On the Live tab, you can find out the current traffic status in Buenos Aires (Argentina).
The map shows every detected traffic incident as a map marker or as a colored line between two markers.
Each map marker indicates a different type of traffic incident:
Traffic Jam and
Works on Road.
At the top of the right-side white bar, you will find a refresh button.
Use this button to update traffic status.
Moreover, you can click on a map marker to visualize on the right-side bar which tweets reported the corresponding incident.
On the Demo tab, you can understand the analysis process of Manwë. Here, you can write down a "fake" tweet in Spanish language and observe how Manwë analyzes it.
After you write the tweet, just press the Analyze button
and Manwë will analyze your "fake" tweet with the aim of detecting traffic incident reports.
Remember: this demo is only configured to detect traffic incidents located in the city of Buenos Aires.
Attention: This may take from the order of a few seconds to one minute depending on server current load and the server load of third-party geocoding services. Please, be patient
Once the analisis ends, you can see the results of the five Manwë steps:
Preprocessing: Simplifies the text by removing unnecessary text elements.
Tokenization: Separates the text into tokens. On the demo, each token is enclosed by brackets.
NER: Recognizes named entities and traffic-related Spanish words on tokens.
On the demo, if the NER recognizes something on an specific token, this token will be shown in blue color.
If you hover the mouse pointer over a blue token, you will see what named entities or Spanish words were recognized.
Each recognition is detailed by the name of the entity/word, the similarity score between the entity/word and the token [0;+1], and the entity/word category (if any).
Relates the tokens by taking into account which named entities or Spanish words where recognized on them.
This step attempts to detect textual locations and the type of traffic incidents.
Geocoding: Geocodes traffic incidents locations detected in the analysis. Each geocoded incident is detailed by a textual description.
If you click on an incident description, the incident will be centered on the minimap of the bottom of the page.
1 / 6
Manwë is a prototype tool designed to detect traffic incidents by analyzing tweets in real-time.
In this page, you will be able see a live demo of Manwë. However, let us first briefly explain how it works...
2 / 6
Twitter users continuously tweet or comment diverse topics in this microblogging site...
3 / 6
...some of those tweets may report recent traffic incidents
4 / 6
Manwë filters traffic-related tweets from the public Twitter stream by using Machine Learning techniques
5 / 6
Afterwards, Manwë applies NLP techniques to detect traffic incidents from filtered tweets. Consequently, Manwë geocodes the incidents locations by means of a third-party geocoding service.
6 / 6
Finally, a map shows the current traffic status in the target city.