A small server/worker you can run to turn Use of Force Report PDFs into structured data for anaylsis. Uses machine learning via AWS Textract.
We imagine two future steps for this tool. 1) Improving the flexibility in analyzing use of force reports that differ drastically in format. 2) Allowing the analysis of police union contracts to find ways that block police officers from being held accountable.
If you're interested in helping, please send a message to: [email protected]
This can be run without any hosting on an external site/server but it will require intermediate programming experience.
This tool relies on AWS S3 + AWS Textract. Textract is free up to 1,000 pages of analysis. Because there is rate limiting, this setup requires a database to save the jobs and run them at delayed intervals.
Once you have your AWS credentials, create a file called .env with the following consts:
AWS_REGION=
AWS_S3_BUCKET=
AWS_ACCESS_KEY_ID=
AWS_SECRET_ACCESS_KEY=
DATABASE_CHARSET=utf8
DATABASE_HOST=postgres
DATABASE_NAME=foilthepolice
DATABASE_PORT=5432
DATABASE_USER=foilthepolice
DATABASE_USER_PASSWORD=foilthepolice
PORT=3000
The database credentials mirror the postgres conifg in docker-compose.yml which is only used for local development.
To run, have docker installed, your credentials saved, and simply type in your root directory docker-compose up. You should see a message saying that a server is running on port 3000. There is no interface for tool at the moment. To process documents and retrieve results, you'll want to be aware of the three endpoints. You can submit information to these with a tool like Postman.
To analyze a PDF for structured data (ex: useful on a packet of use of force reports), use:
POST: localhost:3000/v1/analyze/form
To analyze a PDF and get back plain text (ex: useful for police contracts), use:
POST: localhost:3000/v1/analyze/text
Request POST Body:
{
pdf: [form-data] // PDF file submitted as form-data
}Response Data:
{
"jobId": 96,
"success": true
}When you get back your job id, you know the document is being processed in the background by AWS. You can now make a request for it using this GET endpoint which will return when it's been finished by AWS. For big PDFs and structured data, you might be waiting a few minutes.
GET: localhost:3000/v1/analysis/{jobId}
Form Job Response Data:
{
"data": [
{
"date": "12/26/15",
"time": "23:07",
"day_of_week": "Saturday",
"incident_number": "15-12345",
"location": "Whippany, NJ",
"officer_badge_number": "123",
"officer_name": "Bob Woodsman",
"officer_race": "Cauc",
"officer_sex": "M",
"officer_age": "34",
"officer_rank": "Patrolman",
"officer_on_duty": "Yes",
"officer_uniform": "Yes",
"officer_assignment": "Patrol Division",
"officer_years_of_service": "12",
"officer_injured": "No",
"officer_killed": "No",
"subject_race": "Asian",
"subject_sex": "",
"subject_age": "34",
"subject_under_influence": "",
"subject_unusual_conduct": "",
"subject_injured": "",
"subject_killed": "",
"subject_arrested": "",
"subject_charges": "",
"subject_actions_resisted_officer": "",
"subject_actions_threat_attack_physical": "",
"subject_actions_threat_attack_blunt": "",
"subject_actions_threat_attack_knife": "",
"subject_actions_threat_attack_vehicle": "",
"subject_actions_threat_attack_firearm": "",
"subject_actions_fired": "",
"subject_actions_other": "",
"incident_type_crime_in_progress": "",
"incident_type_domestic": "",
"incident_type_other_dispute": "",
"incident_type_suspicious_person": "",
"incident_type_traffic_stop": "X",
"incident_type_other_type_of_call": "",
"officer_force_used_compliance_hold": "X",
"officer_force_used_hands_fists": "",
"officer_force_used_kicks_feet": "",
"officer_force_used_chemical_agent": "",
"officer_force_used_blunt_strike": "",
"officer_force_used_canine": "",
"officer_force_used_firearm_intentional": "",
"officer_force_used_firearm_accidental": "",
"officer_force_used_firearm_shots_fired": "",
"officer_force_used_firearm_shots_hit": "",
"officer_force_used_other": ""
},
],
"status": "done",
"success": true
}Text Job Response Data:
{
"data": "Chapter 29: Law Enforcement Article 14: Pea........ e Peace Officer's Employer- Employee Relations Act [29-14-1 NMSA 1978].",
"status": "done",
"success": true
}You can also download jobs as CSV or plain text files using the format query parameter with the values csv or text
GET: localhost:3000/v1/analysis/{jobId}?format=csv
date,time,day_of_week,incident_number,location,officer_badge_number,officer_name,officer_race,officer_sex,officer_age,officer_rank,officer_on_duty,officer_uniform,officer_assignment,officer_years_of_service,officer_injured,officer_killed,subject_race,subject_sex,subject_age,subject_under_influence,subject_unusual_conduct,subject_injured,subject_killed,subject_arrested,subject_charges,subject_actions_resisted_officer,subject_actions_threat_attack_physical,subject_actions_threat_attack_blunt,subject_actions_threat_attack_knife,subject_actions_threat_attack_vehicle,subject_actions_threat_attack_firearm,subject_actions_fired,subject_actions_other,incident_type_crime_in_progress,
"12/26/15","23:07","Saturday","15-12345","Whippany NJ","123","Bob Woodsman 764","Cauc","M","34","Patrolman","Yes","Yes","Patrol Division","12","No","No","Cauc","M","19","X","","No","No","","","X","","","","","","","","","X","","","","","X","","","","","","","","","0",""
...