This blog post is a publication of the chief executive officer of TC&C, Csaba Juhász’s article published in IT Business.

Csaba Juhász

No Word Is Lost

Analyzing organizations’ customer related communication holds great potential. The recognition of content and corresponding emotional patterns provide decision makers with extremely valuable knowledge.

Customer service and help desk centers have been recording incoming calls for quality assurance purposes for several years now, which means huge unexploited data assets. It is possible to infer customer satisfaction, quality of associates’ work and even the reception of certain products and solutions from the emotional charge of conversations. Since the toolkit of contact centers also includes chat and video in addition to phone calls, the required system must be able to automatically detect and evaluate emotions via all the aforementioned channels in real time.

Telltale Emotions

“We have been dealing with audio recording solutions that can be connected to IP-based telephony systems for nearly two decades, so we have a good understanding of this customer demand. We decided to develop our CARIN system with an artificial intelligence (AI) based smart recorder and for that project we received a 177,7 million HUF support from the NKFIH ”, says Csaba Juhász, the executive director of TC&C. In general, six basic human emotions get differentiated but for the aim of the project and system it is enough to determine whether the customer is satisfied, neutral or dissatisfied. For that mimicry (frowning, louring, smiling), voice (volume, intonation, interruption) are put to use. Transcriptions generated during conversations can be evaluated by the parser for certain words for e.g. interjections. One of the innovations of TC&C’s smart recorder will be to examine emotions on three channels independently from each other, yet simultaneously. Analyses originating from different sources are accumulated by an additional artificial neural network which guarantees a highly reliable outcome.

Recyclable Knowledge

We are planning to develop a high-precision solution by next fall that is multilingual and can be applied for international environments. There is currently no such system in the world and there probably won't be another one by the time we finish developing ours, which is why we see a serious market potential in this ongoing project”, claims Csaba Juhász. One of the most important benefits for contact center operators is that critical calls get automatically displayed for evaluation and revision without requiring associates to specifically search for these records post-call. However, TC&C strives for making use of knowledge acquired during development in completely different areas as well. Machine learning skills of CARIN can be utilized for other purposes, such as for analyzing Big Data-like knowledge bases (call logs, email archives). This allows, for example, to map how information flows between certain departments and groups of the company, and how the organizations’ communication network toward partners and customers looks - as Csaba Juhász mentions some of the great possibilities and potential use cases of the solution.

Hark to Safety

“NAIH recently imposed a 250 million HUF fine on a bank that has been evaluating call center calls with an artificial intelligence based software analyzing customers’ emotions. Decisions of NAIH regarding this domain are not public yet”, said dr. Zoltán Ormós, the head of Ormós Law Firm. The brief summary of the case shows that authorities do not disapprove or exclude AI-driven call evaluation, but consider this specific approach of financial institutions illegal. In case the bank modifies its data management practice so that it complies with the general privacy policy - for example it provides the affected parties adequate information and the right to protest - then there is no obstacle to the software analysis of calls.

Check out our upcoming event if you want to know more.