the top-down approach has all the chances to oust the bottom-up approach from the Semantic Web. Lee believes that the top-down approach attempts to extract meaning from unstructured or partially structured information, along with Yahoo! embedding a semantic add-in into the index will encourage content publishers to start using semantic modules in information. Lee also noted that many partners have already contacted UpTake about how to organize information in such a way that it is easier for the service to analyze it.

It is assumed that search engines with a vertical system, such as UpTake, will help to more fully represent the needs of users and thus it will be easier to sell the service to the end user in the semantic Web. The ability to search among millions of descriptions and reviews and make the computer show exactly what you need - that’s what the main commercial value of the idea of ​​a semantic Web is. When such technologies become more perfect, there will probably be engines that will become the basis for neuro-linguistic programming, perhaps the Powerset will start to show results much better than Google (but you shouldn’t forget for a minute that Google will not sit back and wait by the sea weather ...)