IRJET - Survey Paper on Tools Used to Enhance User's Experience with Consumer Mobility Applications

International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056
Volume: 07 Issue: 03 | Mar 2020 www.irjet.net p-ISSN: 2395-0072
© 2020, IRJET | Impact Factor value: 7.34 | ISO 9001:2008 Certified Journal | Page 2581
Survey Paper on Tools Used to Enhance User's Experience with
Consumer Mobility Applications
Yashveer Singh Sohi1, Omkar Parab2, Himanshu Pushkarna3
1,2,3Computer Engineering, Fr. C. Rodrigues Institute of Technology, Navi Mumbai, India
---------------------------------------------------------------------***----------------------------------------------------------------------
Abstract - The intent of writing this paper istoreviewsome
of the modern tools that could refine the consumer's
experience while dealing with the sale and acquisition of real
estate commodities. Before implementing anything, it is of
paramount importance that an exhaustive studyofalltoolsbe
prepared. Thus, in the subsequent sections of this document, a
few key functionalities needed in a Consumer Mobility
Application for Real Estate, are highlighted, and the tools
required to implement such features are studied.
Key Words: Hybrid Applications, Compiled Applications,
Document Object Model (DOM), Virtual Reality (VR),
Augmented Reality (AR), Natural Language Processing
(NLP), Regular Expressions (RE), Optical Character
Recognition (OCR), Convolution Recurrent Neural
Networks (CRNN).
1. INTRODUCTION
At the time of this writing, the real estate sector in India has
not been digitized on any noticeable scale. This lack of
automation, impacts both, the customers of real estate
commodities, and the organizations selling such
commodities. Since the transactions made by a customer in
this industry are often times significant, it is natural for
people to be overly curious and inquisitive about the status
of their purchases. The absence of an online platform gives
customers no choice but to satiate theircuriosityfacetoface.
Such interactions, when frequent, become irksome for the
consumers as well as for the real estate corporations. This is
because, in order to uphold high standards of consumer
satisfaction they have to spend humanresources,intheform
of assigning customer relationshipagentsfor eachcustomer,
to cater to their doubts, which are oftentimesrepetitive. The
repetitive tasks are not restricted to straightening out the
queries that a customer may have, but also to verify and
process all the legal documents involved, manually, whichis
a tedious time taking task and hence costs the company
valuable man-hours. Thus, it is not hard to see why there is
dire need of digitization in this industry.
In this survey, we first compare a list of Cross-Platform
mobile application development frameworks toglean which
framework may best suit our use-case. Secondly, weanalyse
various tools, frameworks and methodologies so as to
determine the best approach to implement the following
functionalities –
 Provide customers with a 3D walk-through of the
sample flat instead of a standard 2D blueprint.
 Integrate an interactive chat-bot to handle all
pertinent FAQs from the consumers.
 An Optical Character Recognition software to
automate the data entry task needed to store the
information in legal documents.
2. CROSS PLATFORM MOBILE APPLICATION
DEVELOPMENT FRAMEWORKS
In the modern world, consumers prefer mobile applications
over web sites due to better user experienceandgreaterease
of access. However, there are a gulf of frameworks and
languages to choose from when it comes to mobile
application development.
It is crucial that all the requirements of the application be
studied thoroughly, based on which one can select a suitable
framework or language, which can mitigate the trade-off
between efficiency and cost. In this survey, we have
categorized mobile applications into four broad categories
based on how they are built –
 Applications built on Native Languages (Java for
Android platform, and Swift for IOS platform)
 Hybrid Applications (Phonegap/Cordova)
 Compiled Applications (React Native)
 Compiled Applications (Flutter)
In the following sections, the aforementioned tools are
compared with each other based on a number of
characteristics so as to aid developers in selecting an
appropriate approach, consideringtheir respectiveuse-case.
2.1 How the Applications are made
Firstly, in applications built on native languages, the
applications are written in the native language of the given
platform. This essentially means thatapplicationswrittenfor
Android platform are written in Java (say) and the ones
written for IOS platform are written in Swift (say).
Next, the hybrid applications are actually web applications
built using HTML, CSS and JavaScript. Frameworks such as
Cordova can render these web applications into the native
WebView component of the device so that the users feel that
they are using a mobile application instead of a browser.
Finally, for compiled applications, the applications are
written in a different language, and later compiled to the
native language of the platform on which they run. For

instance, developers need to write code in JavaScript when
using React Native, and Dart when using the framework
Flutter. These languagesarecompiledtothenativelanguages
of Android and IOS platform, and hence here anintermediate
compilation step is needed.
2.2 Code Reusability
Applications builtusingnativelanguagescannotbedeployed
across platforms. Thus, this poses and overload on the
development team to write codes, for the same logic, in
different languages, for Cross-Platform development. Due to
this disadvantage, the development team needs to have a
strong grasp on a minimum of two languages, one for each
platform.
On the other hand, virtually theentireapplicationisreusable
if it is a hybrid one. This is due to the fact that, hybrid
applications are essentially web application and the code
base can be used across all platforms.
In compiled applications, that are builtonReactNative,most
of the logic is usable across platforms, however, one has to
style certain component differently in different platforms so
that the overall look and feel of the applications remains
consistent. On the other hand, developers working with
Flutter need not worry about violating consistency across
platforms, even though components can be separately styled
for different platforms if one may choose to do so.
2.3 Ecosystem
In this context, ecosystem represents the third-party
packages that are available for a particular framework or
language.Italsotakesintoaccountthecommunitythatworks
on a particular technology, and the help and support that is
offers to its fellow members.
In this respect, the ecosystem for native languages is
plentiful, where developers have access to numerous open
source librariesandpre-styledcomponents.Theecosystemis
vaster for native languages in comparison with any other
platform used to build mobile applications. Developers
working with these languages have a gulf of built in APIs that
can be utilized to access almost all native device features
(such as contacts, camera etc.).
When it comes to hybrid applications, the ecosystem here is
well established too. However, since the application is
actually a web application, most of the packages need to be
tailored before putting them to use for building mobile
applications. The challenge here is to use native device
features. Unlike native languages, here developers do not
have packages to access all native device features. Except a
few common features,developerswillhavetobuildtheirown
wrappers if the application needed to use native device
features.
In case of compiled applications, when we look into React
Native, the framework uses JavaScript, which is a well-
established language witharichecosystem.Thus,developers
have access to a wide range of third-party packages.
However, since developers are building a mobile application
using React Native, care should be taken to avoid using
JavaScript packages that interact with the Document Object
Model (DOM). On the other hand, Flutter is a new framework
that still encounters bugs from its users. Thus, even thoughit
does have a rich library of pre-styled components and
widgets, the online community of Flutter is still in its early
adolescent stages. This means developers may encounter
issues that have no precedent and will need to work around
them on their own.
Lastly, for compiled applications, accessingthenativedevice
features is easy as compared to hybrid applications as the
frameworks, React Native and Flutter, both allow the
developers to code for them in the native languages
conveniently(ifthethird-partypackagesareunavailablefora
particular feature).
2.4 Performance and Real-World Usage
When it comes to performance, applications built on native
languages show the best performance. This is because every
aspect of the application is explicitly coded for by the
developers, and hence the developer has the freedom to
optimize the application as much as possible. In addition to
this, there are no wrappers or intermediate compilation
stages in these applications and hence, this furtheraugments
the performanceoftheseapplications.Afewapplicationsthat
are known to have been built, partly or completely, using
these languages are - VLC Media Player, Bitcoin Wallet
(Android), NASA World Wind, Twitter etc.
On the other hand, the performance of hybrid applications
deteriorates typically due to the additional wrapper that
renders the web application into the WebView componentof
the native device. Wikipedia is a famous example of a hybrid
application.
The performance of compiled applications is hindered to
someextent due to the intermediatecompilationstep,butthe
performance is still better when pitted against the hybrid
applications. As far as Cross Platform Mobile Application
Development frameworks are considered,ReactNativeisthe
most widely used, with social media giants such as Facebook
and Instagram being built on it. On the otherhand,Flutterisa
recently developed framework and still has not received any
noticeable traction. Google ads is an example of a Mobile
application which is built on Flutter.
3. 3D WALK-THROUGH
Every single customer, that have ever been in a position to
buy, or inquire about some commodity in the real estate
sector, have seen and attempted to make sense of a 2D

blueprint. This becomes challenging especially in urban
areas, where customers intend to buy flats in residential
complexes. To an untrained eye, it is difficult to visualize a
home by looking at the 2D layout, regardless of how
comprehensive the diagram might be.
Since humans in general are more accustomed to a 3D
perspective of things, it makes sense for real estate builders
to present customersa3Dwalk-throughoftheirfuturehome.
Providing this through Virtual Reality (VR) is possible in a
controlled environment, such as a sales office. Since
customers cannot be expected to have the necessary
equipment, suchas a VR headset, consumers need to cometo
the sellers for this experience. Rather, we aim to bring this
feature to them, through their smartphones. That can be
achieved using Augmented Reality (AR).
In this survey, we compare 2 popular game engines used by
major corporations for building AR products-
 Unreal Engine
 Unity Game Engine
Unreal Engine is a game engine developed by Epic Game in
the language C++. With Unreal, developers can build high-
fidelity visuals in almost no time, which makes Unreal the
preferred choice for developers when the product is to be
deployed on high-end devices.
Unity Game Engine on the other hand is a cross platform
game development engine built by Unity Technologies in the
language C++. With Unity, it is difficult to create high-fidelity
visuals that can compete with Unreal. In addition to this, it
takes much more resources andman-hourstobuildgraphics,
that are at par with Unreal. However, Unity is built to beused
on low-end devices, such as a smartphone. Thus, Unity is the
ideal tool to be used for the 3D walk-through in our use-case.
4. CHAT-BOT
A chat-bot is a software capable of conversing with a human
through textual our auditory means. A successful chat-bot
passes the Turing test, which meansthatahumanconversing
with the bot should always be under theimpressionthatthey
are engaged in a conversation with another human,andnota
computer program.
80% of all businesses are expected to have chat-bot
automation by the year 2020.Surely, there has to beareason
for this remarkable growth in the popularity of chat-bots.
Primarily, this is due to the fact that deploying chat-bots to
solve mundane, trivial, everyday problems faced by
customers is drasticallycheaperwhencomparedtoassigning
customer service agents for such tasks. Chat-bots are
expected to save businesses up to 30% in customer service
spending. Thus, it makessensethatapplicationspertainingto
the real estate sector, which experiencescontinualconsumer
interactions, should look forward to the advantages that a
chat-bot can provide. In addition to being a cheaper
alternative, chat-bots are much more accessible than their
human counterparts, and can look through information in a
database much faster.
Even though chat-bots have gained traction recently, the
concept of computer programs mimicking human
interactions is not new. The first chat-bot was called ELIZA
and was deployed in 1966. From that time, there have been
great feats of achievements when it comes toperfectingchat-
bots. In this survey, we aim to analyze a few key chat-bots
that have come to the market over the years.
4.1 ELIZA - 1966
ELIZA is a Natural Language Processing (NLP) based
conversational program, mimicking a therapist, that was
developed by researchers at the Massachusetts Institute of
Technology (MIT) Artificial Intelligence Laboratory. This
software merely useda pattern matching logic,implemented
using regular expressions. Therefore, the softwaredisplayed
no intelligence, contrary to what many users of the
application thought. ELIZA took the sentences typed by the
user, matched it with a set of patterns using regular
expressions, and returned a response. The response is either
a smart modification over the users input or a vague generic
statement that fits the general context of the conversation
expertly. The figure below (Fig. 1) shows an example of a
conversation made with the ELIZA chat-bot.
Fig -1: Sample Conversation with ELIZA
4.2 PARRY - 1972
PARRY is a chat-bot developed by researchers at the
Stanford University. It marked the first instancewhenachat-
bot passed the Turingtest.PARRYsuccessfullyresembledthe
behavior of a paranoid schizophrenic, because a highly
experienced group of psychiatrists were wrong 52% of the
time when it came to identifying whether PARRY was a chat-
bot or a real person.
The reason for PARRY's success is ascribed to the fact that it
had a dynamically changing demeanor towards the user's

statements, in addition to a patternmatchingmechanismlike
ELIZA. PARRYcould achieve this using 3 Affective Variables-
Anger (ranging from 0-20), Fear (ranging from 0-20),
Mistrust (ranging from 0-15). These variables would adjust
themselves based on certainmathematicalmodelsandhence
give a so-called attitude to the chat-bot. For instance, if the
user implies that PARRY is mentally ill, then there is rise in
values of Anger and Fear by a certain percentage, which in
turn affects the responses generated hence. Likewise, if the
user is using flattery, then these variables decrease their
value by a certain factor, andacorrespondingreactionisseen
in the quality of responses from PARRY.
In the figure below (Fig. 2) we can see a sampleconversation
taken place between ELIZA and PARRY. It is obvious that
PARRY outperforms ELIZA due to the behavioral trait that
was induced in it.
Fig -2: Sample Conversation between ELIZA and PARRY
4.3 ALICE - 1995
Artificial Linguistic Internet Computer Entity (ALICE), also
referred to as Alicebot, isanotherNLPbasedchat-botbuilton
pattern matching algorithm, that is inspired by ELIZA. The
program usedanExtensibleMarkupLanguage(XML)schema
known as Artificial Intelligence Markup Language(AIML)for
specifying certain heuristics that governed the pattern
matching in the case of ALICE.
AIML consists of data objects called AIML objects, which are
made up of units called topics and categories as shown in the
figure below (Fig. 3).
Fig -3: The AIML format
The topic is an optional top-level element, it has a name
attribute and a set of categories related to that topic.
Categories are the basic unit of knowledge in AIML. Each
category is a rule for matching an input and converting to an
output, and consists of a pattern, which represents the user's
input, and a template, which implies ALICE's answer. The
idea of the pattern matchingtechniqueisbasedonfindingthe
best, longest, pattern match.
Despite its superior algorithm, ALICE was not able to beat
the Turing test for prolonged conversations. A sample
conversation with ALICE is shown below (Fig. 4).
Fig -4: Sample Conversation with ALICE

4.4 TAY - 2016
TAY was a chat-bot built by Microsoft, intended to possess
the behavior of a 19-year-old teenager. The software was
deployed on Twitter and was supposed to be a source of
entertainment for the users in the age group 18-24 yearsold.
However, it had to be taken offline within 16 hours of
interaction, since it sparked major controversies on the
platform by posting extremely offensive content.
This mishap took place because TAY was exposedtoahighly
unpredictable environment at Twitter. Itwasmadeprivytoa
huge amount of offensive content, andsubjectedtogreatdeal
of online harassment on the platform in the short time that it
was active. The Machine Learning algorithms and NLP tools
embedded in TAY are capable to learn the positive and the
negative content equally well, given that the data is available
in the right proportions. Therefore, TAYlearntsuchoffensive
content (which was presentinexcess),andreproducedthem,
sparking the controversy.
TAY was taken off the platform and currently Microsoft is
working to build a version of TAY which can take ethics into
consideration before generating responses to user's inputs.
This can be achieved by using a restrictive corpus while
training the chat-bot, or by categorizing sensitive content
based on intent classification and keyword extraction
techniques.
This illustration clearly demonstrates that identifying and
studying the intended audience is extremely important,
especially when designing an intelligent chat-bot.
5. OPTICAL CHARACTER RECOGNITION
Optical Character Recognition (OCR) is the process of
digitizing handwritten or typed text. The text could be fed
into an OCR software in the form of a scanned document, a
photo of a document, a scene-photo (for example the text on
signs and billboards in a landscape photo) or from subtitle
text superimposed on an image (for example from a
television broadcast). The software will be able to convert
that into a text format which can be stored in a standard
database, or operated upon.
This is extensively used in the domain of data entry, where
data from bills, business cards, mail, printouts of static-data
etc. need to be stored in digital format, and where manually
extracting data is not feasible. In the domain of real estate,
users need to input their personal details on numerous
occasions. These details must match the one's in their legal
documents. Instead of manually copying all such details in a
form (say), users can simply upload a scanned copy of their
documents, which can be fed into the OCR software,
providing the digitized text instantly. Furthermore, the real
estate sellers need not employ people in data entry jobs for
the same. Hence, in the real estate sector, OCR can be utilized
to make the work of both, consumers and sellers, easy.
In this survey, we elucidate on 2 approaches to tackle the
problem of Optical Character Recognition.
 Classic Computer Vision Techniques.
 CRNN (Convolution Recurrent Neural Networks)
5.1 Classic Computer Vision Techniques
In this approach, one usually follows the following 3 steps –
 Firstly, we apply filters to the image to highlight the
intended characters (the characters that are
supposed to be detected).
 After this step, we use contour detection to detect
the characters highlighted.
 Lastly, we apply image classification techniques to
identify the detected character in the previous step.
The challenge in this approach is that contour detection is
difficult to generalize, which means that a lot of manual fine
tuning goes intostep 2 before obtainingthedesiredaccuracy.
For instance, this approach works quite well on the
illustration described in Fig. 5 and Fig. 6.
Fig -5: Characters to be recognized are well spaced

Fig -6: Contour Detection is working well for well-spaced
characters
However, the approach requires further fine tuning in the
case described in Fig. 7 and Fig. 8.
Fig -7: Characters to be recognized are not well spaced
Fig -8: Contour Detection is not working well when
characters are close together
5.2 Convolution Recurrent Neural Networks
CRNN (Convolution Recurrent Neural Networks) is a hybrid
end-to-end neural networks architecture that intends to
capture text in 3 steps.
 The first step uses a Fully Connected Convolution
Neural Network, with the last layer, called as the
feature layer, divided into segments, known as
feature columns. Each feature column represents a
certain section of the text.
 The feature columns so generated act as the input
for a deep-bidirectional LSTM. This layer is
responsibleforfindingrelationsbetweencharacters.
 Finally, the sequence generated in the previous step
is utilized by the transcription layer, which removes
redundancies and blank characters from the input
data by using probabilistic methods.
The figure shown below (Fig. 9) summarizes the process
described above.
Fig -9: CRNN Architecture

ACKNOWLEDGEMENT
We would like to extend our gratitudetoMr.MritunjayOjha,
Assistant Professor, Computer Department, Fr. C. Rodrigues
Institute of Technology, who's contribution was
instrumental in finishing this survey successfully on time.
Furthermore, we would like to thank our families and
friends who supported us in the course of writingthispaper.
REFERENCES
[1] Andreas Biørn-Hansen, Tor-Morten Grønli, Gheorghita
Ghinea, and Sahel Alouneh - "An Empirical Study of
Cross-Platform Mobile Development in Industry."
[2] Nitin Nimbalkar - "Top Programming Languages for
Mobile App Development"
[3] Existek, Software Development Company - "Hybrid VS
Native App: Which one to choose for your business?"
[4] Annie Dossey - "[Infographic] A Guide to Mobile App
Development: Web vs. Native vs. Hybrid"
[5] Tim A. Majchrzak, Andreas Biørn-Hansen, Tor-Morten
Grønli - "Comprehensive Analysis of Innovative Cross-
Platform App Development Frameworks"
[6] William Danielsson - "React Native application
development – A comparison between native Android
and React Native"
[7] Narendra Nagpal - "React Native vs Hybrid for mobile
apps: Which is Really Better?"
[8] Ja Young Lee, Tao Dong - "What We’ve Learned fromthe
July 2018 Flutter User Survey"
[9] Shashikant Jagtap - "Flutter vs React Native: A
Developer’s Perspective"
[10] Creative Bloq Staff (3D World) -"UnityvsUnrealEngine:
which game engine is for you?"
[11] Joseph Weizenbaum - "ELIZA A Computer Program For
the Study of Natural Language CommunicationBetween
Man And Machine"
[12] A. M. Turing - "COMPUTING MACHINERY AND
INTELLIGENCE"
[13] MEGAN GARBER - "When PARRY Met ELIZA: A
Ridiculous Chatbot Conversation From 1972"
[14] Bayan AbuShawar, Eric Atwell - "ALICE chatbot: Trials
and outputs"
[15] Wizu, Chatbots Magazine - "A Visual History Of
Chatbots"
[16] Yuxi Liu - "The Accountability of AI — Case Study:
Microsoft’s Tay Experiment"
[17] Narendra Sahu1, Manoj Sonkusare - "A STUDY ON
OPTICAL CHARACTER RECOGNITION TECHNIQUES"
[18] Gidi Shperber - "A gentle introduction to OCR"

IRJET - Survey Paper on Tools Used to Enhance User's Experience with Consumer Mobility Applications

More Related Content

What's hot (13)

Similar to IRJET - Survey Paper on Tools Used to Enhance User's Experience with Consumer Mobility Applications (20)

More from IRJET Journal (20)

Recently uploaded (20)

IRJET - Survey Paper on Tools Used to Enhance User's Experience with Consumer Mobility Applications