Executive Summary

A common challenge facing today’s enterprise, its employees, and customers, is the ability to easily and effectively access corporate data and product/services-related information.  To obtain the accurate and specific information needed from a vast corporate network is a daunting task, especially as data grows more complex and the workforce becomes increasingly mobile.  What if we had pervasive access to it all– enabling us to make timely business decisions while continuing on with our busy days?

Today, it is estimated that 80 percent of corporate information is stored and maintained in unstructured form.  IDC estimates that information workers spend 48 percent of their time searching for and analyzing information; costing organizations $28,000 per worker each year.  Enterprises are experiencing increased difficulty in extending this information – from databases, product manuals, knowledge-base articles, FAQs and support forums to both their employees and customers.  With call center costs averaging $10 per call in some industries, a method of delivering the right answer at the right time is invaluable and offers tremendous return on investment (ROI) across many applications.
 
To reduce call center costs, improve customer relations and better locate corporate information, enterprises need a natural-language, semantic intelligence solution.

The Benefits of Semantic Intelligence for the Enterprise

Semantic intelligence is used to improve the search process by leveraging natural language to disambiguate queries, thereby increasing the relevancy of search results.  Disambiguation is key and as a general rule, the larger and richer the ontology and taxonomy present, the higher the precision and opportunity for true natural language processing. 

Rather than using ranking algorithms such as Google’s PageRank to predict relevancy via presence of keywords, semantic intelligence uses “semantics,” or the science of meaning in language, to produce highly relevant search results.  In most cases, the goal is to deliver the information queried by a user rather than have a user sort through a list of loosely related keyword results.

The word “semantics” related to search today is a term that’s been over-used and under-delivered.  Many search companies claim to have semantic capabilities, but enterprises have been slow to adopt such tools because they’ve repeatedly fallen short.  The human language—especially in the corporate world—is far too complex to rely on keywords to produce results.  The same words have many meanings and a single person should not have to sort through hundreds of pages to get the information that’s truly needed.   

The key component to semantics is the ability to “disambiguate.”  When a term is ambiguous, meaning it can have several definitions (for example, if one considers the word “jaguar," it can be understood as "an animal” or "a car"), it is critical for the system to choose the correct meaning in order to return the most relevant results.

The highest precision can only be accomplished by analyzing the accompanying language in the entire sentence or paragraph, as necessary.  To do this, a deep, broad and rich representation of the language must be available.  The determination of every meaning subsequently influences the disambiguation.  All the fundamental information for the disambiguation process — that is all the knowledge used by the system — is represented in the form of a semantic network, organized on a conceptual basis.

This higher-level understanding of the relationship between words enables users to search by meaning through natural language questions via “faceted search,” or the tagging and categorizing of each document analyzed.  For example, a search for stock intended as a share in a company rather than stock within a soup broth.

True semantic intelligence systems automatically extract rich sets of entities (named and not*) and relationships between those entities, starting from logical ones (subjects and objects) to semantic ones (thanks to the semantic disambiguation), and finally roles and attributes, family relations and much more.  They understand sentence structure as a human would, comprehending the tone, context, slang and even major brand names.

  • Entities extracted: Anniversary, Address, Animal, City, Company, Continent, Country, Currency, Date, Device, Email Address, Entertainment Award, Event, Facility, Fax Number, Food, Holiday, Market Index, Medical Condition, Medical Treatment, Month, Measure, Natural Disaster, Natural Feature, Operating System, Organization, Percent, Person, Phone Number, Plants, Programming Language, State, Sports Event, Sports League, SSN, Time, Tool, URL, Vehicle, Year

Semantic intelligence systems can also be used for customer relations, helping to filter inbound email or text message queries to the most appropriate contact professional, or to provide automated, accurate responses into the user’s inbox or mobile device.  Not only can the disambiguation capabilities help identify the true nature of the request, it can also locate corporate data such as how-to manuals to provide the best results/responses to the customer.

Semantic Search for Mobile Self-help

With the downturn in the economy, many companies including wireless telecommunication carriers and smartphone manufacturers, such as AT&T, Verizon, Sprint, T-Mobile, iPhone, Nokia, Motorola and others, are looking for a way to deliver the highest level of customer service while reducing costs to deliver that service.

Mobile devices are growing more sophisticated with each newly released model, and more users are adopting smartphones such as the iPhone or Google’s G1.  The problem is, the more applications and capabilities these devices hold, the more questions users have about how to use them.  By one estimate, global operators are expected to employ more than 6 million agents — spending more than $100 billion in customer service care — this year alone.

Semantic intelligence can understand the common language query of each individual user on his or her mobile device through a Web-based self-help application, then search the manufacturer‘s premier knowledge base and return the correct answer instantaneously.  The resulting benefits are:

  1. Fewer calls and less cost to the wireless carriers’ call centers on how to use the devices and services.  The cost of a human based call is estimated to be more than $10 on average.  Answering the same call semantically costs pennies. 
  2. Reduced rate of product returns (e.g. less churn) based on the user’s inability to quickly get comfortable with the device and learn how to perform various functions.  Many carriers incur several hundred dollars in acquisition costs and thus look for many months of satisfied use and service to earn back those costs.  Semantics significantly reduces the probably of churn.
  3. Faster time to answers and thus higher product satisfaction.

Final Thoughts

Many enterprises claim some expertise or ability to create their own domain-specific ontology.  While this may be adequate for some uses, the absence of a complete language ontology reduces the holistic benefits of what true semantic search was intended for and can provide.  Enterprises may run the risk of unattainable expectations; quickly finding their initiatives coming up short and missing out on the full potential semantic intelligence technology can truly bring.

Last year, IDC wrote about the expanding gap between exploding volumes of digital information and the lagging tools for finding that information and making sense of it.  This is an expensive problem and getting worse.  As mentioned , analysts predict information workers will spend almost half their time searching for and analyzing information resulting in tremendous costs to their organizations.

Searching for a needle in the digital haystack is growing larger with each passing day and becoming a prevalent issue for both enterprises and consumers alike.