Diamonds in the Rough (Sentiment(al) Analysis

2,376 views

Published on

Power Point from a Webinar which focused on Sentiment Analysis

Published in: Technology, Business
0 Comments
1 Like
Statistics
Notes
  • Be the first to comment

No Downloads
Views
Total views
2,376
On SlideShare
0
From Embeds
0
Number of Embeds
127
Actions
Shares
0
Downloads
32
Comments
0
Likes
1
Embeds 0
No embeds

No notes for slide
  • GS
  • SW
  • SW
  • We will email you and post a copy of this presentation
  • “Semtiment Analysis” is complicated….
  • SW
  • SW
  • SW
  • Sw
  • GA
  • GA
  • GA
  • GA
  • GA
  • GA
  • GA
  • GA
  • SW
  • GA
  • GA
  • SW
  • SW
  • SW
  • SW
  • SW
  • SW
  • Diamonds in the Rough (Sentiment(al) Analysis

    1. 1. Gary Angel and Scott K. Wilder
    2. 2. 3 Gary Angel, President of Semphonic Co-Founder and President of Semphonic, the leading independent web analytics consultancy in the United States. Semphonic provides full- service web analytics consulting and advanced online measurement to digital media, financial services, health&pharma, B2B, technology, and the public sector. Gary blogs at http://semphonic.blogs.com/semangel Introductions Scott K. Wilder, Digital Strategist, WilderVoices Recently was SVP/Social Media Architect at Edelman – Digital. Founded and managed Intuit’s Small Business Online Community and Social Programs. Before Intuit, Scott was the VP of Marketing and Product Development at Kbtoys / eToys, the founder and director of Borders.com, and held senior positions at Apple, AOL, and American Express. Scott is also a founding Board member of the Word of Mouth Marketing Association. He received graduate degrees from New York University, The Johns Hopkins University and Georgetown University. Scott just jump started his own business, “WilderVoices” and blog @ www.wildervoices.com
    3. 3. 4 This will be a good webinar if…. you come away with a sense of how to do sentiment analysis, And how to make it work for your organization!!
    4. 4. 5
    5. 5. 6
    6. 6. 7 Sentiment Analysis – Complex Definition Sentiment analysis or opinion mining refers to a broad (definitionally challenged) area of natural language processing, computational linguistics and text mining. General speaking, it aims to determine the attitude of the speaker or writer with some respect to some topic. The attitude maybe their judgment or evaluation, their affective state – that is to say, the emotional state of the author when writing – or the intended emotional communication – that is to say, the emotional affect the author wants to have on the reader
    7. 7. 8 Sentiment Analysis – Simple Definition The simplest algorithms work by scanning keywords to categorize a statement as positive or negative, based on a simple binary analysis: -“love” is good, -“hate” is bad, -“I don’t know,” which means there probably will not be a second date
    8. 8. 9 Why Brands and Agencies Like Sentiment Analysis • Takes time to read and review every verbatim • Numbers look sexy • Can observe changes in issues / topics over time
    9. 9. 10
    10. 10. 11 How most agencies present Sentiment Analysis
    11. 11. 12 Or they just show you this..
    12. 12. 13 Sentiment Analysis and Market Research Market Research Anecdotal Focus Groups Comment Cards Behavioral Purchase Data Web Analytics Primary Research Online Survey Traditional Survey Sentiment Analysis
    13. 13. 14 Anecdotal Research Good For Not
    14. 14. 15 Behavioral Research Good For Not
    15. 15. 16 Primary Research Good For Not Thematic Tuning Color Behavioral Impact Audience Segmentation Campaign Development Message Tuning
    16. 16. 17 Usage of Sentiment Analysis • Sentiment Analysis is Anecdotal • Sentiment Analysis is Behavioral • Sentiment Analysis IS NOT Primary Research • NOTE: YOU CAN NOT AND SHOULD NOT ASSUME THAT FINDINGS FROM SENTIMENT ANALYSIS ARE REPRESENTATIVE OF YOUR ENTIRE AUDIENCE.
    17. 17. 18 Usage of Sentiment Analysis • Identify how people talk about your product • Understand how people talk about your category • Learn more about the competition • Respond real time to what people are saying Anecdotal • Learn more about the competition • Track changes in perception over timeBehavioral
    18. 18. 19 Sentiment Analysis for Anecdotal Research Manual • Themes and Colors emerge subjectively • Word Clouds are the closest data technique Non-Aggregated • It’s the actual text that matters • Sentiment Analysis can help cull non-neutral messages Classification • Subjective vs. Objective can help refine
    19. 19. 20 Sentiment Analysis for Behavioral Research Automated? • Human analysis of sentiment is more accurate • Consider analysis of API data using 3rd Party tools Sub-Classification • Sentiment classification is easier and more tunable in narrow bands of meaning • Sub-classify products etc.
    20. 20. 21 Listening Platforms • Buzz Metrics • Radian6 • ScoutLabs • Sysomos • Techrigy’s
    21. 21. 22 Text Mining Systems • SAS • Endeca • Autonomy • Lexalytics • OpenAmplify • You can get direct feeds to many social systems and mine text directly using solutions that are more powerful and tunable than those contained in most listening tools.
    22. 22. 23 The Right Tool Depends on the Data • If your analyzing tweets, you need a tool that understands acronyms and emoticons. • Sentence analysis is different that article analysis. Twitter is more like sentence analysis. • Topical classification, Objective v. Subjective, and Sentiment (polarity) are fundamentally different types of analysis.
    23. 23. 24 An example of the Problem: • Major software company looked at their major brand • 80% of the comments were neutral • 61% of the posts marked with a positive or negative sentiment came from microblogs. • They were data oriented, general information • But brands look at ‘Neutral’ and now what??
    24. 24. 25 Issues • Unreliable – not as good as pure human primary research • Not sure what to do with Neutral • No tool consistently identifies positive and neutral • Inconsistency across tools • Power of one or a small group of influencers • Lack of Accuracy
    25. 25. 26 Issues • Inconsistent across social networks (Twitter is usually 60% neutral) • Difficult to determine target • Overlook actual verbatim • No standards
    26. 26. 27 Now What? So…Now What?
    27. 27. 28
    28. 28. 29 10 Steps to find the Diamond 1 Decide on internal customer - which business unit, group will use data 2 Identify and use one tool (standardize on governance and set up) 3 Combine with other metrics/KPIs 4 Determine how broad you will go, but only look at one platform, community at a time 5 Identify emerging, trending topics 6 Take % of positive, neutral and negative 7 Take random sample of a 100 quotes from positive, neutral, negative, groups 8 Read verbatim 9 Categorize issues (For Intuit: Set up software vs. invoices vs. install vs. customer service) 10 Compare with other biggest competitors
    29. 29. 30
    30. 30. 31 Old School • More skin in the game • More engaged leads to greater satisfaction and places you closer to users/people • More confidence in results • Don’t drink and drive (I mean, don’t mix platforms)
    31. 31. 32 Note: It’s Never accurate Note: It’s never totally accurate
    32. 32. 33 • This presentation will be posted @ > http://www.semphonic.com >www.slideshare.net/skwilder >www.wildervoices.com Thank you for your time
    33. 33. 34 For more information Gary Angel: gangel@semphonic.com Blog: http://semphonic.blogs.com/semangel/ @garyangel Scott K. Wilder scott@wildervoices.com New blog: http://www.wildervoices.com @skwilder For other presentations: http://www.slideshare.net/skwilder

    ×