Automated call classification

A method of categorizing digital data records, each data record relating to an interaction between a call center agent and a customer, each interaction being an exchange between parties including one or more of:

• an exchange via telephone, an exchange with an interactive voice response “IVR” system, a web chat, an exchange via instant messaging, an exchange via short message service “SMS” messages, and an exchange via email, each data record having one or more features and the method being carried out in a computer, the method comprising:
• using a set of soft rules relating at least one of said features to a category to determine a first probability that a data record belongs to the category,
• using a set of trained probabilistic models relating at least one of said features to said category to determine a second probability that a data record belongs to the category, converting the first probability into a first vote by comparing the first probability to a threshold and converting the second probability into a second vote by comparing the second probability to the threshold;
• and
• determine on the basis of the first and second probabilities whether the data record belongs to a category by:
• if the first vote and second vote agree, determining whether the data record belongs to the category based on the votes, and if the votes disagree determining that the data record belongs to the category based on the highest of the first probability and second probability weighted by historical performance.

A system for determining reasons for interactions between call center agents and customers, each interaction being represented as a data record having one or more features indicative of the reason for the call, and each interaction being an exchange between parties including one or more of:

• an exchange via telephone, an exchange with an interactive voice response “IVR” system, a web chat, an exchange via instant messaging, an exchange via short message service “SMS” messages, and an exchange via email, the system comprising:
• a memory;
• and
• a processor comprising one or more processing modules including:
• a rules module configured to receive as input a data record and to determine a first probability that the interaction was for a particular reason based one or more rules, each rule defining a relationship between at least one of said features and the reason;
• a probabilistic module configured to receive as input said data record and to determine a second probability that the interaction was for said particular reason based on one or more probabilistic models of relationships between one or more of said features and the reason;
• an arbitration module configured to convert the first probability into a first vote by comparing the first probability to a threshold and convert the second probability into a second vote by comparing the second probability to the threshold and if the first vote and second vote agree, determine whether or not the interaction was for said reason based on the votes, and if the votes disagree determining whether or not the interaction was for said reason based on to the highest of the first probability and second probability weighted by historical performance;
• and
• an action module configured to receive from the arbitration module an indication that an interaction was for said particular reason and to initiate an action appropriate to the reason.