CADIE - a revolution in query flow analysis
- From RSS
Mikhail Daychik, Google
CADIE Product Manager, analyzes your query and, thanks to an analysis of related concepts, suggests how to better reformulate it to get more relevant search results, thereby reducing the barrier between machine logic and human thoughts.
The fact is that, unfortunately, our colloquial speech (for example, “you don’t know how much the dollar is now?”, “Uh, would you like a beer now ...”, “what time is it now?”) Is quite difficult to machine analysis. So far, the semantic context and what was meant between the lines has not been analyzed by anyone. Meanwhile, it is precisely these nuances that often carry a semantic load much greater than the text itself.
CADIE has been analyzing existing idiom dictionaries, forums, blogs, news, and what, how and when people are looking for quite some time. We also tried to reflect the connection between what and how people formulate and what kind of reaction they receive in response. For example, Mr. N. constantly resorts to taboo vocabulary on his blog, which increases his popularity, but reduces citation. We also conducted an analysis of how the reaction of readers corresponds to the expectations of the author and how often he returns to the formats used or tries alternative methods of communication.
Of course, CADIE technology is still far from understanding all the nuances of human communication, but a number of unsuccessful communication methods have been identified that people turn to with enviable constancy - starting with belated birthday greetings and ending with poorly formulated blog posts or incorrect search requests.
To help people avoid them, we’re introducing a series of improvements to the “Maybe you meant” clue system. For example, one of the very popular “mail” requests on google.ru domain, most likely, does not mean paper mail and chain mail (translation from English), but email and can easily be replaced by the word that has become more common in colloquial speech 'soap'. The results of such a replacement are highly likely to please you.
Naturally, each of our users is unique and each of them formulates requests in their own way, and therefore we cannot guess the intention correctly each time. As a result, we decided to be careful with the launch, and so far CADIE will be used only in those requests in which there is a high degree of confidence, and only as a hint, but we are confident that with the improvement of its knowledge base we can expand the volume of requests and accuracy hints.
For more information on CADIE, visit this page .
CADIE Product Manager, analyzes your query and, thanks to an analysis of related concepts, suggests how to better reformulate it to get more relevant search results, thereby reducing the barrier between machine logic and human thoughts.
The fact is that, unfortunately, our colloquial speech (for example, “you don’t know how much the dollar is now?”, “Uh, would you like a beer now ...”, “what time is it now?”) Is quite difficult to machine analysis. So far, the semantic context and what was meant between the lines has not been analyzed by anyone. Meanwhile, it is precisely these nuances that often carry a semantic load much greater than the text itself.
CADIE has been analyzing existing idiom dictionaries, forums, blogs, news, and what, how and when people are looking for quite some time. We also tried to reflect the connection between what and how people formulate and what kind of reaction they receive in response. For example, Mr. N. constantly resorts to taboo vocabulary on his blog, which increases his popularity, but reduces citation. We also conducted an analysis of how the reaction of readers corresponds to the expectations of the author and how often he returns to the formats used or tries alternative methods of communication.
Of course, CADIE technology is still far from understanding all the nuances of human communication, but a number of unsuccessful communication methods have been identified that people turn to with enviable constancy - starting with belated birthday greetings and ending with poorly formulated blog posts or incorrect search requests.
To help people avoid them, we’re introducing a series of improvements to the “Maybe you meant” clue system. For example, one of the very popular “mail” requests on google.ru domain, most likely, does not mean paper mail and chain mail (translation from English), but email and can easily be replaced by the word that has become more common in colloquial speech 'soap'. The results of such a replacement are highly likely to please you.
Naturally, each of our users is unique and each of them formulates requests in their own way, and therefore we cannot guess the intention correctly each time. As a result, we decided to be careful with the launch, and so far CADIE will be used only in those requests in which there is a high degree of confidence, and only as a hint, but we are confident that with the improvement of its knowledge base we can expand the volume of requests and accuracy hints.
For more information on CADIE, visit this page .