This message raises the question of whether we should use a triple-store database for storing the metadata associated with the peer to patent application. Developer community input on this aspect of the technical design is strongly encouraged! For background, please reference Use Case 7 (Submit Prior Art). Many use cases have been recently updated. http://www.communitypatent.org/use_cases/use_case_7_submit_prior_art/index.html Per the use case, prior art will be organized as metadata that can be associated with a patent application. There are various types of prior art reference, and each type has different fields of data. So on a single patent application, you could have a several bibliographic references, as well as a reference to an open source or commercial software application, or files submitted by the citation author. Once a prior art reference is created, it will be associated with the patent application. At this point, community members can add comments and quantitative ratings to the prior art reference. Here is the proposed implementation. An expert using the Peer to Patent application could rapidly create a series of prior art references. Our current plan is to use an HTML or AJAX popup to capture the data and it will be stored using javascript (JSON) on the client side (in the web browser). On the server side we should implement an explicit model for the metadata following the MVC design pattern. So, for example, we could create a data structure to represent each type of prior art reference.. A simple way to do this would be to use Hibernate or pure XML fragments to store each reference. However, we need the ability to search across the fields of a prior art reference. Another requirement is that we need to associate various kinds of metadata with the prior art reference, including citation authors, comment authors, comments, and ratings. Also, we may want the ability to use the reference data across more than one application. Triple-store services (Mulgara, Kowari, Sesame) handle this situation fairly elegantly. First, building a relationship between a patent application and prior art reference is trivial. You add a relation to the triple store that links the two. Once you have the ability to access the prior art reference in the triple store, it also becomes easy to associate user comments and quantitative ratings with the prior art reference. The triple-store typically has an associated query language, which allows a wide range of queries to be constructed even after the system is deployed. To date, the triple-store is the only application framework that I have seen that provides a clean model for building relations between prior art, ratings, reputation data, comments. There are two dimensions, one is for runtime use and storage, the second is for post-facto inquiries and analytics. In order to understand the relationships between users in the community, we are going to need to do data mining on the relationships that build up in the metadata model. The explicit metadata model in the triple-store provides a means to find patterns of community activity in the future. For example, if the triple store has all the citations, as well as all the ratings and comments, you can use the inquiry language to find out the level of community interest in various patent applications. Clearly one can implement relationships between data using SQL tables and it would be speedy to implement but would not give you flexibility over the structure of the metadata model. In particular, there is no defined model for querying the system about metamodel relationships. So you would have to construct this model during the application design. Traditional threaded discussion groups (such as Slash) make it easy to associate comments with an article, but it is not obvious how they address adding comments or ratings to prior art citations. They can handle a 1:1 relationship fairly well, but they have no mechanism for dealing with a 1:many relationship. This situation occurs when a single prior art reference has links back to many claims. Another feature of the triple-store architecture is that you could allow for a concept of external citations. We could de-link the actual prior art citation data from our internal use of that citation data. In this scenario the reference data could be sourced out of a third-party system or database. The triple-store can still point to the reference using a DOI (See http://en.wikipedia.org/wiki/Digital_object_identifier ). The main problem with this is that we cannot search it. However, it is fairly easy to imagine that one can search for prior art in another external database, find it, and then import the reference for use in this universe. Once imported, then we could search on it from this universe. This is interesting in that it creates a possibility to link to other prior art or bibliographic citation databases. Questions: Q1: Is there any reaction to the idea of using a triple store? Q2: Is use of a triple-store ?over engineering? the solution, or is it called for? Q3: If a triple store is over-kill, are there suggestions on the approach for creating a persistent metadata model (as data structures) so that it supports the application at runtime and also analytics about the metadata model after the system is deployed. Sincerely, Eric Hestenes Technical Lead NYLS Community Patent Review project eric.hestenes at communitypatent.org -------------- next part -------------- An HTML attachment was scrubbed... URL: http://lists.osdl.org/pipermail/p2patent-developer/attachments/20061122/11ee5363/attachment.htm
