Social Listening Why, how and when does it work? AND ASKING PUBLIC WEB & Open-ended Answers LISTENING Call Centre Conversations Offline Data 2 AGENDA ARTIFICIAL INTELIGENCE FOR INSIGHTS Dashboard Demo CASE STUDIES Q&A 3 THE DEFINITION OF ARTIFICIAL INTELLIGENCE 4 ARTIFICIAL INTELLIGENCE FOR INSIGHTS ● Noise reduction ● Topic analysis ● Sentiment analysis ● Emotion analysis ● Image processing PROJECT 1: BEER Sentiment Accuracy/Noise->Irrelevant Posts->Homonyms For social insight one thing matters the most: ACCURACY ©2016 DigitalMR® All Rights Reserved Without it we get the proverbial: “Garbage IN Garbage OUT” 8 ©2016 DigitalMR® All Rights Reserved NOISE ELIMINATION Poker NOT Poker AND Machine AUTOMATED TOOL 16,360 posts NOT Poker AND Face listening247 10,505 relevant posts 9 SENTIMENT ANALYSIS AUTOMATED TOOL ANALYSIS 39% 57% 4% RE ANALYSED DATA 3% 9% 5% 15% 56% 88% 37% 80% 7% AUTOMATED TOOL 44% sentiment accuracy listening247 87% sentiment accuracy PROJECT 2: Telecommunications Sentiment & Semantic Accuracy NET SENTIMENT SCORE FROM UNSTRUCTURED DATA TO QUANT. TO QUAL. FROM UNSTRUCTURED DATA TO QUANT. TO QUAL. @Tigo_Colombia Can’t make any calls!!!! To any number, even though I have credits. Wtf? @Tigo_Colombia Your service is terrible. I have changed the company for it was a good offer, but the calls break the same. @Tigo_Colombia My first smartphone! Those were the Times! @Tigo_Colombia Thank you very much, but still having network 2g and 3g annoys a bit. thanks for your attention and good service 1 4 PROJECT 3: DENTURES Any Language SENTIMENT ACCURACY & SOURCES SENTIMENT ACCURACY: 85% 91% Italy Brazil Russia Japan 77% 82% 90% 76% Italy Brazil Russia Japan WHERE PEOPLE TALK ABOUT DENTURES: SOURCES: Blogs 3% 3% 2% 1% 1% 53% Boards 54% 13% 7% 0.3% 14% 24% Twitter 36% Videos 7% 82% 1% 90% 1% 97% 1% 85% 0.2% 23% 1% INTERNATIONAL DIFFERENCES JOKES INSULTS NEGATIVE PERCEPTION OF PEOPLE WITH DENTURES PROJECT 4: BEER EMOTION DETECTION EMOTIONS TOWARDS BEER 42% 51% Positive Joy AnticipationDesire Love Trust Distrust Anger Neutral Hate 7% Negative Sadness Surprise Bravery Calmness Disgust Fear No emotions Base= 8,559 posts ©2016 DigitalMR® All Rights Reserved 1 9 EMOTIONS DRIVE SHARING 2 0 PROJECTS 5/6: Telecomms & Health “The Holy Grail” 2 INTEGRATION OF LISTENING & SURVEYS 2 2 2 INTEGRATION OF LISTENING & RETAIL SALES DATA A Strong Correlation Between Listening Data And Retail Sales* SOCIAL LISTENING WHAT CONSUMERS BUY 2 R =0.81 *based on data for the MENAP region DEEP LEARNING FOR THEME CAPTIONING “1st Attempt: A group of people standing around each other in a restaurant”. “nth” Attempt: Men drinking beer in a pub”. @DigitalMR_CEO ©2017 DigitalMR® All Rights Reserved 2 4 THE LISTENING247 DIFFERENTIATORS Feature listening247 Other Vendors Sentiment Accuracy >75% <60% Noise Significantly reduced 80-90% of the posts are irrelevant Topic (semantic) Accuracy >80% No taxonomy, search based only, no accuracy metric available Languages ALL – 2 weeks set-up if not available Automated sentiment analysis for ususally only 1 language or a handful in the best case scenario Emotion Detection Proprietary model with 14 basic emotions Not available Integration Meaningful integration with existing survey trackers and * Not available Image Processing Logo and Theme detection Not available *online communities platform for co-creation 2 5 DASHBOARD DEMO Q& A Michalis A. Michael [email protected] @DigitalMR_CEO ©2016 DigitalMR® All Rights Reserved
© Copyright 2026 Paperzz