Semantics add value to new travel site as UpTake emerges from private beta
For those of us based in Europe, the announcement of a 'Search and Discovery Site for US Hotels' would normally be greeted with (unfair) muttering about Americans forgetting (again) that there's a world beyond their shores, or (less unfair) grumbling about the relative lack of good and affordable data to support similar ventures elsewhere. I, however, am off to San Jose on Saturday for the Semantic Technology Conference, and therefore had a great excuse to put the site through its paces.
Inputting my search ("Fairmont Hotel" and "San Jose, California") did bring the right hotel back at the top of the list. Rather disconcertingly, it also returned 156 other records for completely different hotels. I queried this with Elliott Ng, VP of Marketing, who explained that their known-item name matching is currently being 'supplemented' with related records from the database. He also made the valid point that the same search in Google returns 67,500 results, commenting that
"the expectations are higher for vertical search such that 157 seems to be a lack of precision."
Indeed. Our expectations are higher for vertical search sites, although there may be some justification for that.
Digging into the detailed listing for the hotel itself, the site does a nice job of summarising sentiment from across the main review sites.
The snippets of text are a little shorter than my real-world need found useful, but the aggregation of views does provide some value, and as UpTake president Yen Lee notes, semantic analysis is at work behind the scenes to attempt an interpretation of the sentiments expressed in those 20 million-plus traveller reviews of some 40,000 US hotels;
"UpTake matches a traveler with the most useful reviews, photos, etc. for the most relevant hotels and activities through attribute and sentiment analysis of reviews and other text, analysis guided by our travel ontology to extract weighted meta-tags. More simply, we break apart and analyze reviews and articles so we can recommend the best products for you"
As today's press release describes,
"For a user looking for a family hotel in San Diego, UpTake analyzes its San Diego hotel catalog for attributes like 'pool', 'babysitting', 'oversized rooms', '3 and 6 year olds' and for sentiments such as 'like', 'love', 'strongly recommend.' UpTake users looking for San Diego romantic hotels will then get different results than those looking for San Diego family hotels. UpTake also tries to understand user intentions. If a traveler is looking for a hotel that is 'good for kids', UpTake interprets it to have the same intent as phrases such as 'child friendly' or 'family vacation.' By aggregating reviews and the most comprehensive selection of products on UpTake, the site will save travelers the time of going from site to site to find the right review for the right product to make your decision."
A traveller searching for a child-friendly break will receive a completely different set of results to those offered to the business traveller. It's an interesting concept, and one that - in principle at least - does a good job of applying some semantic techniques to enrich the experience without forcing the traveller to interact much differently than they would with a regular travel site. Semantic technologies really can add value without needing to be overt or particularly visible to the user.