Unit 6 Honors in AIML-SOCIAL MEDIA ANALYTICS ppt.pptx

SMA Notes

1. UNIT 6 Opinion Mining & sentiment Analysis Prof. Sampada Lovalekar Assistant Professor Dept. of Information Technology, SIES Graduate School of Technology 1

2. The problem of opinion mining Opinion mining, or  sentiment analysis , is a   text analysis  technique that uses computational linguistics and natural language processing to automatically identify and extract sentiment or opinion from within text (positive, negative, neutral, etc.). It allows you to get inside your customers’ heads and find out what they like and dislike, and why, so you can create products and services that meet their needs. When you have the right tools, you can perform opinion mining automatically, on almost any form of , with very little human input needed.  2

3. Opinion Mining Techniques & Types Opinion mining and sentiment analysis models can focus on polarity of opinion (positive, negative, neutral), personal feelings (angry, happy, sad, etc.), and intentions or objectives ( interested  or  not interested ).  Types of opinion Mining Fine-grained sentiment analysis Emotion detection Aspect-based sentiment analysis Multilingual sentiment analysis 3

4. 1. Fine-grained sentiment analysis The most common use of opinion mining works to categorize comments and statements on a scale of opinion polarity. This can be simply positive, negative, or neutral, or you can go beyond this into fine-grained sentiment analysis with a larger scale of categories that include: Very positive Positive Neutral Negative Very negative This is commonly used in opinion polls or surveys, as: Very Positive = 5 stars Very Negative = 1 star 4

5. 2.Emotion detection This is opinion mining aimed at finding and extracting specific emotions (anger, disappointment, irritation, happiness, etc.) from text. Some emotion detection tools use lexicons, or word lists defined by the emotions they denote. This can be problematic as some words that often convey negative emotions, like  bad  or  kill  could also be used to express happiness or approval: “Your brand is killing it!” 5

6. 3.Aspect-based sentiment analysis When opinion mining text about your brand, you’ll probably want to organize it into categories. If you’re  analyzing customer feedback , for example, you’d be able to categorize the text into aspects, like  Usability, Features, Shipping , etc., then analyze each statement as positive, negative, or neutral. 6

7. 4.Multilingual sentiment analysis Multilingual sentiment analysis is often very difficult, as it involves a lot of preprocessing and resources. Some resources, like sentiment lexicons, are available online, while some, like translated corpora and noise detection algorithms, have to be built. And you’ll need coding experience to put them into practice. 7

8. Top Opinion Mining Applications in Business Some of the most popular opinion mining sentiment analysis applications are: Social media analysis Brand awareness Customer feedback Customer service   Market research Evaluating marketing campaigns 8

9. Social media analysis With approximately  6,000 tweets sent every second  and  2.7 billion monthly active users (MAUs) on Facebook , the speed at which opinions move on social media is staggering. Just think how often your business is mentioned on social media. It would be downright impossible to monitor social media mentions manually – there are just too many. Thankfully,  machine learning tools  allow you to perform social listening and social media opinion mining constantly, and in real time, so you’ll never miss a mention. Opinions about brands and products offered on social media are often the most truthful because the user feels compelled to offer them, unsolicited.  9

10. Brand awareness Brand awareness is all about brand recognition – what does your business name, logo, etc. mean to the public? What emotions and feelings do different aspects of your business elicit? Having a strong brand image that evokes positive emotions is essential to remaining competitive.  Monitor news stories, social media, blogs, forums, etc., and find out what words and feelings your brand is associated with. Keep track of your brand over time or take a peek into your brand image at any given moment to monitor your progress. 10

11. Customer feedback Follow conversations about your product or brand and find out what’s working and what may need some work. This can be from surveys or customer service tickets, but you can also tap into all the public feedback available from online reviews, social media, and other web chatter. Even if statements or feedback aren’t directly targeted ‘@’ your company, you can opinion mine mentions of and conversations about your brand.  11

12. Customer service Customer service can make a break a business, but sometimes companies don’t even realize they have poor customer service. Opinion mining can help to regularly analyze your customer service department’s communication with the public from chatbots, emails, online tickets, even phone calls. Are they using a consistent “corporate tone-of-voice?” Are the issues regularly resolved, or do your customers leave unhappy? You can also collect regular  customer service feedback  from in-app or online surveys, customer support tickets, social media, and more, and automatically analyze it. Because, even more than just analyzing how your employees interact with customers, it’s important to understand how the customers  feel  about the interactions. 12

13. Document Sentiment Classification  aims to classify the whole document text as expressing a positive, neutral or negative opinion. This task would be helpful in recommender systems and business intelligence software, where customer negative or positive feedback could be quickly found. Intellexer   Sentiment Analyzer  uses a wide range of text features for document sentiment classification: Part of speech tags Opinion words and syntactic dependency between them Opinion objects with associated sentiment phrases Positions of sentiment sentences through the document text 13

14. Opinion Lexicon Expansion It is  a word-level classification model for automatically generating a Twitter-specific opinion lexicon from a corpus of unlabelled tweets . The tweets from the corpus are represented by two vectors: a bag-of-words vector and a semantic vector based on word-clusters trained with the Brown clustering algorithm. 14

15. 15

16. Examples of positive sentiment words are beautiful, wonderful, and amazing. Examples of negative sentiment words are bad, awful, and poor. Apart from individual words, there are also  sentiment phrases and idioms , e.g., cost someone an arm and a leg. Collectively, they are called sentiment lexicon (or opinion lexicon). 16

17. 17

18. 18

19. 19

20. 20

21. 21

22. 22

23. 23