Sorting Through Data Warehouse Updates
The CLAIMS Direct data warehouse is updated continuously. On average, 1-3% of the entire database is touched weekly. These 1-3 million updates fall into two categories: new documents (roughly 20%) and updated documents (roughly 80%). New documents are, of course, new publications from issuing authorities, US, EP etc. Updates are generally, but not limited to, CPC changes, family changes (adding family ID), IFI integrated content (patent status, standard names and claims summaries), reassignments, legal status, et al. I suggest reading about content updates for a better understanding of load sources and issuing authorities. This blog will describe how to determine what has been updated and further, differentiating new content from updated content at the document level as well as container level.
As described in the content update documentation,
Every document within the data warehouse was loaded as part of a group of documents. This set of documents is identified by a load-id (integer value). There are 3 types of load-ids in the data warehouse: created-load-id, deleted-load-id and modified-load-id. The created-load-id represents the load-id in which a document was added to the database, the modified-load-id represents the load-id that last modified the document and the deleted-load-id represents the load-id in which the document marked as deleted.
xml.t_patent_document_values houses the
load-ids that affect the overall status of the entire document.
load-idthat last modified the document, whether it be addition, correction or deletion of content
load-idthat the entire document was marked deleted
load-idwhen the document was added to the data warehouse
Each satellite table also contains a
modified_load_id column that is specific to the content in that table and will always be <=
xml.t_patent_document_values(modified_load_id). The following table should clarify the sequence of update events and the effect on
load-id values for three database tables.
assuming ifi integrated content is not delivered on date of publication
parties are not updated with IFI content loads
This granularity offers the ability to trace modifications to individual containers (tables) as well as the overall modification status of the document as a whole.
How many US documents published on 20161004 have updated CPC classifications?
Out of 6680 documents published on 20161004, 6145 had CPC modifications after publication.
Knowing which tables and containers have changed is important, but the when and why are also critically important. The information gathered during data warehouse updates is stored in
reporting.t_client_load_process. For a detailed description of the contents of this table, see CLAIMS Direct Design Documentation. Of the many questions that can be posed, for our purposes in this blog, we are only concerned with answering the following questions:
- On any given date, how many documents were added to the data warehouse?
- On any given date, how many documents were modified in the data warehouse?
The answers to these questions are important for a variety of reasons, most importantly for down-stream processing. In the case of CLAIMS Direct primary data warehouses, we trigger indexing based on the answer to these questions. The first thing to look at is what happened on a certain date, for the sake of an example, we will use 20161004.
How many documents were added to the data warehouse?
... and which documents were updated?
In part II of this series, I will detail methods which can be used to monitor this table and react to changes either at the load-id-level or over the course of an interval of time.