Pictured above is what I see when I search “What time is tomorrow’s Giants game?” on Google.
It’s terrible.
Look at the sidebar on the left. What does it say for location? San Jose, CA. And where are the top two results? New York. Nearly 3000 miles away.
The next two hits have the right team (thankfully), but are both from 2010. I had “tomorrow” right there in my query. This is another missed opportunity.
The next two hits are for the wrong city, again, and rounding out what I saw above the fold is ask.com telling me tomorrow is March 27th. Disappointing.
The bottom three results (the rest of Google’s first page) were equally unhelpful. Beneath that were some ads promoting the Giants, and even mentioning their schedule, which is completely infuriating; why are the targeted ads more relevant than the search results themselves?
So, next time someone tells you that search is a solved problem, or that nobody will ever be able to beat Google, you can remind them that search is hard, and even the Giantest search engine of them all still has a long way to go.
2 replies on “We Still Haven’t Solved Search”
Isn’t that actually an “interpreting natural language” problem?
After all, it would take you fewer words to writea search that gets you the info you want on the first try (IE, enter meaningful search terms instead of the useless filler words we use when speaking.)
I agree entirely that we don’t have natural language interpretation down (just look at the mess that is Siri, and how disappointing it is in comparison to the ads and to the classic sci-fi speech interpretation, such as that on ST:TNG). And it’s a very difficult problem, too, as grammar “rules” are fast and loose at best.
I’m just not sure saying this is a problem with search per se is accurate. :)
It’s true, this is more of a natural language problem than a search problem. It’s just that Google has been the best at search for so long and it feels like we’re getting the same search results we did 10 years ago.
None of this seems especially difficult. Goole knows my location, and it should be able to guess I’m asking about sports — which is heavily related to location. Why not use that information to affect the results? It’s a stretch, but it’s far from impossible.
The real killer is that the ads were more relevant than the results. Or really, that the ads were relevant, and the results were not. What does this say about Google Search as a service? Making money and providing the best results aren’t the same thing anymore. Which one is a higher priority?