AI-DRIVEN COMPLAINT MANAGEMENT SYSTEM
Shreyas Shedge, Pratik Wadkar, Atreya Raorane, and Bushra Shaikh
Department of Information Technology
SIES Graduate School of Technology,
Navi Mumbai, India
[email protected], [email protected]
Abstract— Citizens Complaint is important information poor organizational or government services. With a
reflecting citizen's sound. Our main objective of the organization
is to give valuable and productive feedback to the citizens. The priority on grievances, firms can gain an efficient
design we proposed for an AI-driven logging portal will have the success factor by increasing service quality. Each
strength to minimize citizens' worry and additionally, it can organization must therefore establish its internal
inspire people to promote our country by logging complaints on
our website. In this paper, we propose a new model that is an and external customer interaction with its people.
AI-driven logging portal where we try to improve Although adequate communication can minimize
communication between citizens and government, and we try to customer frustration, and complaints cannot be
make transparent communication to make your country a better
place to live. There are different services for different types of avoided.
complaints in your web portal. These services are used by In social organizations, there are grievances about
numerous citizens on basis of their grievances. We have created a the issues requiring an appropriate logging portal.
framework that can recognize citizens' problems and provide
timely feedback to citizens. This system can recognize grievances People complain every day to public servants
by identifying and commenting on each complaint that has been because they feel terrible. All complaints should be
raised. The concern of citizens is treated according to the priority acknowledged and duly handled, regardless of any
in this portal. That is a problem depending on the seriousness of
the situation that will be prioritized. citizen's internal or external bills, such as
face-to-face issues, phone accusations, complaints
Keywords— AI-driven logging portal, e-complaint, Issues letters, and comments on the internet. Also, it
Handling with Priority. should prioritize this complaint, i.e., insufficient
grievance or serious allegation should be given
I. INTRODUCTION priority, and appropriate action should be taken to
A portal is a web-based platform that collects prioritize the complaint handling. To overcome all
citizens' complaints into a single user interface and these issues and to strengthen citizen-government
presents users with the most relevant solution for ties. We have implemented an online portal that will
their context. Simple web portals have developed help citizens save time by lodging complaints
into networks that promote digital customer service online. It will also help the complaint department
initiatives over time. In the early days of offline, solve problems according to the seriousness of a
content discovery was a problem that portals problem that will be helpful for both citizens and
attempted to solve. In the face of a large number of government. The idea of an AI-driven logging
citizens' issues, the complaint logging portal gives portal is implemented for the first time.
administrators and users the easiest way to resolve
issues based on the severity of the problem. Portals II. LITERATURE REVIEW
could display selected content to provide a custom We have reviewed many research papers to gain
experience for users. Likewise, most consumer knowledge and insight into how our project is being
issues are evident where the government and the implemented.
users have insufficient contact with a service
provider. This poor communication may lead to
Electronic copy available at: https://ssrn.com/abstract=3866078
Harne et al. [1] In this publication, we examined diagnostic, and complaint assessment. It represents
how to collect information on citizen data from the the need for an organization's improvement. Also,
web pages that we included in this project. We this paper assesses the framework and its benefits
also learned to analyze the data and perform an automatically in response to various activities,
action on that data. We used this technology to which help decision supports complaints.
make some data analysis changes and performed
some sentimental analysis on citizens' data to R. Razali et al. [6] We have government bodies
prioritize the complaints. This paper also shows (Municipal Corporations) in our country that
converted data to the people.so we have applied oversee sustaining and operating towns. It is their
that in our project to perform analysis and update collective responsibility to respond to citizen
the data and display the status of the complaint to complaints. For this purpose, the municipal
the citizens corporation has two options: first, cameras or other
surveillance devices must be mounted, and second,
Nadar et al. [2] We study how to perform actions people can report their problems to the municipal
on users' data, so we performed operations on corporation. The aim of creating this website to
citizens' data with this paper's help. We used simplify the process of lodging complaints to the
algorithms to find the complaint score and found respective Municipal Corporation and make it
the score by using sentimental analysis. This paper quick and cheaper.
also helped to give them priority. We prioritize
complaints according to the score. The negative Sana Ullah Eduardo et al. [7] In this paper, we
score will be the top priority, and the positive will determine the optimal performance bounds for a
be the least priority. This paper aided us in fixed number of devices in different priority classes
prioritizing grievances and data separation in our with different bit error ratio values. In comparison
project. to low-priority devices, high-priority devices gain
fast and secure access to the error-prone channel.
Hegazy et al. [3] This paper discusses the Our analysis's performance bounds are used to
recording complaints given by citizens. With the understand the tradeoffs between different priority
help of this paper, we stored the data of citizens in levels and network performance. So, we used this
a database. We use PHP, MyAdmin, to store information to prioritize complaints of citizens in
citizens' data and provide excellent feedback to the our project.
citizens on their complaints, whether the
complaint is solved, unsolved, or in progress. Due
to this, citizens and government can have a Yuling Chen et al. [8] In this paper, we determine
transparent conversation. Sentiment analysis of Internet reviews is a hot
research topic in Web information mining. The
R. Razal et al. [4] This paper addresses complaint standard text sentiment analysis approach relies on
procedures to identify reasonable and acceptable either an emotion dictionary or machine learning to
grievances. Many web-based applications are determine sentiment. However, the generalization
being developed in the modern age of technology. potential is limited due to its reliance on emotion
There should also be an online complaint dictionary construction and artificial layout and
management system. A robust foundation design,
extraction functionality. The results of the
particularly concerning the complaint process, is
required to develop a web-based complaint experiments show that the proposed approach
management system. These design findings help effectively increases text sentiment classification
us to create a successful web-based complaint accuracy. So, we have used sentimental analysis in
logging portal. our project to give priority to the complaint.
III. Methodology
Amy J.C. Trappey et al. [5] This paper explores
the impact and the productive dealing with The website that we have designed is coded in
customer grievance. The new system covers the Sublime Text, and we have used HTML5, CSS3 &
monitoring of grievances, compensation PHP for creating a website. Without trouble and
Electronic copy available at: https://ssrn.com/abstract=3866078
some effort, our website can handle the complaint 3) CSS3 – CSS3 is Cascading Style Sheets. It is
information. Since the work is performed manually used for more interactive graphics and animation.
before, it would take a lot of time and a great deal It is used to improve the appearance of the
of time to hold the files. and to store all this data, website.
we have used XAMPP & MySQL. These data can
be managed with a bit of effort and with less time 4) PHP– PHP is a server-side scripting language
by computerizing the system. The chances of that is used to produce dynamically produced
duplicating complaints are negligible because we pages quickly
have used PYTHON to perform operations on data. 5) Xampp– Xampp is used as a database to store
This data is in human language, so it is converted in data which user enters in the website in our
machine language using natural language toolkit project, and this software is free of cost
(NLTK)[1]library, and Sentiment Analysis(SA)[2]is
used to prioritizing grievances by calculating score 6) MySQL Database– MySQL Database is a
to each complaint, the score will be from -1 to 1 Structured Query Language. It is used for
more negative score gets the highest priority and adding, accessing, and managing content in a
second priority will be given to complaints which database. It is also freely available open-source
are pending from many days, and this all is put in software
the table by using Pandas Data Frame(PDF)[3]. Also
used (NUMPY)[4]for a one-dimensional array for 7) Python– Python is a high-level programming
citizens' storing data, the Citizen’s Complaint language. We have used python in our project to
Report can be quickly implemented by collecting perform operations on data entered by users.
data from all required details without any difficulty.
The package is constructed using the GUI concept 8) NLTK Library– NLTK Library is a natural
we used (Tkinter)5]for GUI, and it is very language toolkit in this project. It is used to
user-friendly & easy to use. We will use understand machine-human language and reply
with appropriate responses. Some of the packages
matplotlib[6]for showing the solved no. of cases and
we used in our project were tokenization,
the unsolved no. of cases in graphical form. The
stemming, lemmatization, punctuation, character
people should not go to the local office in the count, and word count. Sentiment analysis is a
proposed framework to address their problems. So natural language processing technique for
we have created a FAQ chatbot on your website by interpreting and categorizing emotions in
using (Dialogflow)[7] this will help citizens to solve subjective data.
basic queries like where can I track my grievances
and so on. By introducing the issue in this proposed 9) Sentiment analysis– Sentiment analysis is a
method, he can discuss his issues and can suggest a natural language processing technique used to
potential solution to the citizens' issues. interpret and classify emotions in subjective data.
In this, we have used giving priority to the
1) Sublime Text – Sublime text is used as a source complaints as -1 to 1 so that the government can
code editor we have used in this project. It can be solve a more severe problem first.
used for a variety of programming and markup
languages, and users can add functions using 10) Emotion detection– Emotion detection is a
plugins. It is also free software. form of sentiment analysis that aims to identify
emotions such as happiness, annoyance, rage, and
2) HTML5 – HTML5 is Hypertext Markup sadness. We will use this to give priority to the
Language. It can be used for a variety of complaints.
programming and markup languages, and users
can add functions using plugins. It is also free 11) Pandas data frame– Pandas data frame is a
software. two-dimensional data structure in which data is
organized in columns and rows in a tabulated form.
Electronic copy available at: https://ssrn.com/abstract=3866078
The data, rows, and columns are the three main Website-We have created a website for citizens
components of a Pandas Data Frame. for logging complaints on government services and
also created a FAQ chatbot for basic queries, After
12) NumPy - NumPy is a Python library used for a complaint has lodged user will get a response that
working with arrays. It also has functions for how much it will take approx time to solve the
working in the domain of linear algebra, Fourier complaint, user can also check how many
transform, and matrices. complaints are solved, unsolved overall in a graph
format.
13) Dialogflow - Dialogflow is a natural language
understanding platform that makes it easy to design Categories for logging complaints are
and integrate a conversational user interface into
your mobile app, web application, device, bot,
1) Electricity: This department will process the
interactive voice response system.
complaints related to electricity, like live wires,
damaged electric poles, and so forth.
WORKFLOW 2) Traffic: This department will process complaints
related to traffic violations, like triple riding,
moving in one way, no parking, riding a bike on the
footpath, and so forth.
3) Police: This department will process the
complaints related to criminal, activities such as
robbery, murder, kidnapping, and so on.
4) Municipality: This department will process the
complaints related to day-to-day life, such as
garbage problems, water leakage, drainage, Etc.
To lodge complaints users must first fill in all
necessary details so that we can check the
complaint is legit or not; we will also show
numbers of solved and unsolved cases so that
FIGURE 1. WORKFLOW DIAGRAM ARCHITECTURE everything will be transparent between citizens
and the government.
Database- There, we have created different
tables for different categories of complaints. In
this database, all user's information and all
complaints are stored. This data is fetched in
python, and operations are performed on data
and again send updated data to the database.
Then the user can see the status of his/her
complaints.
Natural Language Processing - Here Firstly, we
will fetch data from the database. We will use
pandas data frames to put data in rows and
columns. To understand human language into
machine language, we will use the NLTK library.
After collecting all the data, we will perform a
FIGURE 2. A RCHITECTURE sentiment analysis. We will find all main
Electronic copy available at: https://ssrn.com/abstract=3866078
keywords from the description of a problem
given by users, and we will find the score for it.
the score will be from -1 to 1. If the score is
negative, it will be solved first. If the score is
positive, it will be solved later. Suppose there is a
condition both scores are identical. In that case, it
will give priority to the user who has complained
first, and the second priority is given to a
complaint which is pending for many days it will
be solved. After prioritizing the complaints, it
will create a new column of the score, and
updated data will be sent to the database. Then it
will be updating the status of the complaint to the
user on the website
IV. IMPLEMENTATION
Admin Module
Here the admin first has to login with his/her id and
password, and after that, he can see citizens'
complaints. FIGURE 3. A DMIN MODULE(CATEGORIES OF COMPLAINTS)
In this section, the admin will give a response to
citizen about a complaint that it is solved or in User Module References
progress or how much time it will take to solve the
This is the user's dashboard where the user first has
complaint to fill in his/her information. After that, he can
Admin module has various sections like a pending complain about a problem with a photograph for
complaint, closed complaint, the complaint in proof.
progress, admin can add categories and
In the user module users can see how many
subcategories, for now, we have added only a few
complaints have been registered and how many are
categories like water supply, electricity problem,
in progress/unsolved/complaints solved, users can
and road problem.
also see complaints filed by other citizens and their
In this admin module, we can see logs of citizens
progress too.
when they logged out and log in, etc.
Electronic copy available at: https://ssrn.com/abstract=3866078
FIGURE 6. PREDICTION OF TIME TO SOLVE A COMPLAINT
FIGURE 4. USER MODULE (REGISTER COMPLAINTS)
FAQ Chatbot
Sentiment detection and priority allocation
Here we solve basic queries of citizens like how to
Here we will first fetch data of citizens from the lodge complaints, where can I track my grievances,
database and we will perform sentiment analysis on and so on.
the description of complaint given by the user and
we will create a score for that complaint score is
calculated by three-factor negative, neutral, positive
and by using lambda function we will calculate the
score in compound value between -1 to 1, and later
we will sort the complaints according to score.
More negative the score, the more priority to that
complaint.
FIGURE 5. SENTIMENT DETECTION (PRIORITY ALLOCATION)
Prediction of Time takes to resolve a complaint
This is the backend part here we basically predict
how much time it will take to solve the complaint
of the customer. Prediction is done on basis of
seriousness and priority given to the problem. FIGURE 7. FAQ CHATBOT
Electronic copy available at: https://ssrn.com/abstract=3866078
V. CONCLUSION
[10] Yooncheong Cho, Il Im, R. Hiltz and J. Fjermestad, "An analysis of
online customer complaints: implications for Web complaint management,"
This project will improve communication between Proceedings of the 35th Annual Hawaii International Conference on System
Sciences, Big Island, HI, USA, 2002, pp. 2308-2317, doi:
government and citizens; in this project, we have 10.1109/HICSS.2002.994162.
shown how the system will work according to the
complaint’s priority. This project provides an [11] R. Razali and J. Jaafar, "Complaint handling theoretical framework,"
2012 International Conference on Computer & Information Science
overview of the AI-Driven complaint logging portal (ICCIS), Kuala Lumpur, Malaysia, 2012, pp. 382-385, doi:
as a web service design and execution. The 10.1109/ICCISci.2012.6297275.
outcomes of implementation are that people will
[12] C. Chauhan and S. Sehgal, "Sentiment analysis on product reviews,"
communicate directly with the government, and 2017 International Conference on Computing, Communication and
issues will be addressed depending on the problems' Automation (ICCCA), Greater Noida, India, 2017, pp. 26-31, doi:
10.1109/CCAA.2017.8229825.
seriousness. Claims and feedback are essential
sources of knowledge for developing the nation and [13] M. I. Prabha and G. Umarani Srikanth, "Survey of Sentiment Analysis
the facilities used by organizations. As we know, Using Deep Learning Techniques," 2019 1st International Conference on
Innovations in Information and Communication Technology (ICIICT),
offline complaints are not so effective. The Chennai, India, 2019, pp. 1-9, doi: 10.1109/ICIICT1.2019.8741438.
preferred alternative is an AI-driven complaint
logging portal that can directly communicate [14] Q. Ye and D. Doermann, "Text Detection and Recognition in Imagery: A
Survey," in IEEE Transactions on Pattern Analysis and Machine Intelligence,
between citizens and the government and allows the vol. 37, no. 7, pp. 1480-1500, 1 July 2015, doi:
government to solve problems according to priority. 10.1109/TPAMI.2014.2366765.
Further, we will work on the front end to make it
[15] H. Zhang and D. Li, "Naïve Bayes Text Classifier," 2007 IEEE
more interactive, and we will add more categories International Conference on Granular Computing (GRC 2007), Fremont, CA,
to this project. USA, 2007, pp. 708-708, doi: 10.1109/GrC.2007.40.
REFERENCES [16] P. Sethi, S. Sonawane, S. Khanwalker and R. B. Keskar, "Automatic text
summarization of news articles," 2017 International Conference on Big Data,
[1] “Natural Language Toolkit”:NLTK: Natural Language Toolkit — NLTK IoT and Data Science (BID), Pune, India, 2017, pp. 23-29, doi:
3.5 documentation 10.1109/BID.2017.8336568.
[2] "Sentimental Analysis ": Analysis and application: Sentiment Analysis:
Concept, Analysis and Applications | by Shashank Gupta | Towards Data
Science
[3] “Pandas.DataFrame”:pandas.pydata: pandas.DataFrame — pandas 1.2.3
documentation (pydata.org)
[4] “NumPy”:numpy.pydata: NumPy
[5] “Tkinter”:TclTk: tkinter — Python interface to Tcl/Tk — Python 3.9.4
documentation
[6] "Matplotlib": Visualization with python: Matplotlib: Python plotting —
Matplotlib 3.4.1 documentation
[7] "Dialogflow": Dialogflow ES: Dialogflow Documentation | Google
Cloud
[8] A. J. C. Trappey, C. Lee, W. Chen and C. V. Trappey, "A framework of
customer complaint handling system," 2010 7th International Conference on
Service Systems and Service Management, Tokyo, Japan, 2010, pp. 1-6, doi:
10.1109/ICSSSM.2010.5530119.
[9] A. Faed, "Handling e-complaints in customer complaint management
system using FMEA as a qualitative system," 2010 6th International
Conference on Advanced Information Management and Service (IMS),
Seoul, Korea (South), 2010, pp. 205-209.
Electronic copy available at: https://ssrn.com/abstract=3866078