Patent on a new paradigm to solve a problem

[Shayan]

Dear Sir/Madam,

I want to file a patent for a new paradigm for search engines. I just want to know if a new paradigm (without an implementation) is patentable or not. My invention is a new way of looking to the search problem and by this new way of looking and implementing the required components search results could be considerably improved. The point is that I offer a paradigm which consists of several components and each component could be a great research problem and many methods might be proposed for it. But here I do not offer any method and/or implementation for the components since my focus is on the paradigm itself as a whole and not on the individual components. Is this a patentable invention or I have to also include how each component should be implemented?

Regards,
Shayan

[Brad]

Based on your description you may not have enough specific details to file a patent.  In general, "abstract ideas" are not patentable.  A patent must have a "utility" or usefulness so you would have to describe your search solution using specific steps and examples of how it would be implemented by a search engine. 

[shayantabrizi]

In fact the steps are described in the invention. For example, a list of items, X, which represent something (described in the patent), is shown to the user. The user selects one of the X items. The system refines the search results based on the item selected, etc. But for example, it is not described how to extract the list of items, X, and only describes what X should look like and different methods might extract different sets of items. So it could be an open research problem. Is mentioning the steps of each sub-problem mandatory or just mentioning the top-level steps such as the example above is sufficient?

[Brad]

It is hard to give an answer without more detail but in general you must teach each and every step to allow someone to make the invention.  It usually not enough to say my invention does X and results in Y without listing specific examples of how X is done. 

[shayantabrizi]

So sorry for questioning again. Just as an example suppose faceted search (which is already known an used). The user enters a query and the system shows different possible meanings of that query. Then user selects the desired meaning and the system refines the results according to the selected meaning. For example, the user enters "JAVA" then the system shows a list of options including: Java Programming Language, Java Island, Java Coffee, Java Dance, etc. Then the user selects Java Island and the system refines the search results according to his selection and shows only the results related to the Java Island.

In this example, first different systems and implementations might identify different sets of candidates for possible meanings. Second, they might use different algorithms to refine the results based on the selected meaning. These are open research problems in this search paradigm and their solutions are not proposed by the patent. But the overall structure is described and the definition (input/output) of each component is known but specific implementations are not described.
If this system were not publicly known would it be possible to patent it?
Again sorry for taking your time.

[Brad]

It may be possible but you would have to give at least one working example of an algorithm/process for each step.   Ideally your patent would not be limited to that specific example algorithm but you should still teach or show one example of how it can be done. 

[shayantabrizi]

Thanks for your help

[shayantabrizi]

Hi Brad,

Sorry for taking your time again. But some idea jumped out of my mind that might solve the problem and I want to know your opinion. In the example of faceted search, a sub-optimal implementation of the search is to employ collaboration of the users (an idea similar to Wikipedia in which users produce the contents) in the form of tagging the webpages with the tags representing the meanings of the words in the pages to solve the problem of meaning identification and presenting them to the end-user (like folksonomy sites such as delicious.com, but tagging for the meanings of the words). For example, tagging a page about Java in which Java is a programming language and not a coffee with #JAVA:Programming_Language.
Although having a fully-automatic AI subsystem to extract such information is much more efficient and is probably easier to implement because you don't have to give incentive to the users to tag the pages and the machine does that for you, it is still some sort of implementation and a possible solution to that problem, although the validity of the business model might be in doubt and must be further probed (however we have seen many successful ones such as the well-known Wikipedia).
Thus, is it possible to propose this manual process as a possible solution of the meaning identification (in your terms: "at least one working example of an algorithm/process for each step") but not limiting the proposed framework to this specific implementation and have your patent accepted?

[Brad]

You have to be a little careful here.  The manual process can be patented IF there is also some algorithm or processing step being done by the computer.  Bottom line - the human cannot be the only one solving the problem, the machine has to have some unique processes running as well. 

[shayantabrizi]

OK. There is such processing. Thanks again