Scribd
Upload
1 views1 page

ISKCON Case Questions

The document outlines a series of analytical questions related to reviews for ISKCON, focusing on aspects such as the length of positive versus negative reviews, the likelihood of re-reviewing based on sentiment, and the semantic similarity of reviews. It also includes inquiries about seasonal trends in review counts, the impact of weekends on visitor numbers, feature analysis using Tf-Idf, and the application of machine learning models for sentiment classification. Additionally, it suggests the development of a dashboard for automating review collection and sentiment analysis monitoring.

Uploaded by

hillol kashyap
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
1 views1 page

ISKCON Case Questions

The document outlines a series of analytical questions related to reviews for ISKCON, focusing on aspects such as the length of positive versus negative reviews, the likelihood of re-reviewing based on sentiment, and the semantic similarity of reviews. It also includes inquiries about seasonal trends in review counts, the impact of weekends on visitor numbers, feature analysis using Tf-Idf, and the application of machine learning models for sentiment classification. Additionally, it suggests the development of a dashboard for automating review collection and sentiment analysis monitoring.

Uploaded by

hillol kashyap
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 1

ISKCON Case Analysis Questions

Q1. Are positive reviews (5 & 4) longer than negative reviews (2, and 1)? How frequent are the
expression of specific emotions in the reviews (e.g., anger, fear, trust, joy, disgust, sadness, surprise,
and anticipation, NRC Lexicon)
Q2. Test the hypothesis that visitors who post negative reviews are less likely to review again than
those who post positive reviews. (dissatisfaction – negative reviews –churn)
Q3. "All happy families are alike; each unhappy family is unhappy in its own way." – Tolstoy in Anna
Karenina. The quote suggests that positive reviews are more like each other than negative reviews.
Check if that is true for positive vs. negative and 5 vs. 1 rated review. (Convert reviews to document
embeddings using a sentence transformer (like in BERTopic modelling) and compute average
pairwise cosine similarity (i.e., semantic similarity)). Also test the preceding hypothesis for ten most
frequent words in positive vs. negative reviews and 5 vs. 1 review.
Q4. Is there seasonality across the months of a year. Plot the monthly review count across months to
determine if there is a tourist season. Also plot proportion of negative reviews across the months of a
year to see whether the tourist season differs in the mix of positive and negative reviews.
Q5. Do weekends get more visitors to the temple? Plot review count and share of negative reviews
across days of the week (weekend vs. weekdays).
Q6. What are the top 20 Tf-Idf features? Tabulate their tf-idf and tf (excluded stop words: default and
custom beforehand)
Q7. Does the frequency distribution of top features differ across the classes (positive, negative, and
neutral)? Do Chi-sq test of independence to find out. For the test, select 2 positive (negative)
interesting features from the 5 most frequent features in positive (negative) reviews.
Q8. What top features account for positive vs. negative sentiments? What are satisfiers and
dissatisfiers (e.g., food and crowding). For a few aspects of interest (e.g., food and dress), find out
what are the positives vs. negatives? Use topic modeling to define/select important aspects/topics. You
may seed topics and analyze positive vs. negative reviews assigned to those topics with high
confidence.
Q9. Use supervised ML models, Multinomial Naïve Bayes and SVM, to predict the four classes of
sentiment (human labeled) (refer to similar prior work with IMDb movie reviews). To deal with the
imbalance in data, use up sampling, down sampling, and the synthetic minority oversampling
technique (SMOTE). Which of the preceding three techniques yields the best results?
Q10. Suggest a dashboard design (e.g., shiny app) for automating review collection (using web
scraping and surveys) and sentiment analysis. Also monitoring review count, proportion of negative
reviews, counts/proportions of specific emotions of interest, and change in the proportions of topics of
interest over time.

You might also like

pFad - Phonifier reborn

Pfad - The Proxy pFad of © 2024 Garber Painting. All rights reserved.

Note: This service is not intended for secure transactions such as banking, social media, email, or purchasing. Use at your own risk. We assume no liability whatsoever for broken pages.


Alternative Proxies:

Alternative Proxy

pFad Proxy

pFad v3 Proxy

pFad v4 Proxy