Provide a summary of your idea, application, product or service:*


NOTE: A natural language application cannot be taken seriously until it is in a demonstrable prototype form.  Please see the “current development stage” section for important details to help you evaluate this contest application.

Large online retailers are faced with a problem when it comes to offering specialized services for certain goods.  The profit margins on many items are too slim to allow providing more than a baseline level of service when it comes to helping a customer evaluate and select a product.  Amazon's online ordering and customer service servers are the best in the world.  They provide extended services such as customer reviews and ratings, similar product offerings, and now include social networking ideas like micro-forums and product tagging.  Yet there is a limit to how much service even Amazon can afford to provide on a general product line basis. Past a certain point, any additional software processing to aid the user in product selection could bring the server farm to its knees.

The solution is narrowly targeted software based purchase advisers that use the latest in natural language processing, supported by Amazon Web Services.  Luxury Purchase Advisers or LPA's for short, combine the latest in natural language processing with Amazon's Mechanical Turk service to provide a deeper level of online customer assistance in making a purchase decision. Constraining the inventory handled by one particular LPA by topic allows the scope of the natural language interface to be reduced to a size manageable by today's natural language technology. By also limiting the inventory to highly profitable items, it becomes economically feasible to maintain a healthy profit while paying for server time on Amazon's Elastic Cloud (EC2) servers, which is where the natural language processing code will reside.

All real life natural language services need constant refining.  It is impossible to anticipate the shape of every possible question a user will pose to the system. One user might ask to see "all Nikon cameras with red eye reduction" while another might say "show me a camera made by Canon that eliminates red eye".  This is where Amazon's Mechanical Turk service comes into play. Usually it is necessary to hire and train query analysts, an expensive process, to analyze the exception logs of failed queries and make progressive refinements to the software.  Fortunately the problem reduces down to that of having an editor see what the user asked and then figure out how to map it to terms the system already understands. This is a task any educated person can do since we all know how to rephrase a sentence, especially with when prompted with appropriate related suggestions, which the system will provide.  Therefore, this problem can be farmed out to the Mechanical Turk service where people will figure out the rephrasing and post the results via the service.  The LPA server software will take the refinements and incorporate them into the system, making the system ever more powerful as time goes by and with minimal intervention by a programmer or senior editor.  Refinements can be keyed to a particular customer profile, which will allow the system to adapt to a particular person's style of asking questions. Amazon's S3 storage service is the vehicle for storing the customer profiles while the SQS services would be used to coordinate job requests and data exchange between the system components.

Note: Over time there will be many LPA's.  As the time progresses, an LPA will reach the level of automation where the only needed human help is that provided automatically through the Amazon Mechanical Turk integration.  Once this happens a new LPA in a new luxury product area is created and the process continues. For example, the first LPA could be Digital Cameras costing more than $500.  The next might be Golf Equipment costing more than $500, etc.


When an LPA first comes online in one topic area, it will make money through mostly Amazon affiliate commissions. This allows the system to grow stronger, faster since there is no barrier to its use in the form of fees.  Once it reaches the level of sophistication where it runs virtually on automatic, and sufficient popularity has been achieved through marketing and word of mouth, a reasonable annual or monthly subscription fee could be charged.  For example, something like the Amazon prime membership fee.  Other special services and premium reports would be offered in the future as additional profit items. One possible idea for an add-on service would be a clipping service that monitors the web for news related to whatever products they bought, and would subsequently alert them, perhaps via E-mail or text messaging, when the news appears.

(c) 2007 Android Technologies, Inc and its licensors