Page 59 - CBM13

~ Viva Brazil ~

rtificial Intelligence (AI) companies like As a cutting-edge AI and legitimate technology, NLP
the Brazilian Intellimetri are looking to and voice assistants are now a part of our everyday lives,
Avastly improve how businesses interact with whether on e-commerce platforms, reservation centres or
customers. They do this to increase lead conversion, home assistants like Alexa, Siri and Google Home. NLP
improve customer experience and foster loyalty. and AI leverage will bring intelligence to the call centres
and contribute to reduce labour costs, freeing humans from
One of the areas that AI disrupts is customer care, especially redundant work as machines make call centre services more
the repetitive and labour-intensive world of the e-commerce personalised, automated, efficient and convenient.
call centres. Call centres are being revolutionized through
AI-empowered technologies such as Natural Language At Intellimetri, we develop NLP solutions from scratch
Processing (NLP). using speech to text technology, machine learning and deep
learning associated with domain knowledge to develop
There are mainly two types of call centres. First, there is customised solutions. Our algorithm identifies words used
inbound call centres where customers call the operators and by the client and the call centre operator. After that, we
tend to be operated by a company to administer its post- create word embeddings analysing the content and context of
sales services. Second, there’s outbound call centres, where those calls, looking for the degree of similarity between the
the operators call the customers. Outbound call centres are script suggested by the call centre’s manager and the script
normally used for pre-sales, sales campaigns and market effectively used by the operator.
research.
This way, we can find inconsistences between these two
The technology used in call centres is divided into three scripts. It guides decision makers regarding the level of user
categories: real-time speech recognition, conversation experience and degree of attachment of the operator script
management - which ensures simultaneous processing to that one provided by managers. We can also automate
of multi-pass conversations - and conversational and call classification regarding its quality. After this analysis, a
intent analysis. The latter broadly analyses users’ dialogue, huge amount of data is generated and is available in the form
considering semantics. of strategic insights. Dashboards are then generated in real
time so that operators get their feedback and improve their
E-commerce call centres are focused on inbound customer performance.
service operations. Here, intelligent voice assistants and
chatbots are already commonly used to deal with countless Given that most libraries in NLP are developed for the
customer queries. Systems can understand and meet needs for English language, we did a special effort to adapt and use
various types of e-commerce services. these libraries for the Portuguese language. Word embeddings
of trained Wikipedia where used and similarity coefficients
Data analysis of call records is the brand new and fast- helped us to make our analysis, including sentiments involved
growing application for NLP technology in call centres’ in the transactions.
scenario. In this case, in order to provide better and more
personalized services, machines can be assigned to deal We consider human interaction and NLP the most beautiful
with users’ emotions, intentions, and potential thoughts by area to direct our efforts and one of the most promising areas
analysing speech. for real artificial intelligence beyond the hype. These are real
business applications that will help transform organisations
Demand is increasing for this type of service in the into innovative, efficient and sustainable companies of the
e-commerce sector. E-commerce platforms usually have future. ••
massive amounts of user phone consultations and text
message data. Through NLP technology, a machine can About the authors:
automatically organise and analyse this data. Unstructured Rubens Zimbres is Chief Data Scientist at Intellimetri. He has a Master
data analysis technologies can be used, for instance, to and PhD in Business Administration with an emphasis on Artificial
provide users with more customised product recommendation Intelligence. He develops Algorithms for Machine Learning, Deep
and help reduce user complaint rates. Learning, Natural Language Processing and Cloud solutions.

As more and more merchants are now recording and Gerson Rolim is Partner and Director of Vecto Mobile Intellimetri and
retaining their customer service data, analysis of calling Observatore, and Director of ABINC and camara-e.net. He is a pioneer of
records can be expected to grow. NLP technology is able the Brazilian Internet, connecting the first public networks. Rolim has
to parse text data from a calling system. This way, it can 23 years of experience in ICT and is a post-grad in Computer Network
discover and understand the content to leverage e-commerce and a grad in Computer Science. He also has experience in the technical,
merchant’s profitability and increase customers’ satisfaction commercial and management sectors.
and their online experience.

59