Lake Corporation’s XCalibur® Voice Analytics phonetic search engine provides three valuable advantages when compared with conventional speech-to-text recording:  

• The first of these is a superior understanding of what transpired during a call – not just what was said, XCalibur® Voice Analytics captures the nuances as well as the words

• Next, much faster searching of call recordings; both to look for clues to improved business outcomes and pointers to agent and operational effectiveness

• And third, the technology cost advantage; XCalibur® Voice Analytics requires far less investment in processing power

 

Superior understanding – Whereas XCalibur® Voice Analytics captures what is actually meant as well as what is said, conventional speech-to-text conversion – otherwise known as Large Vocabulary Continuous Speech Recognition – is only able to record words anticipated by a nominated vocabulary dictionary.

​

So unanticipated words will not be found when the recording is subsequently searched. This won’t matter if the caller uses a colloquialism in a context that makes its meaning obvious. If, however, the caller names a competitor offering better terms, and that company’s name is not in the dictionary, an opportunity may have been lost.

​

Faster searching – Speech-to-text technology searches audio recording many times faster than searching manually. However, the prerequisite conversion of the recording into a searchable format is time-consuming and the typical end result is a text search capacity only two or three times that of manual searching.

​

The XCalibur® Voice Analytics phonetic search technology is much more advanced. It transforms recorded audio into phonetic representation – including the way words are pronounced – rather than into written words. Preparation of the content for searching is very much faster than manual time and searching itself can be many thousands of times faster than other methods. 

​

Cost advantage – Converting recorded calls for searching using speech-to-text technology demands a sophisticated language model and a large vocabulary dictionary for successful word recognition (and no matter how big the vocabulary, many potentially important words will never be captured).