#Twitter4Uni

Project presentation for #Twitter4Uni, 2015

Transcript

1. #Twitter4Uni - Stage proposal Adriano Di Luzio & Danilo Francati May 2015 1
2. 2 Before we start, a few things about us… We’re 23 years old Sapienza University of Rome students, 
 Computer Science master degree,
 Networks and Security curriculum. aldur d392 aldurd392
3. Crimes, shoplifting, violence, pickpocketing. Car crashes, bad parking, infractions. Public transports, infrastructures… … and yes, lot of other things. 3 We all know that big cities lead to big problems.
4. Problems of this kind are hard to be dealt with, monitored or solved. But we live in a new era, filled with technological and social interactions. 
 … and we have Twitter on our side. 4
5. So, why don’t we let people play a front-line role? Hey, I’ve just seen something wrong! I’d like to help! 5
6. Tweets are short, immediate and descriptive. In a matter of seconds, anyone could signal the problems she sees around her. With an appropriate system, we could even solve these problems. 6
7. This leads us to our first idea.
 Creating, through Twitter, a direct communication link between 
 citizens and authorities. 7
8. 8 Let’s take Rome as an example. @ComuneRomaCapitale I’d like to report a crime… Hey, there’s a flood here! I just witnessed a pickpocketing! * Yes, those are all real-life examples.
9. We want to build an automated system that parses and extracts informations from all the tweets sent by citizens to a specific account or containing a peculiar hashtag. We’ll then forward all the relevant informations to the
 competent authorities:
 Fast response and reaction times. Endless possibilities. 9
10. 10 Could we have more geeky details, please? Of course! Tweets Natural language details GPS positions Hashtags & Media Information retrieval engine Filter and retrieve informations from tweets Authorities reaction Different triggers for different services
11. 11 Tweets Information retrieval engine Authorities reaction Each tweet contains:
 Natural language details. Hashtags, media and GPS coordinates. We’ll feed those informations to our engine. The core of our system. It will clusterize incoming tweets, filter them and extract all the relevant informations by processing their content and metadata. We’ll employ here Machine Learning algorithms and, if needed, Big Data computing techniques. We’ll then trigger specific events to each competent authority. After processing data, we’ll alert the authorities. We’ll provide different kind of triggers, priorities and labelling. All about flexibility and fast responses.
12. Other similar systems already exists… That’s only partially true. Similar initiatives are starting all around us. (Florence, as an example.)
 But our system aims to be far better. We’ll process data automatically, without needing any human interaction.
 We’ll let users and authorities collaborate directly, 
 but we’ll filter noise and fake alarms. And there’s more… 12
13. Here it comes our second idea. 
 In fact, once we have our Information Retrieval Engine, why should we stop there? We could configure it for many different tasks. e.g. we could employ it to help organise the logistic behind large events, such as EXPO Milan 2015. 13
14. 14 Every day, hundreds of thousands of tweets regard EXPO 2015.
15. 15 Feeding those tweets to our system 
 we could predict EXPO’s daily affluence and accordingly adapt the required infrastructure: adding more train carriages, scheduling more flights, preventing and monitoring highways traffic jams, preparing more food, scheduling sales and offers, and so on… +
16. And, eventually, our final idea. We want to build a modular engine.
 By exploiting the pipeline model you have seen before,
 we want to detect events/anomalies and to react accordingly, by triggering specific alerts. We want it to be configurable, so to be as flexible as possible, and to provide really unlimited use cases. We want to build it as part of the future. 16
17. Modular engine 17 Tweets Natural language contents GPS positions Hashtags & Media Core engine Semantic information retrieval Event/anomalies detection Statistics and monitoring Triggers Every use case will come with its own trigger. e.g. alerting the authorities
18. Everyone will be able to use our system 
 through ad-hoc plugins that will interact with the core: The intelligent brain of the whole. 
 (Machine Learning powered) 18
19. 19 We have shown you only many of the
 interesting applications we have in mind.
 The possible usages are really endless. 
 Thanks for your attention. Image credits: freepik.com & Michael Filippoff