2019/09/28 Code for Japan Summit ( )× 11
Open311:
(Google Scholar 2014 400 )
Adversarial Adaptation of Scene Graph Models for
Understanding Civic Issues
Shanu Kumar
Indian Institute of Technology Kanpur, India
[email protected]
Shubham Atreja, Anjali Singh, Mohit Jain
IBM Research AI, India
{satreja1,ansingh8,mohitjain}@in.ibm.com
ABSTRACT
Citizen engagement and technology usage are two emerging trends
driven by smart city initiatives. Governments around the world are
adopting technology for faster resolution of civic issues. Typically,
citizens report issues, such as broken roads, garbage dumps, etc.
through web portals and mobile apps, in order for the government
authorities to take appropriate actions. Several mediums – text,
image, audio, video – are used to report these issues. Through a
user study with 13 citizens and 3 authorities, we found that image
is the most preferred medium to report civic issues. However, ana-
lyzing civic issue related images is challenging for the authorities
as it requires manual eort. Moreover, previous works have been
limited to identifying a specic set of issues from images. In this
work, given an image, we propose to generate a Civic Issue Graph
consisting of a set of objects and the semantic relations between
them, which are representative of the underlying civic issue. We
also release two multi-modal (text and images) datasets, that can
help in further analysis of civic issues from images. We present
a novel approach for adversarial training of existing scene graph
models that enables the use of scene graphs for new applications
in the absence of any labelled training data. We conduct several
experiments to analyze the ecacy of our approach, and using
human evaluation, we establish the appropriateness of our model
at representing dierent civic issues.
KEYWORDS
Civic Engagement, Scene Graph Generation, Adversarial Training,
Smart Cities, Intelligent Systems on Web
ACM Reference Format:
Shanu Kumar and Shubham Atreja, Anjali Singh, Mohit Jain. 2019. Adver-
sarial Adaptation of Scene Graph Models for Understanding Civic Issues. In
Proceedings of ACM Conference (Conference’17). ACM, New York, NY, USA,
Article 4, 11 pages. https://doi.org/10.475/123_4
1 INTRODUCTION
In recent years, there has been a signicant increase in smart city
initiatives [9, 30, 31]. As a result, government authorities are em-
phasizing the use of technology and increased citizen participation
for better maintenance of urban areas. Various web platforms –
SeeClickFix [27], FixMyStreet [1], ichangemycity [16] – have been
Permission to make digital or hard copies of part or all of this work for personal or
classroom use is granted without fee provided that copies are not made or distributed
for prot or commercial advantage and that copies bear this notice and the full citation
on the rst page. Copyrights for third-party components of this work must be honored.
For all other uses, contact the owner/author(s).
Conference’17, July 2017, Washington, DC, USA
© 2019 Copyright held by the owner/author(s).
ACM ISBN 123-4567-24-567/08/06.
https://doi.org/10.475/123_4
Figure 1: Comparison between Civic Issue Graph and Scene
Graph for the same image. The scene graph provides a com-
plete representation of all objects and relationships in the
image, while the Civic Issue Graph only consists of relations
representative of the civic issue.
introduced across the world, which enable the citizens to report
civic issues such as poor road condition, garbage dumps, missing
trac signs, etc., and track the status of their complaints. Such ini-
tiatives have resulted in exponential increase in the number of civic
issues being reported [2]. Even social media sites (Twitter, Face-
book) have been increasingly utilized to report civic issues. Studies
have found the importance of civic issue reporting platforms and
social media sites in enhancing civic awareness among citizens
[36]. These platforms help the concerned authorities to not only
identify the problems, but also access the severity of the problems.
Civic issues are reported online through various mediums – textual
descriptions, images, videos, or a combination of them. Previous
work [10] highlights the importance of mediums in citizen partici-
pation. Yet, no prior work has tried to understand the role of these
mediums in reporting of civic issues.
In this work, we rst identify the most preferred medium for
reporting civic issues, by conducting a user study with 13 citizens
and 3 government authorities. Using the 84 civic issues reported by
the citizens using our mobile app, and follow-up semi-structured
interviews, we found that images are the most usable medium for
the citizens. In contrast, authorities found text as the most preferred
medium, as images are hard to analyze at scale.
To ll this gap, several works have proposed methods to auto-
matically identify a specic category of civic issues from images,
such as garbage dumps [28] and road damage [24]. However, their
methods are limited to the specic categories that they address.
arXiv:1901.10124v1 [cs.AI] 29 Jan 2019
https://doi.org/10.475/123_4
https://doi.org/10.3390/ijgi8030115
RESEARCH ARTICLE
Structure of 311 service requests as a
signature of urban location
Lingjing Wang1,2, Cheng Qian1,2, Philipp Kats1,3, Constantine Kontokosta1,4,
Stanislav Sobolevsky1,5,6*
1 Center for Urban Science and Progress, New York University, Brooklyn, New York, United States of
America, 2 Tandon School of Engineering, New York University, Brooklyn, New York, United States of
America, 3 Kazan Federal University, Kazan, Russia, 4 Department of Civil and Urban Engineering, New York
University, Brooklyn, New York, United States of America, 5 Senseable City Laboratory, Massachusetts
Institute Of Technology, Cambridge, Massachusetts, United States of America, 6 Institute Of Design And
Urban Studies of The Saint-Petersburg National Research University Of Information Technologies,
Mechanics And Optics, Saint-Petersburg, Russia
*
[email protected]
Abstract
While urban systems demonstrate high spatial heterogeneity, many urban planning, eco-
nomic and political decisions heavily rely on a deep understanding of local neighborhood
contexts. We show that the structure of 311 Service Requests enables one possible way of
building a unique signature of the local urban context, thus being able to serve as a low-cost
decision support tool for urban stakeholders. Considering examples of New York City, Bos-
ton and Chicago, we demonstrate how 311 Service Requests recorded and categorized by
type in each neighborhood can be utilized to generate a meaningful classification of loca-
tions across the city, based on distinctive socioeconomic profiles. Moreover, the 311-based
classification of urban neighborhoods can present sufficient information to model various
socioeconomic features. Finally, we show that these characteristics are capable of predict-
ing future trends in comparative local real estate prices. We demonstrate 311 Service
Requests data can be used to monitor and predict socioeconomic performance of urban
neighborhoods, allowing urban stakeholders to quantify the impacts of their interventions.
1 Introduction
Cities can be seen as a complex system composed of multiple layers of activity and interactions
across various urban domains; therefore, discovering a parsimonious description of urban
function is quite difficult [1–4]. However, urban planners, policy makers and other types of
urban stakeholders, including businesses and investors, could benefit from an intuitive proxy
of neighborhood conditions across the city and over time [5–7]. At the same time, such simple
indicators could provide not only valuable information to support urban decision-making, but
also to accelerate the scalability of successful approaches and practices across different neigh-
borhood and cities, as urban scaling patterns have become an increasing topic of interest [8–
12]. As the volume and heterogeneity of urban data have increased, machine learning has
PLOS ONE | https://doi.org/10.1371/journal.pone.0186314 October 17, 2017 1 / 21
a1111111111
a1111111111
a1111111111
a1111111111
a1111111111
OPEN ACCESS
Citation: Wang L, Qian C, Kats P, Kontokosta C,
Sobolevsky S (2017) Structure of 311 service
requests as a signature of urban location. PLoS
ONE 12(10): e0186314. https://doi.org/10.1371/
journal.pone.0186314
Editor: Yanguang Chen, Peking University, CHINA
Received: December 4, 2016
Accepted: September 28, 2017
Published: October 17, 2017
Copyright: © 2017 Wang et al. This is an open
access article distributed under the terms of the
Creative Commons Attribution License, which
permits unrestricted use, distribution, and
reproduction in any medium, provided the original
author and source are credited.
Data Availability Statement: The data underlying
this study was obtained from a third party. The 311
data for the three cities is publicly available and can
be freely downloaded from: 1) New York City Open
data Portal https://nycopendata.socrata.co 2)
Boston Open data Portal https://data.cityofboston.
gov 3) Chicago Open data Portal https://data.
cityofchicago.org The census data is available at
census.gov. Zillow data is available from https://
www.zillow.com/research/data/. The authors did
not have any special privileges in obtaining this
data.
https://doi.org/10.1371/journ
al.pone.0186314