July 14, 2017 by
The Women's March was a worldwide protest on January 21, 2017, to advocate legislation and policies regarding human rights and other issues, including women's rights, immigration reform, healthcare reform, reproductive rights, the natural environment, LGBTQ rights, racial equality, freedom of religion, and workers' rights. For demonstrating the use case for descriptive analysis on text data, we extracted the data of speech transcripts of multiple speakers of Women's March. The speakers include Angela Davis, Madonna, Gloria Steinem, Ashley Judd, and America Ferrera. We cleaned the data set by removing stop words, fixing normalization of keywords, and performed tokenization. The free flowing text tokens were then aggregated, linked together and run through our theme, sentiment and tone detection in order to unearth insights with respect to different speakers. For theme and sentiment detection, we used IBM Watson that takes the text as input and outputs the tone and sentiment of the text. All of the derived data was then intermingled together with the real domain knowledge. Contexts from their speeches, their themes, and keywords used were checked in other datasets such as twitter, news postings etc. In short, the following process represents the analysis flow: 1. Dataset Curation 2. Dataset Cleaning 3. Text Tokenization 4. Token Normalisation and Aggregation 5. Theme, Sentiment and Tone Detection 6. Contextual Insights using Business Analysis 7. Domain Knowledge Insights Here are the results: The overview: The term “Trump” was directly mentioned by his name in the speeches of Angela, Gloria, and Ashley in contrast to speeches of Madonna and Ferrara where “president” keyword was used to refer Donald Trump. The term “People” was used mostly by all the women, the Phrases include: "It is not - I the president, It is - We, the people (repeated thrice in Ferrera’s speech)" "Freedom struggles of black people", "Resistance to attacks on disabled people", and "Danger faced by marginalised people" Context of “I” vs “We” The term "I" was mentioned in the following statements: “I'm angry. Yes, I am outraged”- Madonna, "I am deeply honored to march with you today"- America Ferrera, “I am a nasty woman. I'm as nasty as a man who looks like he bathes in Cheetos dust”-Ashley Judd "We" was mentioned in these statements: “We dedicate ourselves to collective resistance”- Angela Davis, “We choose love. We choose love. We choose love”-Madonna, “We are here and around the world for a deep democracy that says we will not be quiet, we will not be controlled, we will work for the world in which all countries are connected”-Gloria Steinem, “We are America”- America Ferrera Top Common Themes: Main themes observed from the speeches are - Racism, Immigration, Religion and Women's Rights. The general sentiment and tone associated with these themes are negative (specifically sadness & anger on the current situation of Muslims, Immigrants, Blacks, and Women). Top Observed Tones and Personality Trait: From the overall analysis - Sadness, Disgust, and Fear are the top three observed emotional tones. Openness was observed as the major personality trait. Here is the analysis speaker by speaker 1. Angela Davis: Top Quote: “The next 1,459 days of the Trump administration will be 1,459 days of resistance.” Her top themes expressed in the speech are - Racism & Slavery, Human Rights, Violence, Religion & Immigration, Liberalism. “Resistance” was the most used word by Angela in her speech. Her top personality trait observed from the speech was “Conscientiousness” (Acting in a thoughtful way). Context - Angela Davis appealed for "Collective Resistance": Resistance to the attacks on Muslims and on immigrants, disabled people etc..She called out Trump administration will be 1,459 days of resistance Overall Sentiment - Negative (Disgust and Sadness as the emotional tone) Top Positive Words Used: ‘supremacy', 'united', 'freedom', 'rising', 'celebrate', 'thank', Top Negative Words Used: ‘murder', 'worse', 'dying', 'demands', 'attacks', 'resistance', 'struggles', 2. Madonna Top Quote by Madonna: “Welcome to the revolution of love. To the rebellion. To our refusal as women to accept this new age of tyranny. ” Her top themes expressed in the speech are -Love, White House, Unity, Tyranny. “Love” was the most used word by Madonna in her speech of total 5:10 minutes and 358 words. Her observed personality trait from the speech was also “Conscientiousness” which means - The tendency to act in an organised or thoughtful way. Context - Madonna appealed people to not fall into despair and choose Love Overall Sentiment - Positive ( -0.05 in the range of -1 to +1) Top Positive Words Used: ‘hallmark', 'good', 'right' Top Negative Words Used: ‘false', 'danger', 'refusal', 'fuck', 'shake', 'awful' Most observed Emotional Tone: Anger, likely due to her mentions most negative mentions on “White House” Top Madonna’s Tweets about the rally - Yesterday's Rally was an amazing and beautiful experience. I came and performed Express Yourself and that's exactly - Express Yourself...............So you can Respect Yourself. On Stage at the Women's March In D.C. - With My Girl Amy at the Women's March in D.C. We Go Hard or We Go Home. To view the analysis of other speakers, please share your email id. We will be happy to email it. Feel free to share your views in the comments section. :)

June 23, 2017 by
Let’s look at few insights related to social and digital media: - 22% of the world’s total population uses Facebook (Source) - Over 50 million businesses use Facebook Business Pages. (Source) - 2 million business use to Facebook for advertising (Source) - 88% of businesses with more than 100 employees use Twitter for marketing purposes (Source) - 38% organizations spend 20% more than their advertising budgets on social channels (Source) There is no doubt that social media is an essential driver for a brand’s success. However, not many brands are using analytics and data science to gain social media benefits. Even if they are using it, are not aware of key areas of analytics. In this article, I will talk about eight aspects in which brands can leverage social media analytics for their advantage. 1. 360 degrees Brand Tracking Brand tracking is the process of monitoring the presence of a brand - their activities, their competitor activities, consumer trends, customer behaviors etc.  across the entire social landscape. This cross-platform and in-depth monitoring of a brand provides invaluable insights which give them a competitive advantage as well as keeps them up-to-date with their audience needs. For example, using brand tracking, detailed data-driven insights can be used by the brands for an added advantage such as: What is the right platform, right time and right method to reach out to specific customers, How good is their promotional and offers strategy as compared to multiple competitors in same space. 2. Target Audience/Customer/Followers Analysis The audience is an integral part of a brand’s success. Audience analysis includes a deeper analysis of a brand’s followers, fans, and customers in different verticals such as audience engagement, audience sentiment, and audience influence etc. Audience analysis requires different sources of data such as social profiles, timelines, survey records, transactional records and demographics data. Audience analysis unearths insights related to segmentation, demographic, mindshare, sentiment, top questions, top needs, top queries and themes associated with the customer. 3. Content Analysis Content analysis of social media posts (such as tweets, posts, emails, blogs etc.) of a brand could provide myriad of recommendations and facts that will ensure higher engagement and traffic. For example, by using natural language processing and machine learning on historical, present and competitor data, one could suggest right keywords and hashtags, optimal length and post type (pic, video, link), best times to post, so that a brand would garner higher and effective response. 4. Influencers Identification - Global and Brand Specific Influencer analysis includes identification and ranking of celebrities and personalities across different dimensions such as - engagement rate, overall digital presence, followers acquisition and engagement rates. Influencers are important entities that share and endorses the brands and their products, analytics helps to segregate right influencers for right brands with associated insights. 5. Promoted Post-Detection Using classification models and sophisticated feature engineering techniques on the data of competitor social media posts, it can be predicted that which of the tweets or posts are promoted by a brand. Promoted post detection can help in demystifying competitor’s promotional, boosting and monetary strategies. By understanding these strategies, a brand looks closely into their own approaches and optimize the practices they follow. 6. Impression Burnout Optimization Advertisements are the key part of brand’s marketing strategies online, However, every posted advertisement is not completely effective, people start losing interest in an Ad with time. The likelihood of people clicking an ad decreases with time - this is called impression burnout. Analytics can help to optimize the features and content of an ad so that the burnout time of an ad is higher. Data Science can predict what is the true burnout period for an Ad, so that brand can either monetize the good ads and discard the bad ads. 7. Engagement Prediction ML models can be trained to predict what is the likely engagement of their posts, if a post is expected to gain higher engagements, a brand can promote it more, or vice versa - if a post is performing expected to perform too bad, they can either alter its content or completely discard it with a new post. 8. Cross Platform Correlation Analytics and Data Science can be used to identify which of the cross-platform variables are correlated with each other. For example, a restaurant in the US saw higher positive reviews on yelp.com when they started promoting personalized offers on Twitter. Another example is the increase in email opening rates of a brand when they initiated a social media sales across Facebook, Twitter, and Instagram. With the massive boom in social media interactions, brands have to identify right patterns using analytics and data science so that they can adopt a proactive and intelligent approach towards meaningful, significant social media success. Feel free to share your thoughts in the comments section and share this blog. :)  

June 21, 2017 by
Telecom companies often ignore best practices about basic data analytics and go directly to adopting artificial intelligence and other advanced technologies.  When set properly with Machine Learning, Telecommunication Company can predict with 75 times more accuracy whether its customers are about to leave. But they could only achieve this if already automated the processes that made it possible to contact customers quickly and understand their preferences by using more standard analytical techniques. Operators that rush into sophisticated artificial intelligence before reaching a critical mass of automated processes and structured analytics can end up paralyzed – and  in the middle of nowhere.   So here are 3 things to check before going for AI and other advanced technologies : ·        Automating basic processes Managers should ask themselves if they have automated processes in problem areas that cost significant money and slow down operations. Companies need to automate repetitive processes involving substantial amounts of data — especially in areas where intelligence from analytics or speed would be an advantage. ·        Structured Data Analytics When processes critical to achieving an efficiency or goal are automated, telecoms need to develop structured analytics as well as centralize data processes so that the way data is collected is standardized and can be entered only once. Set of structured analytics provides product sales managers with a complete picture of historic customer call data; shows them which phone/broadband/iptv/VOIP ETC. were popular with which customers; what sold where; which phone/data packages customers switched between; and to which they remained loyal. ·        Experimenting with AI After these standard structured analytics are integrated with artificial intelligence, it’s possible to comprehensively predict, explain, and prescribe customer behavior. In telecommunications companies, managers can understand customer characteristics. But they need lots of practise & experiment artificial intelligence to analyze the wide set of data collected to predict if customers were at risk of leaving. After machine learning techniques identified the customers who presented a “churn risk,” managers then go back to their structured analytics to determine the best way to keep them — and use automated processes to get an appropriate retention offer out fast. Companies are just beginning to discover the many different ways that AI technologies can potentially reinvent businesses. Not only Telecoms, but also Social Media, Finance, Retailers etc. primarly because of data based on historic performance, but also with their unstructured data.