New Voice-Recognition Technology to Improve Quality of Call Centers

ATLANTA -- Nexidia, a provider of audio search and speech analytics, launched the Language Assessor yesterday. This automated technology will evaluate offshore agents on three criteria: ability to read, accuracy of pronunciation, and fluency compared to the clarity of the targeted audience.

Anna Convery, Nexidia's senior vice president of marketing and product management, says that the new product addresses two main components necessary to running a call center: recruiting (or "on-boarding") of agents, and constant attention to the quality of those agents. The issue of agent quality, she says, is a critical building block for success. "The quality of agents is very tightly aligned with the quality of service the customers are getting."

And comprehension, as always, remains a two-way street. Elizabeth Herrell, a vice president with Forrester Research, acknowledges that offshore agents may be very capable in writing and reading another language, but "regardless of how good you are, if you can't be understood [by the customer], it's very limiting." To address this, Nexidia says that Language Assessor uses the basic component of a language—a phoneme, or the representation of sound—to assess agents.

Currently, many corporations that use offshore agents conduct their evaluations manually. After prospective agents are given a script to read, the reading is recorded and evaluated. Scripts, Convery says, are "very deliberately architected" to target phonemes that are typically difficult for offshore agents to pronounce when speaking the language of the customer. The primary problem with manual evaluations, however, is inconsistency due to what Convery calls "listener fatigue." Language evaluators have to listen to hundreds upon hundreds of recordings; eventually, their tolerance for what qualifies as an "acceptable" pronunciation of a particular word becomes much more lenient.

Language Assessor relies on automation to resolve the problem of inconsistency. Similar to the manual process, a prospective agent is
recorded reading a script. The recording is then loaded into the application, which gives the individual a "global score," a "pronunciation score," and a "fluency score." The program even breaks down each sentence, phrase, or word the individual has read according to a color-coded system that highlights the phonemes she is good, questionable, or poor at pronouncing. These scores allow organizations to determine who is highly proficient, needs training, or cannot manage even with training, which could subsequently influence their hiring decisions.

Another advantage of Language Assessor is that it can provide a convenient means for organizations to periodically evaluate their existing employees to ensure that they continue to possess the skills necessary to service customers. "Working so hard to run a call center, it is often the training and support of the agents that suffers," Convery says. If skills are not up to standards, organizations will then be able to consider supplementary training options that can, perhaps, target only specific areas of concern.

Convery also adds that Language Assessor, though automated, can be tailored to the needs of the individual organization. For instance, a North American company that uses outsourcing to non-English-speaking countries can decide that, although certain phonemes are difficult to pronounce, they generally do not impede customer comprehension, and thus, can be overlooked during evaluation.

With the advent of this technology, Nexidia believes call centers can take their customer service to the next level. While this is, for now, a niche market, Herrell says, "[Nexidia] is responding not just to a demand, they are solving a problem."