SSS - Stay Safe Selly

Conception

The student community of Selly Oak, Birmingham has faced elevated levels of crime over the past months. As a group we decided a more robust system to view and report crime was needed.

Currently crime is unofficially reported via the university Facebook page 'Fab & Fresh' however, the issue with this is with the volume of other posts the reports can often be lost or not seen. We decided an application that could scan posts on this group, detect crime reports, categorise based on location and present this data in a easy to view format would be beneficial to the community at large.

Method

Scanning data from Facebook was easier said than done in the wake of the Cambridge analytica scandal. We required facebook approval for our app before we could proceed even testing on a private closed group, a process that takes at least 2 days.

In order to get some working code we decided to write our own sample JSON data mimicking that which we would extract from Facebook. With this data we created a basic python script that analysed it for road names in Selly Oak, the tag '#staysafeselly' and detected satirical posts based on the types of reactions to the posts.

We conceptualised several different methods for this:

We would assign each post a "weight", with the weight corresponding the the engagement amount and ratio (Likes / Comments)
- We would modify this weighting depending on the reacts ("haha" reacts decreased weight, "angry"/"sad" reacts increase weight)
- Ignoring satire (Where "haha" reacts make up more than 20% of the total engagement)
Search for key crime words for categorising posts
Time based grouping
- Displaying all posts in the last 6 hours, ordered by post weight

Future Extensions

In order to proceed in our project we would need approval from the Guild of Students who run the 'Fab and Fresh' Facebook group as well as wait on Facebook approval. Once we get permission, we would continue to develop the backend, inferring event type and being able to detect things such as slang names for roads.
We also need to develop the user facing end, i.e. a mobile application connected to a cloud server and database in order to serve the parsed information to the user. Something we began working on at the hackathon, where users will use their facebook account to login, potentially allowing users to post to the facebook group via the app.

The main ways we wish to expand on are:

"Pinning" posts made by either admin or verified sources, such as the newly added West Midlands Police rep on fab n Fresh
Tagging related data together as collections
- Grouping posts based on word/synonym match count, for example a post mentioning the word "knife" and "weapon" or "stabbing" being linked
- Limiting this in a certain timeframe, so instead of linking all posts related to, say, anti-social behaviour, only linking posts mentioning ASB in the last 4 hours.
- Tagging data together if they mention the same or nearby streets
Heatmaps of crime shown on street maps, with heat proportional to the amount of posts and the weight of said posts.
Work with informal slang
- As the platform is facebook, it is likely that including features for mapping posts with informal mentions of road names, such as just "heeley" for Heeley Road or "tiv" for Tiverton Road, would increase the amount of data we can see and group together
- Match abbreviations and slang
Integration of data tagged with #staysafeselly on twitter, instagram, or local news sources and display this on a separate panel
Sending push notifications to users that wish to receive notifications about influxes of information
- For example if a user wanted to know if Heeley Road got any new posts related to crime, getting a push notification once a Heeley Road related post gains traction