interaction mining: the new frontier of call center analytics
DESCRIPTION
Paper presented at the DART 2011 workshop in Palermo. The paper introduces a new type of call center analytics based on interaction mining. It shows how advanced metrics and KPIs for call center quality management can be implemented through interAnalytics NLP technology.TRANSCRIPT
© 2011 interAnalytics 1
Interaction Mining: the new frontier of Call Center Analytics
Vincenzo Pallotta Rodolfo Delmonte Lammert Vrieling
David Walker
© 2011 interAnalytics 2
Outline
• Call Center Analytics• Automatic Argumentative Analysis for
Interaction Mining• Experiments with Call Center Data• Conclusions
3
CALL CENTER ANALYTICS
© 2011 interAnalytics
© 2011 interAnalytics 4
Call Center Analytics
• Call centers data represent a valuable asset for companies, but it is often underexploited for business purposes because:– it is highly dependent on quality of speech recognition
technology– it is mostly based on text-based content analysis.
• Interaction Mining as a viable alternative:– more robust– tailored for the conversational domain– slanted towards pragmatic and discourse analysis
© 2011 interAnalytics 5
Mainstream Call Center Analytics
Does not unveil real
insights about customer
satisfaction
© 2011 interAnalytics 6
Call Center Analytics: metrics and KPIs
• Agent Performance Statistics: – Average Speed of Answer, Average Hold Time, Call Abandonment Rate,
Attained Service Level, and Average Talk Time. – Quantitative measurements that can be obtained directly through ACD
(Automatic Call Distribution), Switch Output and Network Usage Data.• Peripheral Performance Data:
– Cost Per Call, First-Call Resolution Rate, Customer Satisfaction, Account Retention, Staff Turnover, Actual vs. Budgeted Costs, and Employee Loyalty.
– Quantitative, with the exception of Customer Satisfaction that is usually obtained through Customer Surveys.
• Performance Observation: – Call Quality, Accuracy and Efficiency, Adherence to Script,
Communication Etiquette, and Corporate Image Exemplification. – Qualitative metrics based on analysis of recorded calls and session monitoring
by a supervisor.
© 2011 interAnalytics 7
Four objectives
1. Identify Customer Oriented Behaviors, – which are highly correlated to positive customer ratings
(Rafaeli et al. 2007);
2. Identify Root Cause of Problems – by looking at controversial topics and how agents are able to
deal with them;
3. Identify customers who need particular attention – based on history of problematic interactions;
4. Learn best practices in dealing with customers – by identifying agents able to carry cooperative
conversations.
© 2011 interAnalytics 8
ARGUMENTATIVE ANALYSIS FOR INTERACTION MINING
© 2011 interAnalytics 9
Argumentative Structure of ConversationsDISCUSS(issue) <- PROPOSE(alternative)
1702.95 David: so - so my question is should we go ahead and get na- - nine identical head mounted crown mikes ? {qy} 61a
REJECT(alternative)1708.89 John: not before having one come here and have some people try it out . {s^arp^co} 61b.62a
PROVIDE(justification)1714.09 B: because there's no point in doing that if it's not going to be any better . {s} 61b+
ACCEPT(justification)
1712.69 David: okay . {s^bk} 62b
PROPOSE(alternative)
1716.85 John: so why don't we get one of these with the crown with a different headset ? {qw^cs} 63a
ACCEPT(alternative)1721.56 David: yeah . {s^bk} 63b1726.05 Lucy: yeah . {b} 1727.34 John: yeah . {b}
PROVIDE(justification)
1722.4 John: and - and see if that works . {s^cs} 63a+.64a 1723.53 Mark: and see if it's preferable and if it is then we'll get more . {s^cs^2} 64b1725.47 Mark: comfort . {s}
PROVIDE(justification)1714.09 John: because there's no point in doing that if it's not going to be any better . {s} 61b+
Why was David’s proposal on microphones rejected?
© 2011 interAnalytics 10
Automatic Argumentative Analysis
• Based on the GETARUNS system1.• Clauses in Turns are labelled with Primitive Discourse
Relations: – statement, narration, adverse, result, cause, motivation,
explanation, question, hypothesis, elaboration, permission, inception, circumstance, obligation, evaluation, agreement, contrast, evidence, hypoth, setting, prohibition.
• And then Turns are labelled with Argumentative labels:– ACCEPT, REJECT/DISAGREE, PROPOSE/SUGGEST,
EXPLAIN/JUSTIFY, REQUEST EXPLANATION/JUSTIFICATION.
1 Delmonte R., Bistrot A., Pallotta V.,Deep Linguistic Processing with GETARUNS for spoken dialogueUnderstanding. Proceedings LREC 2010 (P31 Dialogue Corpora).
© 2011 interAnalytics 11
Evaluation
Correct Incorrect Total Found Precision
Accept 662 16 678 98%
Reject 64 18 82 78%
Propose 321 74 395 81%
Request 180 1 181 99%
Explain 580 312 892 65%
Total 1826 421 2247 81.26%
Precision: 81.26% Recall: 97.53%
ICSI corpus of meetings (Janin et al., 2003)
Delmonte R., Bistrot A., Pallotta V.,Deep Linguistic Processing with GETARUNS for spoken dialogueUnderstanding. Proceedings LREC 2010 (P31 Dialogue Corpora).
© 2011 interAnalytics 12
EXPERIMENTS WITH CALL CENTER DATA
© 2011 interAnalytics 13
Rationale: implement the four objectives
1. Identify Customer Oriented Behaviors, 2. Identify Root Cause of Problems 3. Identify customers who need particular
attention 4. Learn best practices in dealing with
customers
© 2011 interAnalytics 14
The Data
• Corpus of 213 manually transcribed conversations of a help desk call center in the banking domain.
• Average of 66 turns per conversation.• Average of 1.6 calls per agent. • Collected for a study aimed at identifying
customer oriented behaviors that could favor satisfactory interaction with customers (Rafaeli et al. 2007).
© 2011 interAnalytics 15
Identify Customer Oriented Behaviors
• Based on the work of Rafaeli et al. 2006.• Customer Oriented Behaviors– anticipating customers requests 22,45%– educating the customer 16,91%– offering emotional support 21,57%– offering explanations / justifications 28,57%– personalization of information 10,50%
© 2011 interAnalytics 16
Significant correlation with argumentative labels
© 2011 interAnalytics 17
Identify Root Cause of Problems
• Cooperativeness score – a measure obtained by
averaging the score obtained by mapping argumentative labels of each turn in the conversation into a [-5 +5] scale.
• Sentiment Analysis module.
Argumentative Categories Cooperativeness
Accept explanation 5
Suggest 4
Propose 3
Provide opinion 2
Provide explanation/justification 1
Request explanation/justification 0
Question -1
Raise issue -2
Provide negative opinion -3
Disagree -4
Reject explanation or justification -5
© 2011 interAnalytics 18
Top 20 Controversial Topics with average cooperativeness scores and sentiment
© 2011 interAnalytics 19
Cooperativeness of speakers on top discussed topics
© 2011 interAnalytics 20
Identify problematic customers
© 2011 interAnalytics 21
Select a specific customer
© 2011 interAnalytics 22
Visualize a selected call
© 2011 interAnalytics 23
CONCLUSIONS
© 2011 interAnalytics 24
Conclusions
• New Generation Call Center Analytics requires Interaction Mining– Call Center Qualitative metrics and KPIs can be only
implemented with a full understanding of the customer interaction dynamics
• Argumentation is pervasive in conversations.– In order to recognize argumentative acts, advanced Natural
Language Understanding is necessary.• Future work:– Scalability: need to process millions of call per day!– Multi-language: call centers all over the world.
© 2011 interAnalytics 25
The Team
www.interanalytics.ch…find us at