Entry Name: LIST-Medoc-MC3

VAST Challenge 2019
Mini-Challenge 3

Team Members:

Hélène Bourgoin, Luxembourg Institute of Science and Technology (LIST), helene.bourgoin@list.lu

Nicolas Médoc, Luxembourg Institute of Science and Technology (LIST), nicolas.medoc@list.lu PRIMARY

Philippe Pinheiro, Luxembourg Institute of Science and Technology (LIST), philippe.pinheiro@list.lu

Mohammad Ghoniem, Luxembourg Institute of Science and Technology (LIST), mohammad.ghoniem@list.lu

Student Team: NO

Tools Used:

- Papyrus, software described in: Nicolas Médoc, Mohammad Ghoniem, and Mohamed Nadif. 2016. Visual exploration of topic variants through a hybrid biclustering approach. In Actes de la 28ième conference francophone sur l'Interaction Homme-Machine (IHM '16). ACM, New York, NY, USA, 103-114. DOI: https://doi.org/10.1145/3004107.3004116

- Excel

Approximately how many hours were spent working on this submission in total?

80 hours

May we post your submission in the Visual Analytics Benchmark Repository after VAST Challenge 2019 is complete? YES

Questions

The City has been using Y*INT to communicate with its citizens, even post-earthquake. However, City officials needs additional information to determine the best way to allocate emergency resources across all neighborhoods of St. Himark. Your task, using your visual analytics on the community Y*INT data, is to determine the types of problems that are occurring across the St. Himark. Then, advise the City on how to prioritize the distribution of resources. Keep in mind that not all sources on Y*INT are reliable, and that priorities may change over time as the state of neighborhoods also changes.

1 – Using visual analytics, characterize conditions across the city and recommend how resources should be allocated at 5 hours and 30 hours after the earthquake. Include evidence from the data to support these recommendations. Consider how to allocate resources such as road crews, sewer repair crews, power, and rescue teams. Limit your response to 1000 words and 12 images.

To answer this question we loaded all messages from April 8 between 08:35AM and 1:35PM in the Papyrus software.

In Fig. 1, the Topic Overview tab shows the most important terms as a word cloud and a ranked list based on the TF-IDF score.

The ranked list suggests that “power”, “water” and “bridge” are the most interesting terms during this time period, which seems well aligned with the resource allocation concerns.

To learn more about the situation, we studied other terms related to these three terms.

Fig1 The topic overview tab

Here is our summary of the situation 5 hours and 30 minutes after the catastrophe

In terms of power:

In Fig2, the “Topic details” tab shows a sunburst view of hierarchized term associations extracted from the text of the messages, where each branch stands for a term association.

By clicking on any branch in the sunburst view, we list all related messages on the right.

In this case, the term tree is filtered to only retain branches including the term “power”.

In Fig2, power lines seem to be falling causing power outages.

Other messages describe specific situations (not shown in Fig2), e.g. the GOLDCARE HOSPITAL is out of power. “The hospital by the neighborhood's power is out! Terrapin Springs” 11:39AM

A few gas stations have power and the lines are being repaired. “People are working on the lines! Please let there be power soon!” 1:00PM “The gas station by the fire station has power and can pump gas. #fireWonder” 10:10AM

Fig2 tree of terms co-occurring with “power”; showing messages having both “power” and “line”

In terms of water:

Six neighbourhoods have sewage issues; water might be contaminated in Old Town, Safe Town, Scenic Vista, Broadview, Chapparal and Easton (Fig3).

A broken sewer pipe is also filling Old Town neighbourhood with sewage (Fig4).

Fig3 Whole term tree, showing messages for term association “water, broken, neighbourhood, pipe, sewer, drinking, risk”

Fig4 Term tree filtered for “sewer”

In terms of roads:

There is a 20 feet deep sinkhole in Jade Avenue in Oak Willow in which several cars are trapped (Fig5). North Street has several cracks, streets in Old Town and Scenic Vista are filling with rubble and some streets are flooded, which makes them impracticable.

“Lot of cracks on North St going from Old Town to Northwest!” 12:43PM

“Street outside my house is flooded! Scenic Vista” 11:34AM

“The brides and the roads are unsafe. Please don't try to leave your neighborhood.” 08:45AM

Fig5 Term tree filtered for “people”

In terms of bridges:

-Magritte Bridge (A) reopened at 1:23PM with only one lane, before that only emergency vehicles could use it (Fig6).

-Friday Bridge (B) opened shortly between 10:05 and 10:46, but it closed again (Fig7).

-Jade Bridge (C) has a large crack and is still closed (Fig8).

-12th of July Bridge (D) has collapsed and is closed (Fig9).

-Himark Bridge (E) is closed

-The Wilson Forest Highway (F) reopened at 10:46.

“Dept of Transportation: As a precaution we have closed all bridges and the forest viaduct for inspection.” 9:00 AM

“The Wilson Forest Highway has is re-opened. Please be courteous in your use.” 10:46 AM

In terms of infrastructure:

Several buildings have collapsed and some people are still trapped (Fig10). Fires have started at numerous places in town.

“Extensive damage on the north and south sides. Although no neighborhood has escaped damage. Severals fires throughout City.” 09:30AM

Fig10 Whole term tree, showing messages for term association “fire”

In terms of communications:

The massive download of the rumble app is helping the city distribute resources (Fig11).

However, the global system is starting to overload due to people uploading videos and calling the emergency number for non-life threatening situations.

“City Emergency Operations Center asks that everyone stay off their phones. You are overloading the system. Stop uploading videos. Also stop calling the emergency number to report the earthquake. We know. Call if you have a life threatening emergency.” 12:30PM

Fig11 Messages about the city EOC

In terms of injuries:

Most hospitals are full of rubble which is hindering treatments (hospitals in: Palace Hills, Southton, Broadview, Terrapin Springs and Old Town) so only two hospital are working normally (Fig12).

The water in Southton hospital might also be contaminated, thus the neonatal unit patients are transferred elsewhere.

An aid station for non-critical patient has been installed in Hippocrates Medical Clinic parking lot (Fig12).

“Hospital has rubble! EEEEK!!” 1:09PM

Fig12 Messages about the state of the hospital

In this situation, how should we allocate resources?

Road crews: The first priority is to ease traffic to allow rescue teams to navigate freely around town. To achieve this, the sinkhole in Jade Avenue must be repaired and crews must focus on reopening the less damaged bridges, such as the Friday Bridge and the Himark Bridge, to let people evacuate the city if they wish. Next, they have to clear rubble from the streets and fill the cracks on North Street. Only then they might focus on the most damaged bridges (Jade Bridge and 12th of July Bridge).

Sewer repair crews: Sewage must be kept off the street, by first fixing the broken sewer pipe in Old Town and then working in the 5 other contaminated neighborhoods.

Power: First, power must be brought back to the Goldcare hospital. Then the tilted power lines must be redressed to restore power across the city.

Rescue teams: The firefighters must put out all fires throughout the city. After that, they will be able to save the people trapped under the rubble. Volunteer rescue teams can give first aid to people and help clear out the street.

2 – Identify at least 3 times when conditions change in a way that warrants a re-allocation of city resources. What were the conditions before and after the inflection point? What locations were affected? Which resources are involved? Limit your response to 1000 words and 10 images.

The first inflection point is certainly the earthquake on April 8 at 8:35. The conditions, locations and resources have already been discussed in the previous question.

To spot the other inflection points we look for peaks in the rate of messages over time by plotting them in Excel (Fig13).

We see immediately various peaks.

We ignore the peaks before 8:35AM on April 8 as they occur before the big earthquake since they do not require important re-allocation of resources.

The first peak corresponds to the earthquake on April 8 8:35AM. The following peak is on the same day at 2:36PM when the city strives to rescue people. The last one is the aftershock on April 9 at 3:05PM.

To analyze the second and third peaks, we take a similar approach to question 1, by loading messages occurring in a 6-hour window centered on the peaks.

We suspect that the problems will be similar to the first case, concerning utility and infrastructure networks.

Fig13 Number of messages per minute

Point of inflexion #2: Finding new way to rescue people/ The aftermath of the disaster

In terms of power outage, contamination of water and state of the bridges, the situation did not change.

In terms of roads and infrastructures

Once the shock of the earthquake is over the state of the roads is worsening and people are starting to blame city officials.

“Bricks are scattered all over the roads. Why didn't the city work on getting those buildings up to code sooner?” Scenic Vista. 3:46PM

“the really pretty old church across from my house has a giant hole in it now. government protect the historical infrastructure challenge” West Parton. 2:03PM

More roads crew resources need to be sent to neighborhoods like old town, where people feel abandoned, and have lots of rubble (see Fig14)

As a lot of building collapsed, elephants from the circus are sent to help clear rubble in order to get access to trapped people (see Fig15). This is the main factor of inflexion here as it allows rescue teams to be dispatched elsewhere for example in Old Town.

Fires are now under control which frees firefighter to go give first aid or clear out rubble.

“Fire Command reports all major fires are under control. All citizens are asked to be cautious with candles during the power outages.” 2:00PM

Fig14 The state of Old Town

Fig15 Messages about the elephants

In terms of communication

We saw in the previous question that the system was already starting to overload before the inflexion point. As the day is advancing, people are more and more annoyed when they don’t get a clear answer from the city officials, rumors start to inflate (see Fig16 and example underneath) and people in real emergencies cannot reach the emergency number (see Fig17).

“My boss heard from their children that the city's evacuating. #WinsevacuatingWaste” 2:29PM

This deterioration of events means that more resources must be allocated to emergency center services and city officials must communicate more on social media to avoid calls from people in non-emergency situations.

Fig16 Rumors on the number of fatalities

Fig17 People can’t contact the emergencies

In terms of injuries

We saw in the previous question that before the inflection point the hospitals were already crowded. While people are getting out of the collapsed buildings, there are also many people with minor injuries in the streets heading to the hospital.

Blood donation centers are set up in several neighborhoods as hospital blood reserves are depleted.

“We have one hospital that is functional right now. Roads are damage and the situation is still critical. Stay Safe” Downtown. 2:28PM

“City Health Department: Do not go to the hospital for minor injuries or illness. Also we are establishing blood donation stations at the libraries in neighborhoods 6 11 and 13. Please donate.” 2:25PM

Point of inflexion #3 : the aftershock

The third dataset captures the aftershock (Fig18). This aftershock means that resources will have to be redistributed in light of the recent events. One example of this is that bridges that had reopened are closed again.

“The Magritte Bridge is closed until further notice. This is a precautionary action pending safety inspection.Use other routes. FleetWilliamsCandy1967” 3:03PM

Fig18 Whole term tree, showing messages for term “shuddering”, they were sent during the aftershock

In terms of infrastructures

Many people are camping in the park, a storm is coming and a lot of resources are used to get people to shelters.

“We are offering transportation to diaster shelters for campers in the parks. A storm is coming and you do not want to be living in a tent. #transportation” 2:00PM

Yet the buses for the shelter are late and crowded (see Fig19), this situation is aggravated by the rubble and the flooding across town.

“Not sure if I can open my car door without water seeping in!” 5:27PM

“Large sinkhole swallows entire pond in someone he anyone anyone it users download the he app for more information” 2:09PM

Fig19 Reaction from citizens trying to reach the shelters

In terms of communication

People are trying to get in touch with their relatives, which tells us more about the problems (see Fig20). This suggests that the city should work on a system to let people report that they are safe.

Since the city emergency system is overloaded, people have no easy access to information (see example underneath) and rumors about what is happening keep going (see Fig21)

“I called the fire dept. and they told me to get off the phone. For rude!” 3:38PM

People are rallying to clean the town.

“Lets help each other out and clear that rubble off the sidewalks. People with disabilities like me and the elderly can't get anywhere until the sidewalks are cleared.” 1:30PM

Fig20 We learn from those messages that the City Hall and an High School collapsed

Fig21 A new kind of rumor

3 – Take the pulse of the community. How has the earthquake affected life in St. Himark? What is the community experiencing outside the realm of the first two questions? Show decision makers summary information and relevant/characteristic examples. Limit your response to 800 words and 8 images.

To know how the earthquake affected St. Himark we analyze all messages since the main earthquake (Fig22). Outside the realm of the earthquake, many things are happening in St. Himark.

Fig22 Topic overview, the words on the side are a starting point for our analysis

Several things are happening outside the realm of the first two questions :

- Ongoing debate around the use of nuclear energy and its safety : “Should nuclear power is a nuclear power plants less than 93 million miles from anyone | #MakesItlessCan”

- Solidarity inside the community : As in all crisis situations many good and bad things happen, and if there is a few rumors of people looting and stealing there is way more evidence of solidarity and people helping with the reconstruction of the city, see below.

“I heard someone say there is looting downtown - can you believe it? People are horrible.” 08/04/2020 09:08AM

“Not the theater! Heading down there with my shovel!” 10/04/2020 11:57AM

“#alwayssafe Please! I want to help” 08/04/2020 2:45PM

- Organization of the ferret convention: The earthquake (Fig23) disturbed the organization of this convention.

- Party in the park while camping there: As many people have seen their house destroyed or are afraid to go inside in case of an aftershock, the city has opened the park to campers. This lead to a festive atmosphere (Fig24).

- Always safe company is advertising the help that they are distributing to the city (see Fig25), however is not always seen positively by everyone : “AlwaysSafePowerCompany using the situation as free advertising” 12:05PM

- As many things where damaged during the earthquake it opens opportunities that would not arise in normal situation. For example, there is a distribution of free ice cream in Old Town due to the power outage (Fig26). However, it caused people to try to get to this very damaged neighborhood. Thus sending even more people into the streets.

- Sadly, many people are traumatized by the events and have decided to leave St. Himark, see Fig27.

- The local TV is keeping people in touch with the local news and are reporting numerous local events as life go on in St. Himark (Fig28).

- Life would not be going on if people where not back to complaining about traffic which is made worse by the rubble (Fig29).

Fig23 A ferret owners convention was supposed to take place next week

Fig24 The party in the park was not perceived in the same way by everyone

Fig25 Initiatives from the always safe company

Fig26 Reaction to the free ice cream

Fig27 A lot of people are leaving St. Himark

Fig28 Krak TV features many different type of news

Fig29 Traffic is a constant throughout the event and seems to go back to normal level during April 10 morning as people are commuting back to work

4 –– The data for this challenge can be analyzed either as a static collection or as a dynamic stream of data, as it would occur in a real emergency. Describe how you analyzed the data - as a static collection or a stream. How do you think this choice affected your analysis? Limit your response to 200 words and 3 images.

The Papyrus software supports static analyses of text collections. It includes traditional NLP, and topic extraction algorithms, and provides interactive visualizations to explore the extracted topics and term co-occurrences using biclustering.

In this context, it was critical to define appropriate time intervals for analysis.

For example in the first question, we only analysed data between 8:35AM and 1:35PM. It allowed us to have a global overview of the situation without being overwhelmed by the data.

In the second question, we first drew a timeline in excel in order to slice the data.

This affected our analysis in several ways: first it meant that we had to choose the right dataset or we would get irrelevant answers. Secondly, sorting messages in chronological order in the message panel helped us resolve apparent contradictions. For example in the first question we found that the Magritte Bridge was both opened and closed (see Fig30 and Fig31). Only a quick check on the timestamp allowed us to sort out the information. However this case is not frequent and when it occurs it is very easy to resolve interactively with Papyrus.

Fig30 At 10:46AM the bridge is closed

Fig31 At 1:23PM the bridge is open