Attensity Company Blog

• Sentiment Analysis, Hard But Worth It!

  March 11, 2010 • Author: Michelle de Haaff • 0 comments • Leave your comment
  Tags: Text Analytics, Voice of the Customer

  Last week our CTO David Bean joined analyst and industry expert Esteban Kolsky (@ekolsky), the CTO from ScoutLabs, Jochen Frey and architect Franco Salvetti from Microsoft Bing (and formerly from Powerset) on a panel about sentiment analysis. One of my favorite topics to talk about with Dr. Bean - it was great to hear the whole panel discuss this topic both from an academic and practical point of view. (Here's a picture of the guys on the panel getting into the discussion!)
  
  So, in a nutshell - here are some of the learnings from that event:
  
  1) Sentiment analysis is hard!   Yes, sentiment analysis is hard for at least a few reasons.  First, in some cases, even humans have a hard time understanding the sentiment of what someone else is saying.  Commonly used statements when humans interact: "What do you mean?" "What are you trying to say?"  And that's when they ask - a lot of times people just don't understand the sentiment and don't ask for clarification.  Second, beyond the issues of ambiguity, for computers, being able to pull out the tone and meaning in a statement or set of statements is hard because people express things in different ways and finding the sentiment in a sentence is hard using certain statistical approaches. Many applications that try to understand sentiment use keywords or clusters of keywords to understand sentiment.  If "happy" is used in the sentence then it is positive sentiment.  Well - one can see why this can be brittle. First, users need to list every word that they can think of that is indicative of positive or negative sentiment and track for it - this takes time and effort to create and maintain the rules needed to do this. Second, what if "happy" is negated somewhere in the sentence? Proximity analysis helps with this "look for any negations in the sentence within two words from happy."  The problem is more rules to write and the sentences where the negation is more than a few words away are misclassified.
  
  To accommodate for these issues Attensity invented (and has patented) a completely unique method for getting to accurate sentiment - we call it Exhaustive Extraction. This technology enables us to parse sentences and accurately look for instances of sentiment expression and get it right more so than any other vendor that attempts to do this.  The image below depicts how we parse sentences to find both sentiment and meaning. The process is analogous to what we learned in grammar school English classes. In this example we parse the sentence: I really love my iPhone, but the reception here is very bad.
  
  

  
  
  
  Next we extract and aggregate the facts from the content. This allows us to not only find the words that are indicative of sentiment, but to find the relationships between words so that we can accurately identify both words that modify the sentiment (even if they are not close to the sentiment) and what the sentiment is about. Below is an example of the facts that we automatically extracted from this example:
  
  
  
  

  
  As you can see from this example not only do we pull out the facts - but we can also find the degree of sentiment not just "love" but "really love" and not just "bad" reception but "very bad."
  
  
  Another example that highlights our ability to both find sentiment and get it right is illustrated below. In this example the negation on the sentiment is not near the sentiment - but we are still able to get it right! The content in this one is: I have an iPhone, but I am not really feeling very happy about the iPhone.
  
  
  
  
  
  
  2) Sentiment analysis is needed! While it's a hard subject and in some cases impossible for both people and technology to tackle - the one thing that everyone agreed on is that understanding it is needed! It helps companies understand what buying customers think of their products, services, buying experience, customer service, even the competition.  It is a leading indicator of purchase intent and churn. It can help organizations identify cries for help and emerging issues. Wow!  Worth the millions of dollars of research that go into making this possible - hey?  We've recently blogged about a few examples of analysis that we've done - valuable insights for companies and based on understanding sentiment. You can find them here.
  
  3) Sentiment analysis is most useful when connected to what the sentiment is about! The ever elusive "why." When listening to customers or conducting market research the thing we want to know most is "why?"  I have already covered this somewhat above - but I wanted to emphasize this again.  The key thing about sentiment analysis is being able to go beyond the sentiment analysis to find out what people were happy or sad about.  So why did they like the iPad and say they were going to buy it? (below is an excerpt from analysis we did on the iPad launch.)
  
  
  
  
  
  Why didn't they like about the Nexus?
  
  
  
  
  
  You get the picture.......We can show you how and why understanding your customer's sentiment is valuable to you.  Let me know if you want to see how!  Signing off for now.....
  

• Listen, Analyze, Relate, Act

  February 25, 2010 • Author: Michelle de Haaff • 0 comments • Leave your comment
  Tags: E-Service, Information Access, Semantic, Social Media, Text Analytics, User Generated Content, Voice of the Customer

  I am sitting in a conference on intelligent content and have seen all sorts of great presentations about how static information, when "annotated" or "tagged" with metadata can come alive and become valuable information to organizations. It reminded me that I've been wanting to write a blog about our methodology, backed by our technology called LARA - LISTEN, ANALYZE, RELATE, ACT.  LARA provides the process and applications by which organizations can harness the massive value hidden in user generated content as a business asset.
  
  So, what is LARA?
  
  LARA is a process by which organizations can take user generated content (UGC) whether generated by consumers talking in web forums, micro-blogging sites, facebook, feedback surveys, emails, documents, research, etc., etc. and use it as a business asset in a business process.  Let me break it down for you...
  
  
  

  • "Listen" - to "listen" is actually a process in itself that encompasses both the capability to listen to the open web (forums, blogs, tweets, you name it) and the capability to seamlessly access enterprise information (CRM notes, documents, emails, etc.) It takes both a listening post (Attensity Cloud), deep federated search capabilities, scraping and enterprise class data integration found in Attensity Integration and wrapped up in a complete, easy to set-up solution for listening!
  • "Analyze" - now here is where the "secret sauce" comes in! How can you take all of this mass of unstructured data and make sense of it?  Using our patented Semantic Server that includes keyword, statistical and natural language approaches we are able to essentially tag or "barcode" every word and the relationships between words making it data that can be accessed, searched, routed, counted, analyzed, charted, reported on and even reused!  We do this using a process that doesn't require the user to define keywords or terms that they want their system to look for or include in a rule base, but rather it automatically identifies these terms ("facts," people, places, things, etc.) and their relationships other terms or combinations of terms - making it easy to use, maintain and also more accurate than what others in the text analytics space have to offer.
    
  • "Relate" - ok, so now that you have found the insights and can analyze the unstructured data -the real value comes in when you can connect it to your "structured" data. Your customers (which customer segment is complaining about your product most?), your products (which product is having the issue?), your parts (is there a problem with a specific part manufactured by a specific partner?), your locations (is the customer that is tweeting about wanting a sandwich near your nearest lunch store?) and so on. Now you can ask questions of your data and get deep, actionable insights. 
    
  • "Act"  - now here is where it gets exciting! What do I do with the new customer insight I've obtained? How do I leverage the problem resolution content a customer created that I just identified? How do I connect with a customer who is uncovering issues that are important to us or asking for help? How do I route the insights to the right people? And how do I engage with customers, partners and influencers once I understand what they are saying?  This is our act step. We've invested in a suite of applications that offer great value to end customers and companies when unstructured data is accessed. These applications include Voice of the Customer analysis, Intelligence Analysis, Research and Discovery, Risk Identification, and E-Service/Self Service. 
    
  
  If you want to learn more about our lovely LARA - let us know - we will send you a white paper on the subject....now I've got to get back to listening to this conference!!
  
  

• Leveraging Communities through Analytic Engines

  February 22, 2010 • Author: Esteban Kolsky • 0 comments • Leave your comment
  Tags: Customer Service, E-Service, Social Media, Text Analytics

  The driving force for the Social Customer era is the participation in communities both for social and professional purposes.  From the structured social networks (e.g. Facebook and Twitter) to company-owned or company-sponsored communities used for support, sales prospecting, or research and development, through communities used internally for collaboration between workers – communities are showing up just about anywhere.
  
  This change brings vast amounts of content generated by the communities.  In spite of the extensive experience gained by organizations in the past few years dealing with large data sets and knowledge, the user-generated content still remains untamed.  What to do with it, and how to leverage it for value, are almost as mysterious today as they were when we first began accumulating Knowledge in the 1980s.  Organizations are struggling to understand how to utilize it and how to derive value from it.  Alas, Content Management Systems and similar enterprise tools can help manage the creation and processing of structured content – but the largest problem still remains the unstructured content produced in these communities.
  
  Realizing Value from the New Large Volume of Content
  
  Consider the size of some of these communities: Facebook is close to 500 million people, Twitter nears 100 million, and a few of the corporate-sponsored communities have over two million members.  The amount of content generated is bringing organizations that were already drowning in data from transactional CRM systems to desperate levels.  They are now saddled by massive volumes of knowledge and feedback that makes finding the needle in the haystack look like child’s play.  In spite of the amazing volume, the storage and management of the content is not the problem – storage space is cheap these days so virtually any amount of content and data can be stored for – well, as close to forever as we need to.  The solution of cheap storage has given place to a bigger problem: what to do with it?
  
  An organization wants to capture and leverage critical information from their customers’ needs and wants to deliver better experiences and products.  On the other hand customers fear that their feedback is not being heard and used. To show customers they care about their opinion, companies must act on the feedback.  Alas, given the volume, and short of scanning each entry posted in any community for useful information or data, how can they capture and act on this feedback?
  
  
  
  Enter analytical engines.
  There are two roles that an analytical engine can play in a community – they can either be used to monitor and report on usage, sentiment and trends, or they can be used to structure the unstructured. 
  
  Monitoring for the Sake of Monitoring
  
  Social Media brought with it standard monitoring tools.  Whether from Social Media Monitoring (SMM) vendors like INgage, Radian6, ScoutLabs, and Visible Technologies, or embedded within the products of other vendors, these tools are quickly becoming the “first line of defense” for the barrage of data produced.  The ability to collect the raw data, summarize it and report on specific terms is valuable for organizations that are suddenly overwhelmed by these new channels. 
  
  These tools are used for monitoring specific words and phrases, brand mentions (or competitors’ brands), and people talking about industries or products.  For example, during the TV airing of Super Bowl XLIV there was an analysis of brand mentions done by Radian6 and partners, called  BrandBowl 2010, which resulted in the naming of a winner by number of mentions and “positive” (like or dislike expressions) sentiment.  During the same event, another analysis done by MarketIQ contrasting Coke and Pepsi, aptly named the SodaBowl, also looked at mentions and sentiments for both drink manufacturers.  Again, the conclusion was to which was more popular – they actually used the term “buzzworthy” – not who gained what from their different approach to promoting themselves.
  
  While certainly entertaining, it yielded no value to the brands mentioned on the success of failure of their campaigns – just whether they were popular or not. 
  
  Although there is room for improvement in sentiment analysis, the near-real-time analysis of these events allows marketers to identify which communities are important to them, and which ones need further attention.  It also allows them, for the first time, to understand immediately what effects their actions have and adjust campaigns and plans in real time –invaluable to improve the message and ensure a good reception by the public.
  
  However, monitoring for the sake of monitoring yields limited value to businesses on their way to becoming social.  Listening is the first step, but engaging with the customer and providing a return on their feedback is closer to becoming a social entity.  Organizations leveraging analytical engines to find and structure this feedback are on a more interesting path to assess.
  
  Structuring the Unstructured
  
  Among the contributions to communities by their members there are very interesting nuggets of information, opinions, and suggestions that are often lost since there are no tools that can extract it, organize it, and use it.  This information could be used to improve products, create better experiences, or to better understand the needs of the customers and prospects.  Customers are more open in their opinions among peers than when being asked to complete surveys or participate in focus groups.  This candor and openness often results in very valuable data – which is not always leveraged. 
  
  
  
  Analytical engines can find that information and structure it (create a data record from it), distribute it to the specific system that can utilize it, and keep track of trends and patterns on the data they find.  Organizations use them to carry out actions like ideation (the creation of new products and services), feedback management (understanding how customers really feel beyond the surveys), social prospecting (finding more about their prospects and segments to target in sales), and virtual focus groups (leveraging customers’ opinions without formally convening a group).
  
  Good analytical engines will automatically classify all the information collected (using an SMM – social media monitoring tool – is the best way to collect all this information) into different buckets, and analyze those buckets to generate insights.  This categorized information in its raw form is somewhat valuable, but the use of workflows and databases to store this data and process it further yield very powerful knowledge for the use cases mentioned above.
  
  Integration Rules the Analytics World
  
  The most valuable output an analytical engine can produce is the ability to take different inputs, across channels and across functions, and use all that in search of insights.  Organizations receive communications via email, chat transactions, online comments, surveys with free-text boxes, and many other methods.  To focus the efforts only on the communities, because they are the “hot item”, leaves a lot of potentially valuable data un-examined.  This data must be merged and integrated with the community insights for further analysis.  Analytical engines cannot stop at simply producing a report for each community; they have to become a critical part of the platform used by the organizations to interact with and manage their customers.
  
  This platform will then integrate the content generated by all channels and all methods the organization uses to communicate, and produce great insights that can be analyzed for different channels and segments, or altogether.  This analysis, and the subsequent insights, yield far more powerful customer profiles and help the organization identify needs and wants faster and better.
  Alas, the role of analytical engines for communities is not to analyze the community as a stand-alone channel, although there is some value on that as a starting point, but to integrate the valuable data from the communities into the rest of the data the organization collects and produce insights from this superset of feedback.
  
  What do you think?
  This is the first in a series of posts I will be writing with Attensity to look at the value and purpose of deeper analytics on communities (i.e. beyond simply mentions and sentiments-like words and phrases) and social channels.  Any ideas or areas I should explore further?
  
  

• Have You Heard the Buzz?

  February 16, 2010 • Author: Catherine van Zuylen • 0 comments • Leave your comment
  Tags: Social Media, Voice of the Customer

  Less than a week ago Google launched Google Buzz.  Just two days after the launch, Google reported on their blog that there were over nine million buzz updates posted from mobile devices alone – that’s a rate of 200 mobile posts per minute.
  
  Did you know that Attensity now listens to Google Buzz in addition to the over 110 million blogs, 20,000 online mainstream news sites, Twitter, rich media sites (like YouTube), review sites, and other social media sites we already were monitoring? We now cover approximately 4.5 million Google profiles and our system is automatically and rapidly building on this number, prioritizing the most active profiles. We will be looking to extend this coverage to include Buzz comments as well in the coming weeks.
  
  Attensity also allows you to analyze "internal" customer content from your surveys, CRM systems, private web communities from providers like Lithium, and other information to gain a 360-degree view of your customers.
  
  And unlike most social media monitoring tools, which use fairly unsophisticated keyword analysis for reporting, Attensity automatically analyzes and extracts cries for help, compliments, complaints, at-risk behavior, intent-to-purchase behavior, conditional behavior (“if it did this, I would do that”), and so forth, along with deep clausal sentiment analysis that provides C-level reporting and explains the "whys" behind the buzz.
  
  To learn more, contact us.
  

• When Acting on Customer Insights, Accuracy Matters

  February 09, 2010 • Author: Jeff Johnson • 1 comment • Leave your comment
  Tags: Attensity, Text Analytics, Voice of the Customer

  I have been spending a lot of time with some of our customers lately, learning how they are listening to and engaging their customers. One thing that really sticks out in every discussion I have is how important it is that the information that they get is accurate. Why?  Because in all of our implementations, our customers are using their analysis to take action to improve products or to prevent customer churn or to drive a marketing program…..to do something that impacts the customer and the business.
  
  Buyer Beware
  
  It’s amazing to me how many companies miss this point during the buying process, only to find out later that they purchased a system that might create nice charts and graphs, but that doesn’t have the technology behind it to easily and accurately get to actionable information.  In early markets, with limited coverage and no "Magic Quadrant" or "Wave", its buyer beware.  The core technology used to understand customer conversations is text analytics.   But don’t feel too comfortable when your Customer Experience Management vendor tells you they have a sophisticated text analytics engine because “truth be told” they come in all shapes and sizes… from statistical word counting and targeted keyword lists, to sophisticated natural language processing or NLP. When evaluating text analytics solutions it’s very important to not only evaluate the output – dashboards, reports and exploration capabilities, but to also inspect the quality of the data being extracted.  Solutions from the vendors that participate in our space, big and small offer very different technologies. 
  
  Chose the One with Natural Language Technology That is Actually Used in the Solution
  
  Vendors that invest in NLP technology to model the structure of words in order to understand a sentence can yield accuracy similar to that of a human.  Statistical approaches (keywords, classifiers, etc.) are considerably less accurate because they typically require the user or implementer to create rules to model what data is pulled from the system.  In a lot of cases our customers don’t know what they are looking for and because they have to know, to create these rules, they inherently miss critical information.  A couple of years ago it might have been okay to capture general themes and trends but today companies see the value in not only listening to get a general understanding of the customer, but listening to determine what to do to improve customer satisfaction, competitive position and sales.   Correctly understanding all issues about every customer must be the goal for Customer Experience Managers because the impact in today’s socially networked world is too great.  We see the affect of not hearing the individual daily, from an airline who can’t hear a guitar man http://www.youtube.com/watch?v=5YGc4zOqozo, to an auto manufacturer who can’t figure out if it’s a floor mat or electrical interference.  If you use NLP technology and that technology does not require you to write rules (and therefore know everything you are looking for) then you have the best chance of finding new issues, uncovering opportunities you never saw before.
  
  When evaluating vendors do a simple test and ask all of us how our applications stand up to the accuracy question and enable users to find issues they weren’t aware of before.  Do they parse sentences to find entities (people, companies, brands, etc?) Do they relate those entities to sentiments, events and other issues, intent (to purchase, to leave…), conditions (actions a customer would take if some condition existed)? If so, they have sophisticated NLP.  Another good test….ask if they need to create rules or word lists before they run your data through their system?  If so, you need to be a mind reader or you’ll  miss information and worst of all if you expose the text behind that pretty bar chart you might find that 20 to 30% of your facts are wrong… 
  
  Nice to Meet You!
  
  This is my first official blog.  Since I manage both the sales and service delivery parts of our business, I am going to focus on discussing customer examples of receiving value by getting both accurate and actionable information out of customer conversations in blogs, forums, tweets, surveys, CRM notes, warranty or even claim information.  I hope you learn something from these posts. If you agree, or don’t agree – feel free to comment!