Social Media Analytics PDF

SOCIAL MEDIA ANALYTICS Connect, Analyze and Leverage By: Prof. Atlee Fernandes DEFINITION  Social Media Analytics is an Enterprise Solution that uses social media to create:  New opportunities  Build better relationships with citizenry, customers and partners  Enhanced talent...

SOCIAL MEDIA ANALYTICS Connect, Analyze and Leverage By: Prof. Atlee Fernandes DEFINITION  Social Media Analytics is an Enterprise Solution that uses social media to create:  New opportunities  Build better relationships with citizenry, customers and partners  Enhanced talent pol  Increased resiliency and efficiency S, M, A: ASPECTS OF SOCIAL MEDIA Individual Sentiment Data Aggregation Social Media Analytics Team Unmet Needs Smart Filtering Organization Talent Discovery Meaning Extraction Society Reasoning and Decision Consumable Analytics Support Process Orchestration Crowdsensing, Stream Processing Crowdsourcing Teaming, Incentives, Motivation WHY SOCIAL MEDIA ANALYTICS? STAGES OF IMPLEMENTING SOCIAL MEDIA ANALYTICS Identify and build Connect communities around a given objective Understand the structure & dynamics Analyze of community level interactions Manage communities to Leverage achieve specific goals or business objectives KNOWLEDGE GRAPHS Node 1 Edge 1 Node 5 Node 2 Node 4 Node 3 DATA SOURCES OUTCOME FROM DATA SOURCES Consumer Preferences Consumer Consumption Consumer Feedback APPLICATIONS OF SOCIAL MEDIA ANALYTICS Brand Position Competitive Research Campaign Planning and Measurement Audience Interest and Demographic Insights Industry and Product Trend Discovery Performance Benchmarking Early Warning Crisis Management PERCEPTION GAP CRM Get more customer per customer Up Sell Cross Sell Referral Sell Identify Customer Satisfaction Index Calculate Customer Loyalty Score EVOLUTION: CRM TO SOCIAL CRM MILWARD BROWN'S BRAND Z AND OGILVY ONE LOYALTY INDEX INSIGHTS FROM SOCIAL CRM Branding Product Development Sales Leads Real Time Insights Segmentation Ogilvy One Loyalty Index Advocates Prospects Customer 360 SENTIMENT ANALYSIS  It is The field of study that analyzes people’s opinions, sentiments, evaluations, appraisals, attitudes, and emotions towards entities such as products, services, organizations, individuals, issues, events, topics, and their attributes.  Sentiment analysis is a process of extracting and analyzing the sentiment of a given text. It is a form of natural language processing (NLP) that uses algorithms to identify and categorize opinions expressed in a text. The goal of sentiment analysis is to identify the attitude of the author towards a given topic. It can be used to measure the sentiment of a customer towards a product or service or to analyze the sentiment of a text written by, say, a politician or celebrity.  Sentiment analysis involves the use of machine learning models to identify and classify sentiment in text. The models analyse the text and assign each sentence or phrase a sentiment score. The sentiment score is a numerical value that indicates the overall sentiment of the text. The sentiment score can range from -1 (very negative) to +1 (very positive). LEVELS OF SENTIMENT ANALYSIS Document Level Sentence Level Entity and Aspect level Document-level sentiment analysis Sentence-level sentiment analysis Aspect level sentiment focuses on is a process of analysing the whole evaluates the sentiment of a single analysing the sentiment of sentiment of a given document. It is statement or phrase. It can be used individual aspects of a product or a process of determining the to detect the overall sentiment of service. It involves extracting overall sentiment of a document, a text, such as whether it is aspects from text and then such as a review, article, or blog positive, negative, or neutral. It is analysing the sentiment of each post. It involves analysing the text usually done by assigning a aspect separately. It is useful for for positive, negative, and neutral sentiment score to each word in understanding customer sentiment sentiments. It is used to identify the statement and then combining on specific topics and can be used the sentiment of a document and them to determine the overall to improve customer experience can be used to gain insights into sentiment. This type of analysis can and product design. Aspect level customer sentiment, product be used to quickly analyse sentiment analysis can be used to sentiment, and overall market customer feedback, identify trends identify customer needs, sentiment. in public opinion, and detect preferences, and complaints, as well sentiment shifts in conversations. as to gain insights into customer sentiment on a more granular level. SOLVING SENTIMENT ANALYSIS SENTIMENT LEXICONS  Sentiment lexicons are a type of natural language processing (NLP) tool used to analyse the sentiment of text. They are composed of words or phrases that have been assigned a sentiment score, which can range from positive to negative. For example, good, wonderful, and amazing are positive sentiment words, and bad, poor, and terrible are negative sentiment words. Apart from individual words, there are also phrases and idioms, e.g., cost someone an arm and a leg. SENTIMENT CALCULATION: LEXICON BASED APPROACH Mark Apply Handle Sentiment Aggregate Sentiment But Words & Options Shifters Clauses Phrases STEP 1: MARK SENTIMENT WORDS & PHRASES  For each sentence that contains one or more aspects, this step marks all sentiment words and phrases in the sentence. Each positive word is assigned the sentiment score of [ +1 ] and each negative word is assigned the sentiment score of [ -1 ]. “The voice quality of this phone is not good, but the battery life is long” “The voice quality of this phone is not good [ +1 ], but the battery life is long” STEP 2: APPLY SENTIMENT SHIFTERS  Sentiment shifters are words and phrases that can change sentiment orientations. There are several types of such shifters. Negation words like not, never, none, nobody, nowhere, neither, and cannot are the most common type of sentiment shifters.  Modal auxiliary verbs (e.g., would, should, could, might, must, and ought) are also some examples. “The voice quality of this phone is not good [ +1 ], but the battery life is long” “The voice quality of this phone is not good [ - 1 ], but the battery life is long” STEP 3: HANDLE BUT CLAUSES  A sentence containing a contrary word or phrase is handled by applying the following rule: The sentiment orientations before a “but” or contrary word and after the contrary word are opposite to each other as long as the opinion still stays the same. e.g., “Product A is great, but Product B is better.” Words or phrases that indicate contrary need special handling because they often change sentiment orientations too. “The voice quality of this phone is not good [ - 1 ], but the battery life is long” “The voice quality of this phone is not good [ - 1 ], but the battery life is long [ +1 ]” STEP 4: AGGREGATE OPTIONS  This step applies an opinion aggregation function to the resulting sentiment scores to determine the final orientation of the sentiment on each aspect in the sentence. As we understand that sentiment analysis can be processed at various levels, sentiment lexicons work at aspect or entity level giving hard-core evidence to an opinion shared.  Let the sentence be s, which contains a set of aspects {a1, …, am} and a set of sentiment words or phrases {sw1, …, swn} with their sentiment scores. where, ai is the ith aspect of a particular product or service s is the sentence being observed swj is the jth sentiment word in the sentence swj  so is the sentiment score of the jth word dist(swj, ai) is the distance between aspect ai and sentiment word swj in s ISSUES WITH SENTIMENT LEXICONS Ambiguity Interrogation Sarcasm Facts Sentiment lexicons are Interrogative sentences are Sentences that contain Sentences without collections of words that questions that are used to words or phrases with a sentiment words may still are associated with either ask for information.These negative sentiment score convey emotion, but they a positive or negative sentences may contain can be classified as do not contain any words sentiment. However, the sentiment words but may sarcastic. Sarcastic that are explicitly orientation of these words not express any sentiment sentences are used to associated with sentiment. can differ depending on being a question.They convey a message in an the application domain. typically begin with a word ironic way, often to make a Interrogation such as "what," "where," point or to mock "when," "why," or "how." someone. For Example: “This For Example: “This For Example: “Can you For Example: “What a washer uses a lot of camera sucks,” but “This tell me which Sony camera great car! It stopped water” and “After sleeping vacuum cleaner really is good?” and “If I can find working in two days.” on the mattress for two sucks.” a good camera in the shop, days, a valley has formed in I will buy it.” the middle”. OPINIONS OPINION QUADRUPLE OPINION QUADRUPLE  Posted by: John Doe: August 10, 2021  “I bought a Fujifilm X-T4 camera six months ago. I just love it and use it very often for my travel blogs. The image quality is amazing and the battery life is long too. However, my wife thinks it is too heavy for her.” TYPES OF OPINIONS Direct Opinion Indirect Opinion Comparative Opinion Explicit Opinion Implicit Opinion SUBJECTIVITY & EMOTIONS An objective A subjective sentence sentence presents some expresses factual some personal information feelings SENTIMENT EVALUATION Rational Evaluation Rational evaluation of sentiments is the process of evaluating a sentiment using logical reasoning and objective facts. This type of evaluation involves looking at data, facts, and evidence to determine the validity of a sentiment. Emotional Evaluation Emotional evaluation of sentiments is the process of evaluating a sentiment based on feelings and emotions. This type of evaluation takes into account how people feel about a particular situation or issue. It looks at how people react emotionally to certain events or ideas. AUTHOR READER PARADOX  The Author-Reader Paradox is a phenomenon observed in sentiment analysis where the sentiment expressed by an author and the sentiment inferred by a reader can be vastly different. This paradox arises due to the subjective nature of language, which makes it difficult for machines to accurately identify sentiment. While the algorithms in sentiment analysis are effective at recognizing certain types of sentiment, they often fail to recognize more subtle nuances that may be present in the text. For example, an author may express a positive sentiment but the reader may interpret it as negative. This is because the reader may not have the same context or understanding of the author’s words as the author does. The reader may also misinterpret the author’s intent due to their own biases and preconceived notions. OPINION RULES Sentiment word or phrase Decreased and increased quantity of an opinionated item High, low, increased and decreased quantity of a positive or negative potential item Desirable or undesirable fact (Objective Opinion) Deviation from the norm or a desired value range Produce and consume resource and waste STRUCTURED AND UNSTRUCTURED DATA  Structured data in social media is data that has been  Unstructured data from social media is incredibly organized into a specific format, such as a important for understanding consumer behaviour spreadsheet or database. This type of data is easy to and sentiment. Social media has become a major analyse and interpret because it follows a predefined platform for communication, with billions of users structure. Examples of structured data in social around the world sharing their thoughts, opinions, media include user profiles, posts, comments, likes, and experiences. By leveraging unstructured data shares, and followers. Unstructured data in social from social media, businesses can gain valuable media is data that does not have any specific format insights into customer preferences, product feedback, or structure. This type of data can be more difficult brand sentiment, and more. to analyse and interpret because it does not follow a predefined structure. SENTIMENT CALCULATION: MACHINE LEARNING  Machine-learning enables tools to assess multiple factors at the same time, and uncover how they interact to create patterns. Machine learning can be used in social media in a variety of ways. For example, it can be used to improve user experience by providing personalized content recommendations and filtering out spam or inappropriate content.  Recommendation Engine: Recommendation engines are algorithms that use data about a user’s interests, activities, and connections to suggest content or people they may be interested in. For example, a recommendation engine might suggest a new article to read based on the topics a user has previously interacted with, or suggest a new friend to connect with based on mutual interests.  Gmail Filters: Gmail filters are machine learning based automated processes that use artificial intelligence to identify patterns in incoming emails and apply predetermined rules to them. Machine learning algorithms analyse the content of emails, such as sender, subject line, body text, and attachments, to determine whether an email should be flagged for further action. It can also detect emails from specific senders and apply labels or move them to a specific folder. T-MOBILE USE CASE: WITH BASIC SENTIMENT ANALYSIS T-MOBILE USE CASE: WITH MACHINE LEARNING