A Predictive Policing and Threat Aggregation toolset. This modular toolset, powered by Natural Language Processing and Open Source Intelligence, is designed to provide the user with actionable data including: signals, pivots, and risk levels. In collecting this information all that is required from the user is a 'source'(e.g. Twitter) and 'username'(e.g. Jane Doe). Current source collectors include: Twitter, Instagram, Tumblr, and Reddit.
- Collectors - Collect text (e.g. comments, tweets, posts, etc) from a source (e.g. twitter, or instagram) and a user (e.g. Jane Doe).
- Analysers - Apply Natural Language Processing to the collected text (for example entity recognition and sentiment analysis), which is then used by the methodologies.
- Methodologies - Apply frameworks to the collected text (for example has a location been mentioned and is a location known for a specific type of crime)
For more on how Natural Language Processing can be used to predict crime, check out this article and a talk.
If you have a specific use case for Hunch and would like help in it's implementation then please reach out to James Stevenson.
Hunch uses Python 3.x and supports both Windows 10 and Ubuntu 18.x - However, other versions and operating systems may work they simply have not been tested. An installation video can be found on Youtube.
To use the AWS Comprehend aggregation for Hunch you will need to have the AWS CLI installed. As part of setting up the AWS CLI you will need to create an AWS IAM user with the "ComprehendFullAccess" group and set this as your user when it comes to the configure stage.
pip install awscli --upgrade --useraws configureThe Reddit aggregation requires the Reddit PRAW api to be installed. You will also need a Reddit developed application and use the keys created as part of that process. The API keys will need to be added to the Core_Config.Json.
The Twitter collector requires the Tweepy module and requires Twitter App API keys. The API keys will need to be added to the Core_Config.Json.
The core display requires the Dash python plugin to be installed. The display also uses Pandas to display tables - you may also need Microsoft Windows SDK to install Pandas.
Location detection requires the GeoPy plugin that can be installed with the command below. The Location Detection also uses the UK Police Crime API. This API does not currently require an API key.
The URL Recognition script requires the Requests plugin to unshorten url short links.
pip install -r requirements.txtThe Collectors that do not use an API will most commonly use Curl, if you are on a Linux system you will need to have curl installed.
sudo apt-get install curlThere is an example Python script called Example_Profile_Script.py that contains the necessary code to profile all individuals in the JSON FILES folder as well as launching the display interface on localhost. A simplified version of this can be found below:
import Core_Collector
import Core_Display
import Core_Individual
my_collector = Core_Collector.WebsiteToCrawl([{"twitter":"realdonaldtrump"}],"Donald J. Trump")
my_individual = Core_Individual.Individual(my_collector.aggregate_data(), my_collector.name)
Core_Display.create_website([my_individual.profile()]).generate_page()In addition to this you can use the Lazy tool which cuts down on the code needed to profile individuals, however, loses some of the additional functionality.
import Lazy
list_to_profile = [
[[{"twitter":"realdonaldtrump"}],"Donald J. Trump", 1],
[[{"twitter":"AOC"}],"Alexandria Ocasio-Cortez", 1],
[[{"reddit":"user_simulator"}],"user_simulator", 1]
]
lazy_profle = Lazy.lazy_profile()
for individual in list_to_profile:
lazy_profle.profile(individual[0],individual[1], individual[2])
lazy_profle.display_webpage()Pull requests are welcome, including creating new detectors, collectors, and analysers. For major changes, please open an issue first to discuss what you would like to change.