Ferret Guide to Text Analytics


Published on

Feedback Ferret's guide to their text analytics engine. Find out why and how the Ferret's text analytics deliver 93%+ accuracy.

Published in: Business
1 Like
  • Be the first to comment

No Downloads
Total views
On SlideShare
From Embeds
Number of Embeds
Embeds 0
No embeds

No notes for slide

Ferret Guide to Text Analytics

  1. 1. Text Analytics FERRET GUIDE
  2. 2. ___________________________________________________________________________________________________ 2 The Ferret chews through customer comments at lightning speed and delivers real-time results to your organisation. It filters vital feedback from chatter; keeping your finger on the pulse and focusing attention where it matters. Feedback Ferret’s text analysis engine delivers world leading levels of accuracy and we undertake all the coding and programming ourselves. By doing what we do best (text analytics), our clients are able to focus their time and efforts on what they do best (making happy customers). Our engine does not use any type of automatic machine learning. Rather, we have a team of people who have built and constantly improve our Lexicon because we believe this is the most accurate way of interpreting all the weird and wonderful things people say when talking about experiences. The Ferret does all the heavy lifting – you simply receive the results. ContentsDelivering world leading levels of accuracy
  3. 3. ___________________________________________________________________________________________________ 3 How it works Step 1: Topic Extraction: All text comments are processed against the Feedback Ferret Lexicon. The Lexicon is compiled from ‘Case Based’ human interpretation of feedback content to ensure accurate contextual meaning. Every sentence is analysed for every contextual phrase, so you can understand everything customers are telling you. There is no limit to the length of the comments that we analyse. All this ensures:  Accuracy: ensuring that all relevant comments are extracted from the data  Categorization: clustering text into usable Categories  Poor quality raw data: achieving high accuracy rates despite poor grammatical quality of the source data  Sentiment scoring: all text is contextually analyzed and scored for sentiment Step 2: Reporting Topics Reporting Topics are created from the underlying Lexicon results. These can be simple – just one tightly focused Category (e.g. ‘Angry customer’ or ‘Angry staff’’), or they may be complex combinations to accurately define detailed subject matter. We provide several ‘industry standard’ topic sets, plus we tailor topic sets to meet individual client requirements. All Reporting Topics undergo continuous quality checking and improvements so that you can always rely on our high levels of accuracy. All this ensures:  Accuracy: confidence that the topics show correct results  Granularity: the right levels of topic granularity for your business needs  Usability: tailored reporting topics for your business, not simply “one size fits all”
  4. 4. ___________________________________________________________________________________________________ 4 Example topic groups and names Below is an example of actual customer commentary received and coded by Feedback Ferret:
  5. 5. ___________________________________________________________________________________________________ 5 Contextual analysis of feedback comments Spelling & Grammar  No correction is needed for poor spelling, grammar or punctuation in the source data.  Feedback Ferret effectively handles misspellings, poor grammar and bad punctuation to ensure accurate topic coding. Text-speak & Emoticons:    Feedback from SMS text messages and social media channels can be full of emoticons and “text-speak”.  The Feedback Ferret Lexicons have extensive contextual content to categorize this type of input effectively. False Positives & Negatives  Often a problem in automated text analytics, Feedback Ferret ensures very high levels of accuracy for categorizing false positive and negative statements.  When it comes to using the feedback and acting on the results in your business, it is imperative that you should be confident of accurate identification of these positive and negative statements. Sentiment Scoring  All text is scored for sentiment, i.e. how positively or negatively the customer has expressed their views.  The sentiment scoring takes into account the contextual meaning of comments, not simply a count of positive and negative words.  For practical purposes, sentiment scores are grouped into 7 bands, from Extremely Positive to Extremely Negative.  The detailed numeric sentiment scores are also retained for use in analytics and reporting. Only problem at the moment is telephone cues on the inquiries line Telephone Answering – Poor i'm now a positive advocate  Recommendation I COULD NOT HONESTLY THINK OF A BETTER CAR Vehicle – Good
  6. 6. ___________________________________________________________________________________________________ 6 Brand names, Jargon, Colloquialisms  The Feedback Ferret Lexicons are enhanced with your brand names, distributor names, industry jargon, key personnel, etc. and this is continuously updated from the ‘Case Based’ analysis of feedback data.  A wide and ever-growing range of colloquialisms, euphemisms, slang, sarcastic and ironic statements are incorporated into the Lexicon for accurate coding of text comments. Tailored Reporting Topics  We create custom reporting topics for every client based on your business needs and on what your customers are saying.  Standard industry topic sets are available for immediate progress.  The level of granularity for reporting topics is easily tailored to suit your business needs.  The reporting topics can be amended and updated over time. More genuine examples of customer feedback comments show how the Ferret’s automated text analysis engine categorizes the source data into contextually accurate Reporting Topics: I wanted to get a cardigan in size 12 but they only had it in orange Clothing – Jumpers & Cardigans Clothing sizes Availability – Poor Colour - Orange getting new blood in the organization is healthy Staff – Recruitment He drove me potty Dissatisfied I’m still waiting for my first bill Invoice Process – Poor They're all a bunch of cowboys Dishonest At no time did I feel I had been bait and switched Sales Process – Good Honesty - Good The Apple customer service was excellent Apple Technology Customer Service - Good don't believe someone at this point had apologised No Apology I am still waiting for Bill to call me. Telephone Communication – no follow-up Shes clearly not the apple of her manager's eye Staff Quality - Poor I’m on Orange but the connection was poor Orange Mobile Network Mobile Signal - Poor Impact Analysis Your orange juice is the best I’ve ever tasted Orange Juice Good tasting I wanted a Little Angels potty but they had run out Baby Equipment Availability - Poor
  7. 7. ___________________________________________________________________________________________________ 7 www.feedbackferret.com UK / Head Office Piers Alington Managing Director Feedback Ferret Ltd The Old Barrel Store Draymans Lane Marlow SL7 2FF +44 (0) 1628 681 088 info@feedbackferret.com USA Kate Handley Vice President, Client Services Feedback Ferret, Inc. 150 North Michigan Avenue Suite 2800 Chicago, IL 60601 +1 (312) 291 4629 infoUSA@feedbackferret.com South Africa Jason Wilford CEO Feedback Ferret SA (Pty) Ltd 3 Waterford Office Park Waterford Drive, Fourways, 2055 Gauteng +27 (0) 11 2511980 infoSA@feedbackferret.com Poland Grzegorz Popek Feedback Ferret c/o SmartLife Polska, a division of Retail Partner Polska Group ul. 25 Stycznia 18, 62-080 Tarnowo Podgórne +48 (509) 445 564 infoPL@feedbackferret.com Feedback Ferret® and the Feedback Ferret logo are Registered Trademarks of Feedback Ferret Ltd. Contact