extracting a keyword network of flood disaster measures motoki miura, mitsuhiro tokuda, and daiki...

16
Extracting a Keyword Network of Flood Disaster Measures Motoki Miura, Mitsuhiro Tokuda, and Daiki Kuwahara Department of Civil and Architectural Engineering, Faculty of Engineering, Kyushu Institute of Technology

Upload: joseph-owens

Post on 01-Jan-2016

220 views

Category:

Documents


0 download

TRANSCRIPT

Extracting a Keyword Network ofFlood Disaster Measures

Motoki Miura, Mitsuhiro Tokuda, and Daiki KuwaharaDepartment of Civil and Architectural Engineering,

Faculty of Engineering,Kyushu Institute of Technology

Risk of Flooding

• Frequent Extra-tropical cyclones by greenhouse effect increase the risk of flood disaster and localized heavy rainsin the decade

• To prevent/reduce the flooding damage, anti-disaster headquarters must have overall knowledge of anti-disaster measures(ex. know-hows, tips)

Purpose

• To provide overall knowledge of anti-disaster measures,

• (1) we analyze measures from previous disaster, and

• (2) we visualize the structure of measures as “summary”

• The “summary” can be effective to improve quality of disaster-recovery services/cares

Sample of the anti-disaster measures for city-center headquarter (1/2)

• Do not hesitate to counsel refugees.Safety of human life is most important.

• We observed that even in emergency, people do not evacuate. To make people aware of the need for evacuation is crucial.

• Volunteer center should be established immediately. Volunteer is not only a worker, but is also hope for recovery.

Sample of the anti-disaster measures for city-center headquarter (2/2)

• A huge amount of garbage is thrown away. A temporary garbage dump should be constructed immediately. For quick processing, separation of garbage should be promoted.

• Do not hesitate to act for relief even if it costs much money. Any financial issue can be settled afterwards. The head of municipality must prove that we can afford the refugee costs.

Preprocess

• 828 measures are taken from Book “Know-hows of prevention, reduction and recovering of flood disaster --- message from the striken areas”

• We manually labeled a few keywords to each measure (source text) , to represent the original context/meaning

The Labeling Sample (1/2)

• Do not hesitate to counsel refugees.Safety of human life is most important.– Life | Refuge | Refugees Counsel

• We observed that even in emergency, people do not evacuate. To make people aware of the need for evacuation is crucial.– Refuge | Refugees Counsel | Publication | Tense Situation

• Volunteer center should be established immediately. Volunteer is not only a worker, but is also hope for recovery.– Volunteer Center | Volunteer | Victims | Encouragement

The Labeling Sample (2/2)

• A huge amount of garbage is thrown away. A temporary garbage dump should be constructed immediately. For quick processing, separation of garbage should be promoted.– Garbage | Temporal Place | Garbage Separation

• Do not hesitate to act for relief even if it costs much money. Any financial issue can be settled afterwards. The head of municipality must prove that we can afford the refugee costs.– Budget | Immediately After | Head of Municipality

Visualizing System (based on Prefuse toolkit)

Visualize Graph (Node: Measure + KW) • 828 measures, 188 keywords, 2.539 links

Visualize Graph (Node: KW only)• Eliminate Measure-ID nodes, and link neighboring keywords directly• 188 keyword nodes, 3.450 links (911 links are added)

3.450 links

Visualize Graph (KW only, link reduced)• Only the most frequent keyword on the measure can link to another• 188 keyword nodes, 1.961 links (43% of links reduced, but still dense)

3.450 links 1.961 links

Co-occurrence Keyword Network

• Consider co-occurrence degree (numbers of overlapping links) among keyword nodes

419

428

Disposal

Electric Appliance

417

Making Resource

Electric Appliance

Disposal

Making Resource

Degree = 3

Degree = 1

Frequency of co-occurrence degree

6-24 co-occurrence graph• 6 co-occurrence : green 32 links over 6 co-occurrence: red 34 links

Conclusion and Future Work

• We discussed our trials to analyze know-hows (measures) on flood disaster measures

• We found the co-occurrence keyword network can represent meaningful relationships– The graph can provide overview for anti-disaster

headquarters and citizens

• Future Work– Semi-Automatic labeling– Improve interface design for quick graph analysis