Speech Technology Basics, part two: Speech Analytics

"Speech Analytics": In relation to call recording, let's use an example of a Call Center Manager tracking spikes in call volume. In order to help figure out why, an advanced Speech Analytics tool would allow the manager to query a database for the highest used terms and determine what the customers are saying during these calls. Quickly the manager can realize most callers are talking about a recent “Increase in monthly bill”. Another example of Speech Analytics would be to query a database for how many times last month did a caller said the word “Cancel” versus how many accounts were actually closed. This would reveal a “Save” rate for the agents. Let’s say 1,000 Customers called discussing “Canceling their account” yet only 100 accounts were actually closed. This is a 90% Save Rate. These two examples use the concept of “Indexing” to determine what words were said for later query and analys

Speech Analytics can be broken down into two categories: The “Dictionary” based and the “Phonetic” based. The dictionary-based systems require users to define words and phrases in advance so the speech engine can look for or “spot” those words as the engine is analyzing the audio. When these words or phrases are spotted they are indexed for later analysis. The phonetic-based systems make a “Phonetic Track” of the audio as its being analyzed and index all words and phrases for later analysis. At this point you can probably realize that the “Phonetic Based Indexing Engines” are much more robust than the “Large Vocabulary Dictionary Engines”.

From the Wikipedia:
Speech Analytics