Skip to Content

SAP BusinessObjects Text Data Processing Analyzing Social Media Mentions: an ASUG Webcast

  Earlier in the month, Anthony Waite, SAP, provided an ASUG webcast on SAP BusinessObjects Text Data Processing : Analyzing Social Media Mentions.  The usual disclaimer applies as things are subject to change. Agenda: ** Text Data Processing ** Sentiment Analysis ** Key Learning Points h4. +Market Forces+ The world is “connected”.   Through social media, we are connected and connected to world events In the past, to plot a coup, a television station would be seized.  That has changed; everyone has a TV “in their pocket” with phones, giving insight to uprisings.  Anyone can tune into “information highway”. image  Figure 1, Source: SAP   80% of data is unstructured data, according to Gartner.  The majority of data is unstructured such as e-mails, CRM system and social media.  For businesses, structured data supports automated data processing; it conforms to a data model in databases and spreadsheets.  It is granular.  image  Figure 2, Source: SAP   Figure 2 shows an example.  Structured data has a specific data domain.  As an example it is related to a customer or materials. Unstructured data model is not easily understood and there is insufficient metadata.  There is lots of “noise”.  An example is a press release communication.  Content is free-form with “key mentions”.  Entities and mentions are interchangeable.  Another example is a forum posting.  Could it contain a praise or complaint? According to Anthony, structured data tells us “what”.   Unstructured tells us “why”.  For example, J. Crew dresses are out of stock tells you what, but doesn’t tell you why.  Then you analyze social mentions you see Michelle Obama wore that dress the day before and that will tell you “why”.  This gives you the context. image  Figure 3, Source: SAP   Figure 3 gives you a business perspective; you are missing a business opportunity if you are not combining unstructured with structured data. Prabhakar Raghavan, from Yahoo Research, “The bulk of information value is perceived as coming from data in relational tables. The reason is that data that is structured is easy to mine and analyze Business Intelligence tools run off unstructured data.” How can you extend your BI investments to unstructured text data?   h4. +Text Data Processing+ image  h4. + + +Can we +*+bill +*+you?+ Recognizing bill is a verb *+Bill +*+was the president.+ Recognizing Bill is a noun   What follows next is disambiguation to resolve conflicts. Semantic Disambiguation:  +I talked to +*+Bill +*+yesterday.+ +Proper noun+  +The duck has a +*+bill+*+.+ +Common noun+  +The +*+bill +*+was signed into law+ Text Data Processing learns what a core person is. image  Figure 8: Source: SAP   Figure 8 shows the languages supported in Data Services 4.0 In 2012 additional languages will be added “out of the box”.  You get this with Data Services 4.0 and part of the default installation.   Domain extraction is realized through rules which look for facts which are relationships between entities (mentions) or states involving an entity.   image  Figure 9, Source: SAP   Figure 9 shows the language modules that come with Data Services 4.0, “voice of the customer”, sentiment analysis, was it a strong positive or negative They are also able to extract the problem from the text You can change it as well as business may have different view of terminology – for example, the word “thin” is not positive in hospitality but may be positive in retail.   You can customize it based on domain.   h4. +Sentiment Analysis+ As defined per Wikipedia +“+*+Sentiment analysis +*+or ++opinion mining ++refers to the application of +natural language processing+, +computational linguistics+, and +text analytics +to identify and extract subjective information in source materials.”+   For the voice of the customer “Apply text data processing to enhance customer service and satisfaction by understanding customer opinions on blogs, forum postings, and social media.”   It is trying to understand how people feel about the brand, do they have an affinity, in social media or documents.   According to Leslie Owens, +“The challenge lies in identifying statistically valid data related to specific business priorities from the mountain of available content. You don’t want to overthrow a key marketing campaign because a few bloggers write snide things. ”+

+ + When customers are unhappy, they strike back; this is called “double deviation”.     From MIT Sloan Management Review “customers have been victims of not only a product or service failure, but also failed resolutions.   Betrayal is the “primary driver of what causes customers to complain online”.   It is easy to use as a “bully pulpit”.   According to Forrester Research, “+78% of consumers trust peer recommendations+.”   h4. Key Learning Points:   image

1 Comment
You must be Logged on to comment or reply to a post.