TEXT MINING
RESTAURANT
REVIEW
HELLO!
MEMBERS
- ASTHA SINGH
-SVKSRJL PRIYANKA
- AYASKANTA MOHAPATRA
- SAGAR MHATRE
- YATESH RAJPUT
- AKSHAY D. PATIL
- PRATEEK DEY
- SHIVAM BORIKAR
2
“Without data you’re just another person
with an opinion.”
- W . Edwards Deming
(BSc. Electrical engineering, MSc. Statistics Colorado Univ)
3
INTRODUCTION
• Text Mining is a Discovery
• Text Mining is also referred as Text Data Mining (TDM) and Knowledge Discovery in
Textual Database
• Text Mining is used to extract relevant information or knowledge or pattern from
different sources that are in unstructured or semi-structured form.
• Extract and discover knowledge hidden in text automatically
• Aid domain experts automatically by:
 Identifying concepts
 Extracting facts/relations
 Discovering implicit links
 Generating hypothesis
4
LITERATURE REVIEW
• Restaurant online ratings.
• Existing study of restaurant online rating has been unexpectedly scarce, previous
study on online UGC (not limited to restaurants primarily treat rating as an
antecedent (or an independent variable), with sales being most frequently studied
consequence (or dependent variable).
• Extremely limited number of studies have considered online rating a dependent
variable.
TEXT MINING OF ONLINE REVIEWS
• Text mining has enabled researchers to explore customers overall feelings after
consumption detailed description of a product or and perceived brand quality and
positioning.
• With structurised online reviews, researchers have further investigated into
profitable business issues such as viral marketing, sales prediction, product defect
discovery, and cross cultural consumer.
• It has introduced the possibility of quantitatively correlating online reviews and
online ratings
6
MORE
ABOUT
REVIEW
7
Its thought or belief about something or someone and when it becomes a
review then it can influence decision making of others.
What is an opinion ?
Where can I get these reviews about particular restaurant?
• Personal suggestion
• Travel Blogs
• Advisor websites
• Magazine
• Discussion forums
• Internet
8
How many reviews do you read
before visiting?
Do you check restaurant reviews
before visiting?
How often do you dine out?
From where do you prefer to take
reviews?
Advisory websites
Friend suggestion
Travel blog
Magazine & Discussion
forums
Google
Do you leave your review after visiting the
restaurant from above option mentioned?
Which language do you prefer to give
your review in?
Do you fill restaurant feedback
form?
Have you ever been asked by the
restaurant to fill in fake review?
English
Hindi
Native
language
A PICTURE IS
WORTH A
THOUSAND
WORDS
11
😋😋😋😋
😒😒😒😒
12
13
How is it done?
• Get the word counts or User Generated Content (UGC)
• Screening
• Analysis
• Get insights
• Action
14
Restaurant A Restaurant B Restaurant C Restaurant D Restaurant E
Word Count Word Count Word Count Word Count Word Count
not bad 6915roast duck 6114taste 5299not bad 4566not bad 3848
Taste 4623not bad 4071not bad 5044taste 3725taste 3744
food 3751taste 4057food 4418delicious 3694delicious 2910
environment 3525delicious 3420delicious 4204food 3308like 2496
delicious 3373duckling 3294queue 3657like 2440Bfengtang 2382
Nangou 2698groups 2647Walpojia 3340service 2353dumpling 1597
like 2490food 2574grandma 2223yogurt 2187food 1429
service 2228quanjude 2042like 2135Xibel 1672queue 1155
price 1768duck 2027pork 2071waiter 1584waiter 1151
waiter 1528service 1709chap 1988noodle 1482Rice roll 1107
crab 1409waiter 1701inexpensive 1843northwest 1436environment 1068
flavour 1396peking 1487price 1828roast lamb 1377Duck 1060
Zhangda 1291environment 1235tea 1653lamb 1346Chicken claw 1033
Shanghai 1250like 1083best buy 1652flavour 1230jaw 1029
friend 1226skin 1026chicken 1587lamb2 1228service 1016
oxtongue 1199RMB 943environment 1578lamb3 1000flavour 932
expensive 1189flavour 836flavour 1494environment 917Horg long 799
shrimp 1134duck bone 741green tea 1448meat 839Wujaochang 733
Tofu 1125amount 725waiter 1383gluten 772dessert 706
salad 1069dice 723spicy tofu 1140tofu 732tea café 659
roast 1042Shanghai 650service 1101recommend 719must order 632
pork 1026price 642pancake 960fish 716porridge 589
vegetable 1021dice of skin 629fish 953honey 685price 547
bake 980meat 600friend 905soup 638love 535
window 932soup 571wait 876homemade 630roast pork 535
river 916pancake 361amount 830friend 629pineapple 533
meat 880sauce 527broccoli 830lamb4 670stir fry 533
beacon 871next time 515recommend 813queue 594amount 474
groupon 831pack 477shrimp 791price 583Shrimp 449
beef 806friend 457RMB 766bun 577friend 442
Table: Most frequently used words
Obstacles
15
• Sentence Negation
• Sarcasm
• Terseness
• Language Ambiguity
What happens when your
“AI” isn’t intelligent
enough?
16
Text Mining-
It is an important technique for extracting information and key concepts from collection of
textual data.
17
Implement diff. Models
On the structured data
we may implement
models like predictive
analysis, classification
algorithms, clustering
techniques,
Various coding
languages like python,
nlp can also be used
Collect Data
Obtain data from
various sources like
feedback, customer
calling, surveying,
emails, online surveys.
Construct a Structure
After that the obtained
data needs to be
structured in such a
manner that if any
logic or function is
implemented upon
them they don’t show
error.
49,99,96,800
Whoa! That’s a big number, aren’t you proud?
That’s the amount of tweets just a day.
18
95,007,324
Instagram Posts per day
100%
Market reach to the audience on online platform.
Youtube videos everyday
19
5,76,000(2017)
Using social media for collection of reviews-
Company continually seek improve their business model through feedback and customer
satisfaction survey. Social media provide additional opportunity for explore the mind of the
customer
20
Case Study
online reviews and ratings of your restaurant can drive hungry customers to
your door or scare them away.
21
• 60% of restaurant-goers read online reviews before going out for a meal — a habit that
takes precedence over getting directions to a restaurant or looking at food photos. In
terms of preferences in types of reviews, customer-written reviews on websites like
Yelp, TripAdvisor, and Zomato are preferred by 25% more people than reviews written
by professional food critics.
• Two economists at the University of California, Berkeley found that a half-star
improvement on Yelp’s 5-star rating scale makes a restaurant 30% to 49% more likely to
be fully booked during peak dining times.
• According to ReviewTrackers research, approximately 1 in 3 diners will not choose to
eat in a restaurant with an average rating of 3 stars out of 5 (or below).
22
4 ways to grow your restaurant reviews and
performance
▫ Promote your restaurant’s presence on online review sites
▫ Ask customers for reviews
▫ Create your own landing page for reviews
▫ Conduct the Net Promoter Score (NPS) survey
23
A typical case of how ratings changed the game can be
pointed out by a local restaurant.
“GUPTA SANDWICH”
• A fast food chain started in Mumbai expanded to
its central and harbour regions rapidly and solely
on the basis of reviews and ratings it got.
(Data from tripadvisior.com shows the ratings and reviews
of the restaurant .
Here gupta sandwich got a score of 4 out of 5)
24
• Increase in public reach through online
modules which helped them understand what
is required what is not.
• Reviews and ratings worked as medium of
driving force for the owners of “Gupta
Sandwich”
• It made the platform open for all sorts of
suggestions and remarks people wanted to
give.
(Open reviews help people to decide what to choose.
Following data is taken from Zomato which rates
Gupta Sandwich a 3.9/5)
25
• People generally tend to check reviews before
visiting a place.
• And with many case the restaurants and hotels
don’t disclose it and hence get opted for.
• While open sources handles the data and keep it
open to the general public which eventually helps
in growing the entire business of that restaurant
getting good reviews and visits
(Just dial is one the older players to start their platform
where people can review their favourite places)
zomato4.2
tripadv3.7
swiggy4.0
yelp4.1
swiggy4.0
JD3.2
INFOGRAPH FOR DIFF ONLINE PORTALS
26
Based on where the customers leave
more review and how “Gupta Sandwich”
is rated
A Study of Negative Customer Online Reviews
and Managerial Responses on
Social Media
• Introduction
• Social Media Marketing within the Hotel Industry
 Review system
 Trip Advisor Reference
27
• The Impact of Social Media on Customer Decision-Making and Customer Satisfaction
Consider
Evaluate
Buy
Enjoy/Advocate
Negative reviews and Service failure
• Service Failure
• Service is unavailable
• Service is delayed
• Negative reviews
• Impact on decision making
• Affect the organization
Type of Negative reviews and Responses
• Tangible factors
• Reliability
• Empathy of staff
• Amenities
• Availability
Managerial responses to Reviews
• Excuse and offer a refund or discount
• Justification
• Apology
• Internal attribution
• External attribution
• "We", "I” and "our company" in the
text
Advantages and Disadvantages
32
PROS AND CONS OF DATA MINING
▫ Advantages
• Up-to-date feedback.
• Benchmark results
• Show that you care
• Catching problems
▫ Disadvantages
• Negative reviews
• Change of opinion depending upon
reviews
• Cost of service provided
• Malicious and damaging reviews
• Need to keep update on new
reviews
33
34
CONCLUSION
35
THANKS!
Any questions?
CREDITS
Special thanks to all the people who made and released
these awesome resources for free:
▫ Presentation template by MR. Shivam rapper
▫ Photographs and icons by Google.com
36

Text mining of reviews

  • 1.
  • 2.
    HELLO! MEMBERS - ASTHA SINGH -SVKSRJLPRIYANKA - AYASKANTA MOHAPATRA - SAGAR MHATRE - YATESH RAJPUT - AKSHAY D. PATIL - PRATEEK DEY - SHIVAM BORIKAR 2
  • 3.
    “Without data you’rejust another person with an opinion.” - W . Edwards Deming (BSc. Electrical engineering, MSc. Statistics Colorado Univ) 3
  • 4.
    INTRODUCTION • Text Miningis a Discovery • Text Mining is also referred as Text Data Mining (TDM) and Knowledge Discovery in Textual Database • Text Mining is used to extract relevant information or knowledge or pattern from different sources that are in unstructured or semi-structured form. • Extract and discover knowledge hidden in text automatically • Aid domain experts automatically by:  Identifying concepts  Extracting facts/relations  Discovering implicit links  Generating hypothesis 4
  • 5.
    LITERATURE REVIEW • Restaurantonline ratings. • Existing study of restaurant online rating has been unexpectedly scarce, previous study on online UGC (not limited to restaurants primarily treat rating as an antecedent (or an independent variable), with sales being most frequently studied consequence (or dependent variable). • Extremely limited number of studies have considered online rating a dependent variable.
  • 6.
    TEXT MINING OFONLINE REVIEWS • Text mining has enabled researchers to explore customers overall feelings after consumption detailed description of a product or and perceived brand quality and positioning. • With structurised online reviews, researchers have further investigated into profitable business issues such as viral marketing, sales prediction, product defect discovery, and cross cultural consumer. • It has introduced the possibility of quantitatively correlating online reviews and online ratings 6
  • 7.
  • 8.
    Its thought orbelief about something or someone and when it becomes a review then it can influence decision making of others. What is an opinion ? Where can I get these reviews about particular restaurant? • Personal suggestion • Travel Blogs • Advisor websites • Magazine • Discussion forums • Internet 8
  • 9.
    How many reviewsdo you read before visiting? Do you check restaurant reviews before visiting? How often do you dine out? From where do you prefer to take reviews? Advisory websites Friend suggestion Travel blog Magazine & Discussion forums Google
  • 10.
    Do you leaveyour review after visiting the restaurant from above option mentioned? Which language do you prefer to give your review in? Do you fill restaurant feedback form? Have you ever been asked by the restaurant to fill in fake review? English Hindi Native language
  • 11.
    A PICTURE IS WORTHA THOUSAND WORDS 11 😋😋😋😋 😒😒😒😒
  • 12.
  • 13.
    13 How is itdone? • Get the word counts or User Generated Content (UGC) • Screening • Analysis • Get insights • Action
  • 14.
    14 Restaurant A RestaurantB Restaurant C Restaurant D Restaurant E Word Count Word Count Word Count Word Count Word Count not bad 6915roast duck 6114taste 5299not bad 4566not bad 3848 Taste 4623not bad 4071not bad 5044taste 3725taste 3744 food 3751taste 4057food 4418delicious 3694delicious 2910 environment 3525delicious 3420delicious 4204food 3308like 2496 delicious 3373duckling 3294queue 3657like 2440Bfengtang 2382 Nangou 2698groups 2647Walpojia 3340service 2353dumpling 1597 like 2490food 2574grandma 2223yogurt 2187food 1429 service 2228quanjude 2042like 2135Xibel 1672queue 1155 price 1768duck 2027pork 2071waiter 1584waiter 1151 waiter 1528service 1709chap 1988noodle 1482Rice roll 1107 crab 1409waiter 1701inexpensive 1843northwest 1436environment 1068 flavour 1396peking 1487price 1828roast lamb 1377Duck 1060 Zhangda 1291environment 1235tea 1653lamb 1346Chicken claw 1033 Shanghai 1250like 1083best buy 1652flavour 1230jaw 1029 friend 1226skin 1026chicken 1587lamb2 1228service 1016 oxtongue 1199RMB 943environment 1578lamb3 1000flavour 932 expensive 1189flavour 836flavour 1494environment 917Horg long 799 shrimp 1134duck bone 741green tea 1448meat 839Wujaochang 733 Tofu 1125amount 725waiter 1383gluten 772dessert 706 salad 1069dice 723spicy tofu 1140tofu 732tea café 659 roast 1042Shanghai 650service 1101recommend 719must order 632 pork 1026price 642pancake 960fish 716porridge 589 vegetable 1021dice of skin 629fish 953honey 685price 547 bake 980meat 600friend 905soup 638love 535 window 932soup 571wait 876homemade 630roast pork 535 river 916pancake 361amount 830friend 629pineapple 533 meat 880sauce 527broccoli 830lamb4 670stir fry 533 beacon 871next time 515recommend 813queue 594amount 474 groupon 831pack 477shrimp 791price 583Shrimp 449 beef 806friend 457RMB 766bun 577friend 442 Table: Most frequently used words
  • 15.
    Obstacles 15 • Sentence Negation •Sarcasm • Terseness • Language Ambiguity
  • 16.
    What happens whenyour “AI” isn’t intelligent enough? 16
  • 17.
    Text Mining- It isan important technique for extracting information and key concepts from collection of textual data. 17 Implement diff. Models On the structured data we may implement models like predictive analysis, classification algorithms, clustering techniques, Various coding languages like python, nlp can also be used Collect Data Obtain data from various sources like feedback, customer calling, surveying, emails, online surveys. Construct a Structure After that the obtained data needs to be structured in such a manner that if any logic or function is implemented upon them they don’t show error.
  • 18.
    49,99,96,800 Whoa! That’s abig number, aren’t you proud? That’s the amount of tweets just a day. 18
  • 19.
    95,007,324 Instagram Posts perday 100% Market reach to the audience on online platform. Youtube videos everyday 19 5,76,000(2017)
  • 20.
    Using social mediafor collection of reviews- Company continually seek improve their business model through feedback and customer satisfaction survey. Social media provide additional opportunity for explore the mind of the customer 20
  • 21.
    Case Study online reviewsand ratings of your restaurant can drive hungry customers to your door or scare them away. 21 • 60% of restaurant-goers read online reviews before going out for a meal — a habit that takes precedence over getting directions to a restaurant or looking at food photos. In terms of preferences in types of reviews, customer-written reviews on websites like Yelp, TripAdvisor, and Zomato are preferred by 25% more people than reviews written by professional food critics. • Two economists at the University of California, Berkeley found that a half-star improvement on Yelp’s 5-star rating scale makes a restaurant 30% to 49% more likely to be fully booked during peak dining times. • According to ReviewTrackers research, approximately 1 in 3 diners will not choose to eat in a restaurant with an average rating of 3 stars out of 5 (or below).
  • 22.
    22 4 ways togrow your restaurant reviews and performance ▫ Promote your restaurant’s presence on online review sites ▫ Ask customers for reviews ▫ Create your own landing page for reviews ▫ Conduct the Net Promoter Score (NPS) survey
  • 23.
    23 A typical caseof how ratings changed the game can be pointed out by a local restaurant. “GUPTA SANDWICH” • A fast food chain started in Mumbai expanded to its central and harbour regions rapidly and solely on the basis of reviews and ratings it got. (Data from tripadvisior.com shows the ratings and reviews of the restaurant . Here gupta sandwich got a score of 4 out of 5)
  • 24.
    24 • Increase inpublic reach through online modules which helped them understand what is required what is not. • Reviews and ratings worked as medium of driving force for the owners of “Gupta Sandwich” • It made the platform open for all sorts of suggestions and remarks people wanted to give. (Open reviews help people to decide what to choose. Following data is taken from Zomato which rates Gupta Sandwich a 3.9/5)
  • 25.
    25 • People generallytend to check reviews before visiting a place. • And with many case the restaurants and hotels don’t disclose it and hence get opted for. • While open sources handles the data and keep it open to the general public which eventually helps in growing the entire business of that restaurant getting good reviews and visits (Just dial is one the older players to start their platform where people can review their favourite places)
  • 26.
    zomato4.2 tripadv3.7 swiggy4.0 yelp4.1 swiggy4.0 JD3.2 INFOGRAPH FOR DIFFONLINE PORTALS 26 Based on where the customers leave more review and how “Gupta Sandwich” is rated
  • 27.
    A Study ofNegative Customer Online Reviews and Managerial Responses on Social Media • Introduction • Social Media Marketing within the Hotel Industry  Review system  Trip Advisor Reference 27
  • 28.
    • The Impactof Social Media on Customer Decision-Making and Customer Satisfaction Consider Evaluate Buy Enjoy/Advocate
  • 29.
    Negative reviews andService failure • Service Failure • Service is unavailable • Service is delayed • Negative reviews • Impact on decision making • Affect the organization
  • 30.
    Type of Negativereviews and Responses • Tangible factors • Reliability • Empathy of staff • Amenities • Availability
  • 31.
    Managerial responses toReviews • Excuse and offer a refund or discount • Justification • Apology • Internal attribution • External attribution • "We", "I” and "our company" in the text
  • 32.
  • 33.
    PROS AND CONSOF DATA MINING ▫ Advantages • Up-to-date feedback. • Benchmark results • Show that you care • Catching problems ▫ Disadvantages • Negative reviews • Change of opinion depending upon reviews • Cost of service provided • Malicious and damaging reviews • Need to keep update on new reviews 33
  • 34.
  • 35.
  • 36.
    CREDITS Special thanks toall the people who made and released these awesome resources for free: ▫ Presentation template by MR. Shivam rapper ▫ Photographs and icons by Google.com 36