CONSUMER SENTIMENT ANALYSIS ON RESTURANTS IN

DELHI USING TWITTER DATA

FRP Synopsis

Submitted by

Harshita Bansal

Roll number: 435/2018

Under the guidance of

Dr. Shivani Bali

Lal Bahadur Shastri Institute of Management, Delhi

INTRODUCTION
The wide spread of World Wide Web has brought a new way of expressing the sentiments of
individuals. It is also a medium with a huge amount of information where users can view the
opinion of other users that are classified into different sentiment classes and are increasingly
growing as a key factor in decision making. The process of analysing such a big data and
summarizing the opinions expressed in these huge opinionated user generated data is usually
called sentiment analysis. Hence, in this paper, we analyse the reviews given by the customers
of the restaurant with the help of machine learning classification algorithms which help us to
classify the tweets where opinions are highly unstructured and are either positive or negative.

The main goal of analysing sentiment is to study the reviews and examine the scores of
sentiments. People mostly depend upon user-generated content for decision making. Sentiment
analysis tells the user whether the information about the product is satisfactory or not before
they buy it. Marketers and firms use this analysis data to understand about their products or
services in such a way that it can be offered as per the user’s requirements. Sentiment analysis
is usually done at different levels varying from coarse level to fine level. Coarse-level sentiment
analysis deals with determining the sentiment of an entire document and fine level deals with
attribute level sentiment analysis. Sentence-level sentiment analysis comes in between these
two.

Sentiment analysis can be done either using a knowledge-based approach or machine learning-
based approach. Under knowledge-based approach, we can analyse the sentiment by either
using language processing algorithms or Lexicon methods. The machine learning-based
approach can further be classified as unsupervised and supervised learning. The language
processing algorithms are used to extract features like word frequency, parts of speech tags,
opinion words, and phrases. However, the supervised machine learning algorithms learn the
polarity (positive, negative, or neutral) of the review from a dataset which is initially classified
by a human. In Lexicon-based approach, it uses sentiment dictionary with opinion words and
matches them with the data to determine the polarity. Later, scores are assigned to the opinion
words based on how positive or negative the words contained in the dictionary are. The
machine learning-based approach along with Lexicon based approach is the most commonly
used technique for classifying the sentiment. But it has been claimed that these techniques do
not perform as well in sentiment classification as they do in topic categorization. This is
because the nature of an opinionated text which requires more understanding of the text while
the occurrence of some keywords could be the key for an accurate classification.
The proposed work is to analyse the opinion of customers on a restaurant service. It uses the
natural language processing, text analysis, and computational techniques to automate the
classification of sentiment from the reviews.

LITERATURE REVIEW
As users continue to post a large amount of textual information on various social media sites,
there is a growing interest in using automatic methods such as text mining and sentiment
analysis to process large amounts of user-generated data and extract meaningful knowledge
and insights. The hospitality industry has actively used social media platforms like Twitter and
Facebook as effective marketing tools to improve brand awareness or promote products
or services (Bilgihan et al., 2014; Jayawardena et al., 2013; Kwok and Yu, 2013; Nunko et al.,
2013). Data pre-processing should be done to improve the quality of the structure of the raw
sentence. Techniques like cosine similarity and latent semantic analysis are used for sentiment
analysis (Ramachandran and Gehringer, 2011). During the process four major problems like
opinion extraction, document sentiment classification, word sentiment classification,
subjectivity classification. For subjectivity classification techniques like cut-based classifier,
Naïve Bayes classifier, Multiple Naïve Bayes classifier, similarity dependent were highlighted
(Tang et al, 2009). To overcome the difficulty in lexical substitution, a model was developed
which uses word space formalism. This model represents the local context of the word with its
overall distribution (Baroni et al., 2009)
The Bayesian algorithm was improved by introducing a model which efficiently performs
feature selection, weight computation, and classification. Here, the weights of the classifier are
adjusted using unique features and representative features. Unique feature helps in
distinguishing classes and representative feature represents the class. Using those adjusted
weights, the probability of each classification was calculated and hence improving the Bayesian
algorithm. Among various classification techniques SVM classifier resulted in the highest
accuracy of 94.56% for the given dataset (Krishna, Akhilesh, Aich and Hegde, 2019). In the
collection of tweets related to restaurant reviews majorly positive tweets referred to food
quality, many negative tweets suggested problems associated with service quality or food
culture (Park S., Jang J., Michael C., 2016).

OBJECTIVES
 Classifying user reviews on twitter using sentiment analysis techniques as positive,
negative and neutral.
 Helping restaurants to incorporate AI into their decision making in order to improve
profitability.

RESEARCH METHODOLGY
Twitter offers the tweet data-streaming tool, Application Programming Interface (API), for
researchers and developers. The interface provides users with access keys and tokens to access
approximately 1 per cent of all tweet data. Key words like “delhi”, “awesome food”, “bad
experience”, “lovely place”, “not good” will be used. The data will be pre-processed to create
tokens (tokenization) and collected into bag of words. The data will be then segregate into
training and testing data for further classification. Classification will be done using different
machine learning techniques like Naïve Bayes classifier, Support Vector Machines, Decision
Tree, Random Forest etc would be used to check which algorithm gives the best results with
maximum percentage accuracy.
REFERENCES
 Bilgihan, A., Peng, C. and Kandampully, J. (2014), “Generation Y’s dining information
seeking and sharing behavior on social networking sites”, International Journal of
Contemporary Hospitality Management, Vol. 26 No. 3, pp. 349-366.
 Baroni M, Pucci D, Cutugno F, Lenci A (2009) “Unsupervised lexical substitution with
a word space model”. In: Proceedings of EVALITA workshop, 11th congress of Italian
Association for Artificial Intelligence, Citeseer
 Krishna A., Akhilesh V., Aich A. and Hegde C. (2019), “Sentiment analysis of
restaurant reviews using machine learning techniques”.
 Park S., Jang J., Michael C., (2016),"Analyzing Twitter to exploreperceptions of Asian
restaurants", Journal of Hospitality and Tourism Technology, Vol. 7 Iss 4 pp. 405- 422
 Ramachandran L, Gehringer EF (2011) “Automated assessment of review quality using
latent semantics analysis”. In: ICALT. IEEE Computer Society, pp 136–138.
 Tang H, Tan S, Cheng X (2009) “A survey on sentiment detection of reviews”. Expert
Syst Appl 36:10760–10773.