Text analysis
- 2. Agenda Objective Approach Data Preparation Model Conclusion
- 3. Objective To analyse the “La Binchoise “ béer reviews and provide useful insights to improve the brand quality.
- 4. Approach Taking out La Binchoise beer reviews from all the reviews. Splitting up positive and negative reviews. Removing stop words from all the reviews. Categorization positive, negative and all reviews into single strings each Plotting the word cloud to see the more frequently used words.
- 5. Data preparation We have to subset the dataset into high rating reviews and low rating reviews This high and low rating reviews are classified based on there rating High reviews means >3 Low Reviews means <3 Removing the stop word
- 6. Data Preparation Importing regular expression and removing unnecessary symbols and spaces Eg : ‘n’ ‘---’ ‘____’ ‘***’ etc Removing stop words from all the 3 types of reviews We will remove stop words from high and low rating reviews
- 7. Word Cloud is Generated Plot of all positive reviews Plot of all negative reviews
- 8. Model Generating corpus after tokenizing to look for the concordance of specific words Eg : In high review plot we can see that “light” is used more. To know what exactly people are talking about “light” we need to read the full sentence. For that we use concordance This type analysis is called as polarity analysis, it is useful to give us good context like as see above The polarity of a word can be identified by studying the occurrence frequency of the word in a large annotated corpus
- 9. Model Importing genism package to perform Tfidf model
- 10. Insights Average rating is 4 and mostly this beer has more positive reviews Taste is something which most people are talking about and should improve on taste as it gives lemon and fruit flavour than a usual beer flavour.