FOCUS on…

Focus on Contact Center of the Future

In an intensely competitive global market, companies seeking to differentiate themselves know that offering top-notch customer service is one way to stand out from the crowd. The current focus on customer relationship management (CRM) as a key differentiation strategy is fueling a boom in CRM analytics software, with contact center operations representing a significant portion of the service delivery, generally between 10 and 15 percent. As a major player in speech analytics and other contact center technologies, IBM is strategically positioned to take a leadership role in this market.

Recognizing that contact centers serve as the primary contact point for many customers, smart companies are investing in their contact center operations to strengthen customer relationships, while at the same time using automation technologies to streamline operations. New advances in automated contact center technology can help reduce the workload on human agents and provide a more satisfying experience for clients.
IBM Research Services, a collaboration between IBM Research and IBM Global Business Services (GBS), has taken the lead in this area through its extensive portfolio of speech-enabled solutions that can help improve performance by automating a wide range of functions, from contact center systems to audio content analysis and transcription. By leveraging sophisticated conversational technologies and routing maps to deliver these flexible natural language solutions, companies can reduce costs while maintaining the quality benefits of human interaction.

Speeding the development of conversational self-service solutions

The Reusable Speech Assets and Analytics Reporting Tools initiative, a joint development partnership between Watson Research Labs and GBS, is focused on accelerating development and deployment of speech-enabled tools and applications that can provide businesses with contact centers a shortcut to creating customized speech-enabled applications. Among the tools are a natural language understanding call routing system that captures callers' intent and routes them to the appropriate destination. A directed dialog feature helps resolve any ambiguities when intent is unclear. Other tools allow callers to reset their passwords and provide personal information through a speech interface. One such project involves a conversational help desk application that uses a statistical language model-natural language understanding interface to transform an existing touch-tone-driven interactive voice response system to speech user input.

IBM's CIMA (conversational interaction management architecture) provides a flexible, modular spoken language multimodal application framework for processing incoming calls in a variety of environments. The most common of these are Web servers for VoiceXML based telephony applications and platforms used to deploy embedded applications in the car or portable devices. When customized for contact center routing, CIMA's interaction manager coordinates the operation of its subcomponents – event manager, session manager, strategy manager and action manager – to determine the routing and handling of each phone call. In addition, a "policy" component may apply predetermined rules to the strategy selection process in order to narrow down the choices. Actions are pieces of executable code, with a common interface, that perform the various tasks of an interactive application. When combined with the proper tools, CIMA supports a rapid application development methodology, extending its use to research on human computer interaction and human language technology.

Analyzing customer contact call data to help improve business processes

While technology can play an important role in providing better service and a more satisfying customer experience, it also is a vital link to customer feedback that otherwise might be lost. Contact center analytics tools tap into the data trove generated during customer contact and reveal important information about user behavior and application performance that can then be used to fine tune operations. The unified call log processor tool extracts raw call logs from contact center servers, applies filters and consolidates that data into an XML format, which is then saved in a DB2® database. The unified call log analyzer tool analyzes the database records and provides users with consolidated call data in both its raw and analyzed forms. These tools, used in tandem, provide speech application developers, speech scientists, business analysts and system engineers with technology to support speech application grammar performance tuning, as well as generate business-level and interactive voice response (IVR) channel utilization reports.

IBM Research has also developed an integrated quality monitoring application that transcribes each call and automatically measures the quality level and degree of adherence to stated policies. The application uses IBM's state-of-the-art speech recognition toolkit combined with IBM DB2® information management technology and IBM WebSphere® applications, and is built on the Unstructured Information Management Architecture (UIMA), IBM's open-source text analytics framework. IBM is currently using the system to monitor internal calls to North American help desks. It can be customized to fit clients' needs in any industry.

IBM's text analysis and knowledge mining (TAKMI) system uses natural-language algorithms to sift through written summaries that contact center operators make after each call. While ordinary data mining uses a simple keyword strategy, TAKMI incorporates grammatical relationships into its analysis. By identifying which word is the subject, the verb and the object, TAKMI can categorize calls by product, and whether they are complaints or questions. The tool also generates a color-coded chart that shows which problems are most common with which products, which can be instrumental in catching and correcting problems early on. In addition, the information is useful for creating FAQs (frequently asked questions), which can then be posted on a customer assistance Web page, cutting down on the number of calls human agents must handle. Call operators can also provide more comprehensive information that can help shorten calls and make follow-up calls less frequent.

Boosting contact center agent productivity

IBM's call-center agent buddies (CABs) system is a First-of-a-Kind project aimed at creating new technologies to help boost contact center productivity by decreasing the amount of time it takes to deal with each customer. The effort leverages a series of technologies based on UIMA and IBM Research speech transcription technology to provide an end-to-end solution for clients wishing to improve contact center efficiency, the accuracy of answers and customer satisfaction. First, calls are automatically transcribed into text, which is then analyzed to determine the customer's issue. The problem statement is then dispatched to one of a series of agent buddies, each of which was developed to deal with different types of questions. The correct response is displayed for the human agent to relay back to the customer, significantly reducing the time spent researching answers to customers' questions. An example of an agent buddy is the call transcript matching buddy, which is used to mine previous calls to find the best matches to the incoming call and offer the answer that was given to similar calls.

In addition to using agent buddies to help reduce the average handle time spent on each call, a second component of the system, the transformational diagnostic tool, analyzes the vast pool of previous calls to develop insights into agent and customer interactions. The insights may be used to drive process changes in the contact center, or to build new agent buddies for answering new types of questions.

Making speech recognition more reliable and responsive

Speech recognition technologies rely on clear input that conforms to grammar rules devised for the application. For instance, in an automated drink selection system, the acceptable grammar would be a basic "pick list" of drinks, such as "coffee," "tea" or "soda." A different answer, such as "milk," would generate an out-of-grammar (OOG) exception, prompting the system to restate its query. Other OOG exceptions occur when non-speech events, such as laughter, coughing or background noises, are picked up by the speech recognition system. The voice toolkit contains a grammar editor that enables users to select valid responses, create the grammars, and then create the pronunciations, or phonemes, for any unknown words.

IBM's confidence measure tool is designed to make speech applications more usable when recognition accuracy is less than perfect. Even when accuracy is high, out-of-grammar words or non-speech events must trigger a response of some kind. The confidence measure tool is a method of automatic speech recognition that performs an initial search of a grammar to identify a word hypothesis for an OOG event, and then applies an algorithmic confidence measure to the word hypothesis to determine whether a second search is warranted.

Meeting competitive challenges with the contact center of the future

Enhancing customer satisfaction through conversational technologies and analytics while reducing call volume and cost can help provide a key to success in today's fast-paced global market. The combination of emerging technologies from IBM Research and the CRM expertise of IBM Global Business Services can help provide flexible natural language contact center solutions that allow companies to gain the cost benefits of an automated contact center solution while maintaining a consistent level of service.

For more information on engaging IBM expertise in contact center technologies to help improve your customer relationship management, contact IBM Research Services today