Copyright (c) 2020 Ripple Software. All rights reserved.
This program is free software; you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation; version 2 of the License.
This program is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details.
You should have received a copy of the GNU General Public License along with this program; if not, write to the Free Software Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
Author: Joseph Lee
Email: [email protected]
Website: https://www.ripplesoftware.ca
Github Repository: https://github.com/rippledj/uspto
This python script is based on a project from University of Illinois (https://abel.lis.illinois.edu/UPDC/Downloads.html). Several parts of the script have been improved to increase the data integrity and performance of the original script. The script has been tested with Python 3.6. It should work with Python 3.6 through 3.8, but will not work with any versions of Python 2 or Python 3.9 or higher.
The script is run from the command line and will populate a PostgreSQL or MySQL database with the USPTO patent grant and patent application red-book bulk-data. It is recommended to use PostgreSQL since PG provides better performance over the large data-set.
The usage of the script is outlined below:
There are three steps.
- Install the required database
- Run the parser USPTOparser.py
- Schedule the updater
Run the appropriate database creation scripts depending if you intend to store the USPTO data in MySQL or PostgreSQL. The script will create a user 'uspto' and limit the scope of the user to the uspto database. If you want to change the default password for the user, edit the appropriate .sql file before running it. Also, some configuration of your database maybe necessary depending on the settings choose when running the script. For example the ability to bulk insert CSV files are disabled by default in MySQL.
MySQL or MariaDB
installation/uspto_create_database_mysql.sql installation/uspto_create_database_mariadb.sql
PostgreSQL
installation/uspto_create_database_postgresql.sql
Before the USPTOParser.py can run successfully, the database connection and authentication details must be added (if database storage will be specified). Text search for the phrase "# Database args" to find the location where database credentials must be changed. Enter "mysql" or "postgresql" as the database_type. Enter the port of your MySQL or PostgreSQL installation if you have a non-default port. If you changed the default password in the database creation file, then you should also change the password here.
The default setting is to parse USPTO 'Red Book' bibliographic data which results in a database size of about ~70GB. If you want instead to collect full-text, you must pass the argument '-full' when you run USPTOParser.py. The full-text database size is ~2TB.
Also, you must specify the location for the data to be stored. These options are: '-csv' and '-database'. You must include at least one. These arguments tell the script where you want the data to be stored. You should set the 'database_insert_mode' to specify whether you want the data to be inserted into the database after each data object is found and parsed ('each'), or in bulk post parsing of each file ('bulk'). 'bulk' setting greatly improves database performance and reduces the total time to complete the bulk insertion.
Finally, you can set the number of threads with a command line argument '-t [int]' where [int] is a number between 1 and 20. If you do not specify the number of threads, then the default number of threads will be used, which is 5. Using the '-balance' argument will turn on the load balancer which will limit the threads CPU usage. However, if you do not use the '-balance' flag, your computer may crash if your CPU load is too high.
The following example is the command to store in csv file and database with 10 process threads.
$ python USPTOParser.py -csv -database -t 10
The following example is the command to store full-text data in csv file and database with 20 balanced process threads
$ python USPTOParser.py -csv -database -t 20 -full -balance
Finally, the script can be run in 'sandbox mode' or normal mode by setting a flag in the args_array called 'sandbox' which is at the top of the main function. Running the script in sandbox mode will keep all downloaded .zip files and extracted .xml or .dat files on your computer so that they do not need to be downloaded again if you restart the script or encounter any errors, or so that you may inspect the decompressed data files.
The script will keep track of processed files in the LOG directory. There are log files for grants (grant_links.log) and applications (application_links.log), and a main log file USPTO_app.log which keeps track of errors and warnings from the script. If the script crashes for any reason, you can simply start the script again and it will clear any partially processed data and start where it left off. You can set the verbosity of the stdout and USPTO_app.log logs with the 'log_level' and 'stdout_level' variables at the top of the main function.
You should check of the grant_links.log and application_links.log files after the script has completed to make sure that each line in those files says "Processed" at the end. If the file has not been processed, the line will end with "Processed" and you should run the script again to finish any files that were not processed.
The script will also run in update mode to get all new patent grants and applications which are issued each week by the USPTO. This is done by passing the '-update' argument when running the script as the command below.
$ python USPTOParser.py -update
The script will check your previous data destination(s) settings and get any new patent data release files that have been published on the USPTO website (https://bulkdata.uspto.gov/). The new files are then parsed and stored in the destinations you previously specified. Since database data files are released every week, the updater can be scheduled once a week to keep your database up-to-date.
After the USPTOParser.py finishes building the initial databases, the data can be validated, duplicate records can be removed and some metrics can be calculated such as total forward and backward citations. The scripts for each of these processes can be found in the installation directory.
Validate Record Counts
The total grants and applications can be compared to the per year summary issued by the USPTO using a pre-built script. USPTO per year data summary for grants and applications can be found here: https://www.uspto.gov/web/offices/ac/ido/oeip/taf/us_stat.htm
MySQL or MariaDB
installation/verification/data_summary_mysql.sql
PostgreSQL
installation/verification/data_summary_postgresql.sql
Remove Duplicate Records
The USPTO bulk dataset contains some duplicate patent grant and application records. These records are initially inserted into the database and the GRANT and APPLICATION tables which use a combination of the id and source filename as primary keys. Since the duplicate records may be in different source files, they will not violate the primary key restraint and will be inserted into the database. The duplicate patent grant records can be removed from the database using scripts below, but duplicate application ids are left in the database. If you want to remove them, you can modify the script and run it again to remove duplicate application numbers.
MySQL
installation/post_parse/remove_duplicates_mysql.sql
PostgreSQL
installation/post_parse/remove_duplicates_postgesql.sql
Verify Parser Accuracy
The USPTOParser accuracy can be verified by running a pre-built SQL script installation/verification/create_parser_verification_table_mysql.sql and then running python USPTOParser.py -verify. All files must be processed before you start the verification. Otherwise, you will need to delete the PARSER_VERIFICATION table from the database and run the verification again.
The verification process will first build a database table PARSER_VERIFICATION, then compile a count of all records from each data-table / source-file combination into the table 'Count' column. The verify process with then search each source bulk-data file for XML tags and compile an expected count for the number of records that should be in each database table. Then query to show the completeness of all filename / table combinations.
To maintain the source bulk-data files use the -sandbox flag when processing the bulk-data into the database. Otherwise the files will be downloaded again from the USPTO bulk-data site to do the verification process.
Calculate Additional Statistics
The USPTO data contains a wealth of information. The data can be used to calculate patent metrics for statistical analysis. There are some scripts included to calculate some common statistical metrics such as forward and backward citation counts.
Patch Missing Data With Google BigQuery
The USPTO bulk-data is missing some information. This data can be collected from Google BigQuery Patent database. First the credentials for the Google BigQuery API must be downloaded as a JSON file and added to the __init__ function of the PatentBigQuery Class in the USPTOBigQuery.py file. Then the database can be automatically patched by running the command python USPTOParser.py -patch
MariaDB
installation/metrics/uspto_metrics_mariadb.sql
MySQL
installation/metrics/uspto_metrics_mysql.sql
PostgreSQL
installation/metrics/uspto_metrics_postgresql.sql
The script uses a load balancer which detects the number of CPU cores and creates and sleeps threads to maintain a constant CPU load. The default value is 75%. Therefore, if the 10 minute CPU load is less than 75%, another thread is added. If the CPU load exceeds 75%, threads are forced to sleep. These settings can be adjusted in the script.
The method used to insert data into the database can be configured in two ways. The script can insert each document record immediately after it is parsed or in bulk after a file is finished being parsed. Using bulk storage utilizes .csv files to temporarily store the data before it is inserted in bulk. If the '-database' flag is set but the '-csv' is not set, then the .csv. files are erased after being used to load the data.
If you have questions about the USPTO patent data you can contact: Author, Joseph Lee: [email protected] USPTO: [email protected] USPTO Developer: [email protected]