1 / 16

Sentiment Analysis Approach for Brand Reputation Evaluation

Sentiment Analysis or Opinion Mining is a subfield of Natural Language Processing that analyses user-generated content to identify the underlying sentiment of either positive, negative or neutral. Objective is to evaluate and compare the selected sentiment classification techniques used for evaluating brand reviews. The findings are presented to make informative decisions regarding the adoption of classification techniques.

dejongy
Download Presentation

Sentiment Analysis Approach for Brand Reputation Evaluation

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. Sentiment Analysis Approach for Brand Reputation Evaluation De Jong YeongB.Sc. (Hons) in Computing with Software Development

  2. Motivation • User Generated Content (UGC), e.g. reviews had increased significantly over the past decades. • Websites like Amazon consists on 30 departments and sells more than 353 million of products. • Products contain significant amount of reviews. • Requires significant amount of time to read through each review and classify them into positive, negative or neutral. • Negative reviews cause unintended consequences to businesses, can be used as a guideline to improve product or services.

  3. Introduction • Sentiment Analysis / Opinion Mining to determine the underlying sentiment of either positive, negative or neutral. • Lexicon-based Approach: Dictionary method (SentiWordNet). • Machine Learning: Multinomial Naïve Bayes. • Hybrid Approach: Latent Semantic Analysis + Linear Support Vector Classifier. • Dataset from Kaggle containing 400,000 review of unlocked mobile phone sold on Amazon was chosen.

  4. Research Areas

  5. Objective

  6. Proposed Method

  7. Data Pre-processing Data Pre-processing Pre-processed Data Output (CSV File) Raw Data VADER Sentiment Lexicon Sentiment Label

  8. Lexicon-Based Approach SentiWordNet Tokenize Review Sentiment Label Processed Data Model Evaluation Label Encoder Labelled Sentiment

  9. Machine Learning Approach TF*IDF Tokenize Review Processed Data Results Multinomial Naïve Bayes Labelled Sentiment TF*IDF: Term Frequency - Inverse Document Frequency

  10. Hybrid Approach TF*IDF LSA(n = 100) Tokenize Review Processed Data Results Support Vector Machine Labelled Sentiment Reference: Keith, B., Fuentes, E. & Meneses, C., 2017. A Hybrid Approach for Sentiment Analysis Applied to Paper Reviews. Halifax, Canada, KDD.

  11. Sprints (21st Jan – 14th April 2019) • Basic Project Setup. • Data Understanding and Cleaning. • Data Preprocessing and Preparing 01. • Data Preprocessing and Preparing 02. • Refactor and Lexicon-based Approach Sentiment Analysis. • Machine Learning Approach Sentiment Analysis. • Hybrid Approach Sentiment Analysis and Model Comparison.

  12. Issues Encountered

  13. Key Findings 1 • The accuracy of Hybrid Approach was the highest with 81.2% of correctly predicted observation. • The F1 score of Hybrid Approach was the highest, presenting with 70.2% of harmonic mean between precision and recall. • The precision score of the Lexicon-Based Approach was the lowest, giving 54.0% of correctly predicted positive observations.

  14. Key Findings 2 • The positive sentiment label in Apple and BlackBerry mobile reviews were higher, compared to the negative and neutral sentiment labels.

  15. Conclusions and Future Work • Hybrid Approach to sentiment analysis can effectively used to evaluate brand reviews that benefit businesses. • The underlying sentiment of brand reviews can be evaluated with the use of sentiment classification techniques. • Emoticons and Slang handling may be implemented to improve results of sentiment analysis.

  16. Question? Thank You!

More Related