Podium Review

Transcript

1. Podium Review Summarization Turning Reviews into Action Items Kyle DeGrave
2. Podium Helps Clients Collect Reviews ❖ What if review collecting process is too successful? Client reviews reviews
3. Too Many Reviews!
4. The Deliverable Automatic Parsing of Reviews ❖ Python module to parse reviews for key topics and sentiment ❖ Sentiment is how one “feels” about a specific topic Positive Sentiment Negative Sentiment
5. The Review Data ❖ 10,000 reviews from the automotive industry ❖ Only review text matters for our purposes 10,000 reviews
6. Raw Text to Sentiment Score “These reviews are about some really great products.” review
7. Raw Text to Sentiment Score “These reviews are about some really great products.” stop words review “These reviews are about some really great products”
8. Raw Text to Sentiment Score “These reviews are about some really great products.” stop words lemmatize review “These reviews are about some really great products” “These reviews are about some really great products”
9. Raw Text to Sentiment Score “These reviews are about some really great products.” stop words lemmatize review “review really great product” Final “These reviews are about some really great products” “These reviews are about some really great products”
10. Raw Text to Sentiment Score “These reviews are about some really great products.” stop words lemmatize review “review really great product” Final “These reviews are about some really great products” “These reviews are about some really great products” unigrams (review), (product) bigrams (review really), (really great), (great product) POS chunk (review), (really great product) RAKE (review), (great product)
11. Raw Text to Sentiment Score “These reviews are about some really great products.” stop words lemmatize review “review really great product” Final “These reviews are about some really great products” “These reviews are about some really great products” “The reviews were bad, but now they are good -0.586 +0.382 “The reviews were good, but now they are bad VADER Sentiment Analysis Correct sentiment scores
12. Actionable Insights
13. Actionable Insights
14. Actionable Insights
15. Actionable Insights weighted sum of sentiment summarize reviews
16. Actionable Insights Negative Aspects rude worst_experience service damage horrible worst_customer_service worst_dealership horrible_service engine customer service worst_service wrong battery worst_enemy worst_place disappointed bad_experience hate scam service_department phone response problems dishonest accident horrible_experience terrible_customer_service repair unprofessional Aspects Weighted Sum of Sentiment
17. Actionable Insights ❖ Service ❖ Experience ❖ Car problems ❖ Oil changes ❖ Batteries ❖ Engine ❖ high-pressure sales tactics rude worst_experience service damage horrible worst_customer_service worst_dealership horrible_service engine customer service worst_service wrong battery worst_enemy worst_place disappointed bad_experience hate scam service_department phone response problems dishonest accident horrible_experience terrible_customer_service repair unprofessional Aspects Weighted Sum of Sentiment Negative Aspects
18. Actionable Insights Negative Aspects ❖ Service ❖ Experience ❖ Car problems ❖ Oil changes ❖ Batteries ❖ Engine ❖ high-pressure sales tactics rude worst_experience service damage horrible worst_customer_service worst_dealership horrible_service engine customer service worst_service wrong battery worst_enemy worst_place disappointed bad_experience hate scam service_department phone response problems dishonest accident horrible_experience terrible_customer_service repair unprofessional Aspects Weighted Sum of Sentiment
19. ❖ Service ❖ Experience ❖ Car problems ❖ Batteries ❖ Engine ❖ Oil changes ❖ High-pressure sales tactics Actionable Insights Negative Aspects rude worst_experience service damage horrible worst_customer_service worst_dealership horrible_service engine customer service worst_service wrong battery worst_enemy worst_place disappointed bad_experience hate scam service_department phone response problems dishonest accident horrible_experience terrible_customer_service repair unprofessional Aspects Weighted Sum of Sentiment
20. Kyle DeGrave PhD, Astronomy New Mexico State University