Browse Prior Art Database

Monitoring Voice Data for Specific Content Using Unobtrusive Speech Recognition Technology

IP.com Disclosure Number: IPCOM000030337D
Original Publication Date: 2004-Aug-06
Included in the Prior Art Database: 2004-Aug-06
Document File: 1 page(s) / 33K

Publishing Venue

IBM

Abstract

Call centers often perform quality monitoring. Indeed, many businesses monitor some percentage of inbound / outbound calls to their employees. This is typically done by some sort of pseudo-random, even distribution. A normal statistical quality monitoring system may be in place which records (saves?) the typically statistic samples as normal. This technology can be expanded for additional use in call centers or, in reality, as an adjunct to any telephone switch. This concept proposes an attached speech recognition system that caches and records in temporary memory 100% of the calls. The adjunct system listens for particular keywords or phrases during these calls. If a specified match occurs to various keywords, the application saves the temporary recording (commits it to permanent memory) as well as executing some configured action once the keywords are first identified and once the call is completed. Other systems may be monitoring / recording these calls as well for background, detailed analysis or other purposes). This allows for real-time actions to be taken automatically based on keyword triggers and allowing the organization in question to respond more quickly to critical situations and records the calls in question for later review. This concept has wide-spread potential use: inside call centers, office environments, law enforcement, etc. and can be provided as an interface to any voice/audio communication channel (telephony, voice over IP, standard PC microphone / chat)

This text was extracted from a PDF file.
This is the abbreviated version, containing approximately 52% of the total text.

Page 1 of 1

Monitoring Voice Data for Specific Content Using Unobtrusive Speech Recognition Technology

This concept proposes a speech recognition system is attached to the related voice/audio communication channel which monitors 100% of data interactions and takes actions based on key information.

Examples / specifics:

In a call center, maybe the customer says "This is great! Thank you so much!" or related terminology, we record the call and forward it in an email attached as a WAV file to the agent's manager with a subject line of "good job" In a call center or office perhaps the customer says "I'm so disappointed. I want to speak with your manager" we record the call and immediately attempt to conference in a team leader or manager. Maybe on a public or private switch this application listens for keywords like "hack", "bomb", "operation Mickey Mouse", excessive profanity hits, etc. Once triggered, the recording is forwarded to someone to evaluate the recorded call to determine the various party's intent. As part of intelligence gathering for the government/military, sorting through the various data to determine what is worthy of further scrutiny is important - by searching on known key words calls that fit a confidence level could be recorded and evaluated later by an expert - in addition to a normal statistical sampling. The system could initiate a trace to determine ANI / DNIS / caller id / cell phone geographic location, etc. upon hearing specified keywords. Unless the call data triggers an action to co...