ASPECT BASED SENTIMENT ANALYSIS

14 April 2021
th

ABSTRACT
Product feedback is a very crucial point when a customer/consumer decides to buy a product or
service. Therefore, identifying an individual product with its unique features and analyzing how
each feature is defining the product could be an important way forward to making informed
decisions. On the other side companies are on the constant lookout for individual reviews for
their products and services to improve and get feedback about their products to build their
market strategy afterwards. Therefore, it has become very important to analyze the data present
and create the procedures to automatically classify these sentiments for the respective
products/services. The approach of sentiment analysis can be used to classify the sentiments of
various users on each product. Identifying each feature is an important step to analyze and then
determine the sentiment of the product in terms of its individual features for the user to gain an
understanding about the product details. In this paper, machine learning methods were used to
extract information from the online reviews posted by customers. The main emphasis is to label a
feature-wise score for each product depending upon the individual reviews.

BUSINESS UNDERSTANDING

BACKGROUND
Ecommerce has become the equivalent of the agriculture revolution in the business world. It
seems every day they having more influence in our life. Our country, Ethiopia is slowly catching
to this phenomenon. There are currently more than a dozen ecommerce platforms running in
Ethiopia. Customer sentimental analysis will give a huge boost to their performance, increasing
customer satisfactions, and improving their product recommendation.
Unfortunately using customer review data to optimize user experience does seems to be at the
early stages of development. It is unequivocal to say, if the ecommerce industry is to grow and
catchup to the level of the developed countries, sentimental analysis plays a huge role to ease the
way from a human interactive market to a digital one.

Business Objectives
As more companies make the transition to selling over the Web, an established
computer/electronics e-retailer is facing increasing competition from newer sites. Faced with the
reality that Web stores are cropping up as faster than customers are migrating to the Web, the
company must find ways to remain profitable despite the rising costs of customer acquisition.
One proposed solution is to cultivate existing customer relationships in order to maximize the
value of each of the company’s current customers. Thus, the goal of the project is to identify
overall customer experience, and find ways to elevate customer’s satisfaction, where they,
theoretically, will buy more, stay longer, and refer other customers.

Thus, the project is commissioned with the following objectives:

 To Measures the attitude of the customer towards the aspects of a service or product
which they describe in the feedback.
 To detect changes in the overall opinion towards your brand. Because it provides insight
into the way your customers are feeling when they approach you, you can monitor trends
and see if overall opinion towards your company drops or rises.
 To analyze what the most important issues are.

Business Success Criteria

The most important aspect of the business success criteria includes customer satisfaction, degree
of improvement compared to current conditions, amount of expected use of products by
customers and growth of customer involvement to a certain level of the system.

The project will be judged successful if:

 Cross-sales increase by 10%.

 The study finishes on budget & time

Inventory of resources

 Personnel: It is clear that there is some experience in data mining and managing
server logs and product databases, but no experience in data warehousing and data
cleaning for analysis. Thus, a database specialist may be consulted.
 Data: Since this is an established company, there is plenty of Web log and purchase
data to draw from. In fact, for this project, we will restrict the analysis to customers of
Amazon. If successful, the program can be expanded.
 Hardware: ?
 Software: ?
Requirements
The project is scheduled to be completed in less than three months. Also, since there is no cash
budgeted, it is imperative that the project to take little or no expenditure.

Assumptions
The development resources are assumed to be available from project initiation to project
completion. The project doesn’t require any financial expenditure.

Constraints

 The project must deliver according to the due date.

 Because of the type of the project and requirements the development team members must
be co-located.
 No internal resource has the required level of knowledge about
 IT system must comply with the architectural decision for this type of solution.
 The size of the data set must consider the available hardware capacity.

Risks and contingencies 

 Hardware failure or unavailability: using backup hardware.

 Lack of expertise: experienced individuals should be consulted.
 Financial cost: scope of the project should be reduced.
 Poor result: finding better data sets.

Terminology

 Data warehousing: the process for collecting and managing data from varied sources to
provide meaningful business insights.
 Hardware: a physical device used for computing purpose.
 Ecommerce: is the buying and selling of goods (or services) on the internet.
 Data miner: is an interdisciplinary subfield of computer science and statistics with an
overall goal to extract information (with intelligent methods) from a data set and
transform the information into a comprehensible structure for further use.
 E-retailer: a person or a company that sells goods through the internet.
Costs and benefits
For the moment there are no anticipated costs.

Benefits:

 Increase in profit
 Product improvement
 Higher customer satisfaction

Data mining goals

 Use historical information about previous purchases to generate a model that links each
feature of a project to customer reaction. Companies build market strategies respective of
user feedback (market analysis).
 Identify the aspects of the product/service that attract more customers and/or keep
away people to use/ buy the product/ service.
 Identify changes in the overall opinion towards your brand. Because it provides insight
into the way your customers are feeling when they approach you, you can monitor trends
and see if overall opinion towards your company drops or rises.
 Measuring the attitude of the customer towards the aspects of a service or product which
they describe in the feedback.
 Build a system that satisfies customers feeling and satisfaction.
 Build a system that improve customer experiences and avoid data overload.

Data mining success criteria

 The minimal success criterion of the project is an accuracy greater than 90%.
 To improve quality of a product by 10%.
Project plan
The overview plan for the project is as follows.

 Phases  Time  Resources  Tasks

 Business 1 week All analyst Economic change

understanding

 Data 1 week Some database analyst Data problems,

understanding time technology problems

Data preparation - week Some database analyst Data problems,

time technology problems

Modeling -week Some database analyst technology problems,

time inability to implement
results

Evaluation - week All analysis Economic changes,

inability to implement
results

Deployment -week Data mining consultant, technology problems,

Some database analyst inability to implement
time results

Initial assessment of tools and techniques 

It is important to assess tools and techniques early in the process since the selection of tools and
techniques may influence the entire project. There are data mining tools and frameworks that
support various methods for different stages of the process of the project.