Redesigning Patent Database

McNair Project

Redesigning Patent Database
Project Information
Project Title	Redesigning Patent Database
Owner	Shelby Bice
Start Date	201701
Deadline	201705
Keywords	Database, Patent
Primary Billing
Notes
Has project status	Active
Copyright © 2016 edegan.com. All Rights Reserved.

Documentation on the process and eventual designs for the new patent database. Not to be confused with "Patent Data Restructure", which deals with condensing and cleaning USPTO assignees data.

Deliverables by end of semester: Design and build new patent database with core tables from data, add tables from Patent Data Restructure and Small Inventors Project, write thorough documentation on schema of new patent database and instructions on what documentation should be written when it is altered

TODO on 4/13/2017 Keep working on updating patent data

• How to parse patent data
• How to parse assignment data
• Also update documentation as I go

Redesigning Patent Database

• Design a better representation for database
• Fix scripts if necessary
• Start moving data into new database by querying existing databases (using SQL)
• Use scripts to query new data
• Test database
• Remove extraneous information from database (copies, patents that we're not interested in, etc.)

Documentation I need to include:

• Schema of new database (with justification of design), would like to include a visual representation
• SQL commands that were used to fill database with explanation of what they do
• Clear instructions on where to find scripts in bulk drive and an explanation of what each script does
• Visual representation of example table entries that isn't just copied and pasted from a CSV file

Related Projects

• Patent Data Restructure
• Small Inventors Project - uses Fee Status and Citations
• Medical Centers and Grants - uses patent assignees, specifically their zipcodes and organizations

Documentation Relevant to Current Patent Database

Previous documentation on the patent database:

As of 3/21/2017 the most up-to-date database containing patent data is "patent" not "allpatent" or "allpatent_clone" and "patent" is the database that the the other patent data redesign project, Restructuring Patent Data (link above) is working with.

Patent Data - overview of what the data is and where it came from, probably starting point for changing documentation

Patent Database - overview of schema of database (specifically, the database patent, which includes data from Harvard dataverse (originally stored in patentdata) and USPTO (patent_2015)

USPTO Assignees Database - enhances assignee info in patent database, also being redesigned

Problems with Patent Database - lists issues with current schema

Previous ER Diagram - does not match up with schema described in Patent Database and contains outdated list of what we want to pull from XML files

Processing Patent Data - states that allpatent is the newest database and an amalgamation of patentdata or patent_2015

Description

The purpose of this project is to create a new, redesigned database to hold all of the patent information that the McNair Center has accumulated and document the process so that the design can be easily understood and replicated or edited as needed.

This database will include design patents, utility patents, and reissues.

Development

Design will be built upon a relational database model. I will be referencing this article on database design as I develop the design (http://en.tekstenuitleg.net/articles/software/database-design-tutorial/one-to-many.html), and I will be creating an ER diagram using ERDPlus or Creately.

Current Design and Scripts Documentation

The scripts for querying data for the patent database exist in McNair/software/scripts/patent. If the design for the schema of the database tables are altered too much, may have to write new scripts.

The following pages are relevant to how previous databases are built/how to build tables in the database:

Harvard Dataverse Data - explains how to make tables from Harvard Dataverse data, where to find scripts, etc.

USPTO Data - explains how to make tables from USPTO data, where to find scripts, etc, specifically for assignment data.

Patent Data Extraction - explains locations of XML files and lists (at the bottom) where the Perl scripts can be found

Patent Data Cleanup - explains changes that were made to clean up problems in the database allpatent as a result of merging the Harvard Dataverse data and the USPTO data

Patent Data Processing - SQL Steps - explains SQL needed to merge two existing databases, one that contained the Harvard Dataverse data and one that contained the USPTO data

Here at the instructions I'm developing for downloading, parsing, and hopefully adding new data to the database since the documentation is very sparse (can also be found under McNair/Projects/Redesigning Patent Database/Instructions on how to download patent data form USPTO bulk data.

How to run Perl Scripts to extract Patent Data

1. Log on to the RDP
2. Open powershell
3. Change directory to wherever the script is located by doing: cd e:/mcnair/whatever
4. Run the script by doing:

perl scriptname arg1 arg2 ...
 or
perl scriptname argument1=arg1 ...
 where arg1 and arg2 are arguments passed to the script either directly or by name (e.g., argument1=)

For the Main Patent Data

You'll need to run a series of scripts:

1. USPTO_Parser.pl to get the zip files from the USPTO and unzip them
2. splitter.pl to break the files into individual xml files

USPTO_Parser.pl

USPTO_Parser.pl can be found under

E:/McNair/Software/Scripts/Patent

Notes:

• You may need to fix the base url in the script. The USPTO has changed it before. It is currently: https://bulkdata.uspto.gov/data2/patent/grant/redbook
• Instead of taking a text file containing a url, as the USPTO_Assignee Download does, it takes two arguments, year 1 and year2, which are supposed to represent the range of data that you wish to download (for example, 2015 to 2016).
• The perl script places the downloaded zip files into "E:/McNair/PatentData/name" where "name" is the name of the zip file.
• The folder "Processed" under McNair/PatentData appears to hold all the unzipped zip files that have been downloaded and processed already. So if you are curious if some files have already been processed, you could look there. They are organized by year.

Now to actually run the scripts:

Open a command line or powershell
e:
cd .\McNair\Software\Scripts\Patent
perl USPTO_Parser.pl 2016 2017

Note the zip files should appear briefly (sequentially) in E:/McNair/Software/Scripts/Patent before disappearing and reappearing unzipped in E:/McNair/PatentData

splitter.pl

Now we need to split the files into individual, valid xml files. To do this:

Move the files to be split into E:/McNair/PatentData/Queue
Go to: E:/McNair/PatentData (this is the version of the splitter to use)
Run the command: 
 perl splitter.pl
Each file will then be blown out into a directory of xml files in E:/McNair/PatentData/Processed

Notes:

• Change line 26 of script to reflect the year (name of the directory that you want to put the split files into the appropriate directory)

this will go ahead and put the files in the directory so you don't have to copy them.

xmlparser_4.5_4.4_4.3.pl

The next step is to parse the actual files. Do not use the perl script PatentParser.pl. This script is out of date.

Instead go to:

E:/McNair/PatentData/Processed 

and use the perl script called xmlparser_4.5_4.4_4.3.pl by running (for example):

perl xmlparser_4.5_4.4_4.3.pl E:\McNair\PatentData\Processed\2016

This will load the data into the database. Running splitter.pl should have loaded the files into a directory. Pass the path to the directory to the parser and it will parse the data and load it. Notes:

• The parser will open a connection to a database on the RDP's installation of postgres. It will then put the data directly into this database. Once complete. we manually move the tables to the dbase server's database (i.e. patent).
• The password hint is tsn. You can run pgAdmin III to connect to it.
• The default database is called PatentDB
• If you want to make a new dbase, you'll have to run a sql script to make the tables. It is E:\McNair\Software\Scripts\Patent\createTables.sql
• The script populates an inventors table. We may have failed to move this table over to the production dbase server.
• If you are updating the data, make sure that you don't add duplicate records to the dbase. The easiest way to fix this is to make sure that you don't parse duplicate xml files.

During the last update:

• ipg160322_6745.xml was the last loaded XML file. It covered <doc-number>09295186</doc-number><kind>B2</kind><date>20160322</date>. I confirmed that this was the last file in the E:\McNair\PatentData\Processed\ipg160322.
• I also confirmed that this was the highest patent number with the highest grant data in the dbase PatentDB
• I therefore put every folder from ipg160329 to ipg161227 into the E:\McNair\PatentData\Processed\2016 folder
• The script (xmlparser_4.5_4.4_4.3.pl) required substantial modification. In particular, the data structure implied by E:\McNair\Software\Scripts\Patent\createTables.sql was inadequate. As a consequence, I created a new database called PatentUpdate and modified its structure. All of the additional data is now loaded into there.

Next steps

We now need to:

1. Retrieve the data out of PatentUpdate and reprocess it so that it will fit into patent on the main db server

Cleaning up patent data and inserting it into the existing patent data table

For the table called "patent"

INSERT statement:

INSERT INTO patents_merged ( SELECT patents.patentnumber, patents.grantdate, patents.prioritydate, patents.prioritycountry, patents.prioritypatentnumber, patents.cpcsubgroup, patents.pctpatentnumber, patents.numberofclaims, 
patents.applicationnumber, NULL, patents.filingdate, NULL, NULL, NULL, NULL FROM patents );

Equivalent COPY statements:

First copy data from RDP to a CSV file:

\COPY (SELECT patents.patentnumber, patents.grantdate, patents.prioritydate, patents.prioritycountry, patents.prioritypatentnumber, patents.cpcsubgroup, patents.pctpatentnumber, patents.numberofclaims, 
patents.applicationnumber, NULL, patents.filingdate, NULL, NULL, NULL, NULL FROM patents) TO  '/tmp/merged_patents_export.txt' DELIMITER AS E'\t' HEADER NULL AS  CSV;
--COPY 1646225

Then copy CSV file into table:

  \COPY patents_merged FROM '/tmp/merged_patents_export.txt' DELIMITER AS E'\t' HEADER NULL AS  CSV;
  -- RESULT : COPY 1646225

where "patents" is the name of the table containing the data from the RDP update and "patents_merged" is the name of the table we want to insert the new data into.

Table that we want to insert in patent is called "patent". So you would replace "patents_merge" with "patent".

The data in columns nber, uspc, and uspc_sub is taken from historicalpatentdata. Here is the SQL code to do that after the new data from the RDP has been inserted into the table "patent".

UPDATE patent p
 SET nber = hp.nber,
 uspc = hp.uspc,
 uspc_sub = hp.uspc
  FROM historicalpatentdata hp
  WHERE hp.patentnumber = CAST(p.patent AS varchar);

Missing information (represented by NULL): gyear (int) appear (int)

Schema of patent data when loaded into RDP (based on existin documentation - have not checked schema on RDP postgres server):

patentnumber                  | int | not null
kind                          | varchar    |
grantdate                     | date    |
type                          | varchar    |
applicationnumber             | varchar    |
filingdate                    | date    |
prioritydate                  | date    |
prioritycountry               | varchar    |
prioritypatentnumber          | varchar    |
ussubclass                    | varchar    |
maingroup                     | varchar    |
subgroup                      | varchar    |
cpcsubclass                   | varchar    |
cpcmaingroup                  | varchar    |
cpcsubgroup                   | varchar    |
classificationnationalcountry | varchar    |
classificationnationalclass   | varchar    |
title                         | varchar    |
numberofclaims                | int |
primaryexaminerfirstname      | varchar    |
primaryexaminerlastname       | varchar    |
primaryexaminerdepartment     | varchar    |
pctpatentnumber               | varchar    |
filename                      | varchar    |

Schema of patent table in patent (called "patent"):

patent               | integer           | not null  | plain    |              | 
gdate                | date              |           | plain    |              | 
prioritydate         | date              |           | plain    |              | 
prioritycountry      | character varying |           | extended |              | 
prioritypatentnumber | character varying |           | extended |              | 
cpcsubgroup          | character varying |           | extended |              | 
pctpatentnumber      | character varying |           | extended |              | 
claims               | integer           |           | plain    |              | 
appnum               | integer           |           | plain    |              | 
gyear                | integer           |           | plain    |              | 
appdate              | date              |           | plain    |              | 
appyear              | integer           |           | plain    |              | 
nber                 | integer           |           | plain    |              | 
uspc                 | character varying |           | extended |              | 
uspc_sub             | character varying |           | extended |              | 

Indexes:

   "patents_pkey" PRIMARY KEY, btree (patent)
   "patent_idx" UNIQUE, btree (patent)

For the table called "citation"

Schema of citations table in patent

                      Table "public.citation"
 Column |  Type  | Modifiers | Storage | Stats target | Description 
--------+--------+-----------+---------+--------------+-------------
 patent | bigint |           | plain   |              | 
 cited  | bigint |           | plain   |              | 

We just need citingpatentnumber and citedpatentnumber where citedpatentnumber is an seven digit int from the RDP database.

Use this script to replace textual patent numbers as ints (from existing documentation)

CREATE OR REPLACE FUNCTION cleanpatno (text) RETURNS bigint AS $$ if ($_[0]) { my $var=$_[0]; if ($var=~/^\d*$/) {return $var;} return undef; } return undef; $$ LANGUAGE plperl;

INSERT statement

INSERT INTO citation ( SELECT new_citations.citingpatentnumber,  new_citations.citedpatent number FROM patents );

Equivalent COPY statement:

\COPY citations (SELECT new_citations.citingpatentnumber,  new_citations.citedpatent number FROM patents) TO  '/tmp/citations_merge_export.txt' DELIMITER AS E'\t' HEADER NULL AS  CSV;
--COPY 1646225

Then copy CSV file into table:

 \COPY citations_merged FROM '/tmp/citations_merge_export.txt' DELIMITER AS E'\t' HEADER NULL AS CSV; --1607724

where "citations" is the name of the table containing the data from the RDP update and "citations_merge" is the name of the table we want to insert the new data into.

For the USPTO Assignment Data

Downloading the data

For USPTO Assignment Data, there is a script, under McNair/usptoAssignment, called USPTO_Assignee_Download, which lets a user pass it a text file (file ending in.txt) which contains the url(s) of the assignment data that needs to be downloaded. The script then downloads all the zip files available at that URL. An example called BaseUrls.txt (containing the url that you will probably be using to download the assignment data, unless you're downloading the data from this currrent year, which is in a different link) can be found in McNair/usptoAssignment It then places the downloaded zip files in "E:/McNair/usptoAssigneeData/name", where "name" is the name of the file. If you want to check which files have already been processed, check "McNair/usptoAssigneeData/Finished" to see the finished zip files. (In the future, this should be updated, if possible to specify which years to download, since all assignment data that is not from this current year is under one url, and we've already downloaded most of it.)

Then, to parse the actual files, do the following:

For the USPTO Assignment Data, the parsing file is called USPTO_Assignee_XML_parser. It takes the path to the files that need to be parsed (an example would be ":E/McNair/Files_to_process" where "Files_to_process" is the name of the folder you've placed the xml files to be parsed. It iterates through all the files in the "year" directory that you passed. This file directly loads the information into the database while it parses the file.

For the USPTO Maintenance Fee Data

Download the file manually from https://bulkdata.uspto.gov/data2/patent/maintenancefee/and replace existing file in McNair/Patent Data/Maintenance Fee Data.

For the USPTO Historical Patent Data

Has not been updated since 2015, but if you need to download Historical Patent data this is the link: https://bulkdata.uspto.gov/data2/patent/historical/2014/

Specifications of USPTO Data To Extract

Go to https://bulkdata.uspto.gov/ to bulk data from USPTO.

To see a description of what each file the USPTO bulk data contains, go to the bulk drive and navigate to McNair/Projects/Redesigning Patent Database/2017BulkDataProductDescriptions. This gives an overview, but does not explain how the XML files are structured. Those are the DTDs.

For assignment data, we pull from https://bulkdata.uspto.gov/data2/patent/assignment/. A description of all the elements in an USPTO assignment XML file is called a DTD, and the one for the assignment data can be found in the bulk drive under McNair/Project/Redesigning Patent Database/USPTO Assignment DTD

I'm currently looking at the DTDs for USPTO patent data (2005 and up DTDs have to be opened by Microsoft Visual Studio) to ascertain if there are any fields we're not currently pulling from the bulk data that we should. I am using the following link to figure out how to read a DTD: http://www.ldodds.com/delta/dtd_guide.html. There are only DTDs for 2005 and up, but there is a very long pdf that appears to detail the format of patent files pre - 2005. It has been saved under McNair/Projects/Redesigning Patent Database/

Pre - 2005

According to the documentation, the inventor's name and city are required to be listed, and the street, state, and country of the inventor may also be listed.

Paragraphs of abstract may also be listed under logical group "Abstract" in a field called "abstract":

2005 - present

The USPTO patent data for 2005 seems to include multiple paragraphs for the abstract under an element called "abstract". I've included the line from the DTD below:

<!ELEMENT abstract (doc-page+ | (abst-problem , abst-solution) | p+)>

<!ATTLIST abstract id ID #IMPLIED lang CDATA #IMPLIED status CDATA #IMPLIED >

An abstract is required to be included for all patents that are not design patents, and according to the DTD must have at least one paragraph element.

As far as inventors go, it looks like for 2005 - up patents, the inventor is simply the applicant. this is copied from the DTD:

<!ELEMENT applicant (addressbook+ , nationality , residence , us-rights* , designated-states? , designated-states-as-inventor?)>

<!ATTLIST applicant sequence CDATA #REQUIRED app-type (applicant | applicant-inventor ) #REQUIRED designation (all | all-except-us | us-only | as-indicated ) #REQUIRED >

If for some reason the inventor could not be the applicant, this element would contain the inventor:

<!ELEMENT inventor (addressbook+ , designated-states?)>

<!ATTLIST inventor sequence CDATA #REQUIRED designation (all | all-except-us | us-only | as-indicated ) #IMPLIED >

<!ELEMENT inventors (inventor | deceased-inventor)+>

It seems that we have already pulled patent abstracts before with the existing scripts. I found a huge word document with abstracts under McNair/Software/Scripts/Patent/Abstracts_2016.

Log:

2/16/2017 - Talked over project with Ed, began reading existing wiki pages related to patent data and databases

2/21/2017 - Brushed up on SQL, Entity - Relationship model of designing databases

• In the documentation, I want to briefly explain what the entity-relationship model is before including

the diagram so that readers have a little bit of background

• Found a tool for creating a visual representation called ERDPlus.com - create a standalone instead of an account, can download

Learning commands from Patent Data - SQL Steps

• copy command is PostgreSQL that copies a SQL table to a text file
  • DELIMITER set what will separate columns in text file
  • HEADER specifies that there will be a header in the text file with the names of the columns
• Definitely need to include more detail about what these do in the documentation
• insert into command inserts a new entry into the table

2/23/2017 - Read great database design article, dug through some more wiki articles, started reviewing Perl

• What client do we use to interact with the current patent database?
• Will need to determine all the fields that need to be included in the database before finishing the design and ER diagram, will need Ed's input

3/2/2017 - Started compiling a list of what fields to include and how they would be related.

• Created an Excel spreadsheet that records the each table, their current attributes in the existing patentdata table, what I think the attributes should be in the new table, their relationship to a patent (i.e. one-to-many, many-to-many, etc.), their primary key, questions I have relating to the table, future steps for cleaning up the data in the table (i.e., once all the data has been move to the new database, removing patents that are not US-based), and current problems that have been recorded with the existing table for that information (if an existing table exists)
• Once Excel spreadsheet is completed (and questions in the Questions column are answered or removed from the spreadsheet entirely) I will look into trying to embed it on my "Redesigning Patent Database" wiki page so that future users can sort of follow my thought process. I will also create separate wiki pages to explain each table once the new database is created
  • instructions for adding a table to a wiki page https://www.mediawiki.org/wiki/Help:Tables
• Making spreadsheet led me to realize there is some data that is repeated (filedate in fee table when it is also located in patent table, and the fee table includes the patent

3/7/2017 - Continued working on spreadsheet, update project page with relevant links, developed following plan for documentation

• Most important thing for new documentation in my opinion is a clear flow, i.e., one central page on the database, which contains links to the following:
  • Where data is (patent data) where it came from (Harvard Dataverse, etc), the scripts/SQL used to obtain the data, and where original xml files can be located
  • Next, there should be an ER diagram that shows the general design, followed by short descriptions of each table is, what it's related to, and what it's use is
  • Each table should contain a link to a separate wiki page on the table which will explain all the columns, where the data came from, provide a visual example of data in the table, list issues, and also contain a log of what has been done to the table over time
  • there should also be instructions on the page for how, if someone is going to change the database (add a table, alter a column, etc.) which documentation they should update (the wikipage for the table, for example)
  • Links to USPTO Assignee Database with an explanation of how it is joined with the Patent Database
  • Lastly, everything should be dated. I know that each wiki-page lists when it was last edited, but the main page, next to table descriptions, should have the date the table was originally added. The ER diagram should be updated as needed and also dated so that readers know if it's the most current representation of the database.

3/9/2017 - Finished up first draft of spreadsheet, found ER diagram tools

• Old ER diagram (see Documentation Relevant to Current Databases) is difficult to read - it appears to show tables that don't exist and data that doesn't exist in the current schema
• psql allpatent does not work. psql allpatent_clone does for some reason

3/21/2017 - Working on determining "core" tablse that will always be in database and understanding work on assignees info

• http://postgresonline.com/special_feature.php?sf_name=postgresql83_psql_cheatsheet&outputformat=html
  • helpful commands for interacting with psql
• Notes on determining "core" tables for database are saved under Projects/Redesigning Patent Database/Determining which tables should be core tables

3/23/2017 - Made spreadsheet with core tables and share it with other people working on patent data to get their approval

• This spreadsheet can be found under Projects/Redesigning Patent Database/Core Tables for New Patent Database
• Some tables that will later be deleted were included on the spreadsheet because their are currently being tables built to replace them
• May try to just move all the (twenty-something) "pto-" tables that have been created due to the "Restructuring Patent Data" project from "patent" to the new database
• Will work on understanding SQL for filling new database from this link next week http://mcnair.bakerinstitute.org/wiki/Patent_Data_Processing_-_SQL_Steps and http://mcnair.bakerinstitute.org/wiki/Patent_Data_Cleanup_-_June_2016

4/4/2017 - Found all the pages on extracting data and making tables and databases

• Do not need to pull Harvard Dataverse data again, it's saved in CSV files on the bulk drive
• Started looking through DTDs for USPTO patent and assignment data to determine if there is extra information that we should extract from USPTO data.