160 likes | 259 Views
Kevaughn Gordon Dr. Rakesh Verma. The Classification of Online Reviews. Upcoming Vacation. Review Sites. Honest?. Or Deceptive?. The Problem. Text Classification Determining if an online review is fraudulent or genuine. . How?.
E N D
Kevaughn Gordon Dr. Rakesh Verma The Classification of Online Reviews
The Problem • Text Classification • Determining if an online review is fraudulent or genuine.
How? These books are fantastic. I got the first book in the series on a Friday night, Saturday morning I was at the bookstore for more. The books have elements of both the supernatural and some twists and turns in like a good "who done it" so if you like either I would recommend them. As for the writing style; Mr. Butcher paints very vivid pictures, making it easy to get lost in the story, and the way he writes the characters made them quite believable. No "knight in shining armor" who makes no mistakes and has no fear; Harry Dresden has his issues and was a very believable person as a result. A fun a fantastic read, I hope the series continues. And don't make my mistake of just getting the first one. Get them all at once and save yourself the trouble!
How? These books are fantastic. I got the first book in the series on a Friday night, Saturday morning I was at the bookstore for more. The books have elements of both the supernatural and some twists and turns in like a good "who done it" so if you like either I would recommend them. As for the writing style; Mr. Butcher paints very vivid pictures, making it easy to get lost in the story, and the way he writes the characters made them quite believable. No "knight in shining armor" who makes no mistakes and has no fear; Harry Dresden has his issues and was a very believable person as a result. A fun a fantastic read, I hope the series continues. And don't make my mistake of just getting the first one. Get them all at once and save yourself the trouble!
Features • A document is a collection of features • We humans can recognize the features of a positive (or negative) review and classify based on their presence or absence.
More Examples • After recent week stay at the Affinia Hotels, I can definitely say i will be coming back. They offer so many in room amenities and services, Just a very comfortable and relaxed place to be. My most enjoyable experience at the Affinia Hotel was the amazing customization they offered, I would recommend Affinia hotels to anyone looking for a nice place to stay . • The Affinia is virtually the perfect hotel for a visit to Chicago. My wife and I stayed in an upgraded suite. Very large room, that is modern and has all the little touches. The staff was extremely helpful and professional. The location is excellent. The price is a good value for what you get. We simply couldn't find anything to criticize. Highly recommended!
? • After recent week stay at the Affinia Hotels, I can definitely say I will be coming back. They offer so many in room amenities and services. Just a very comfortable and relaxed place to be. My most enjoyable experience at the Affinia Hotel was the amazing customization they offered, I would recommend Affinia hotels to anyone looking for a nice place to stay. • The Affinia is virtually the perfect hotel for a visit to Chicago. My wife and I stayed in an upgraded suite. Very large room, that is modern and has all the little touches. The staff was extremely helpful and professional. The location is excellent. The price is a good value for what you get. We simply couldn't find anything to criticize. Highly recommended!
Features • A document is a collection of features • We humans can recognize the features of a positive (or negative) review and classify based on their presence or absence. • However we're not so good at recognizing the features of genuine (or fraudulent) review. Therefore
Machine Learning • We can construct a program that can do what we can't • Supervised Learning
Machine Learning: Complications • How should the features be selected? • What machine learning algorithm should we use?
Feature Selection N-Gram: Unigrams, Bigrams, Trigrams Attribute Evaluation: Chi square, Information gain, Gain Ratio, Symmetrical Uncertainty Subset Evaluation
Classifiers Neural Network Support Vector Machine Decision Trees Naive Bayes