Natural language processing in customer service

Natural Language Processing is a branch of artificial intelligence that deals with how computers communicate with human beings naturally. A single application of Natural Language Processing such as a chatbot can handle 80% of customer communications. Recent research by Gartner predicts that 30% of interactions with technology would be through conversations with smart machines.

Here are some of the top uses of Natural Language Processing in customer service

1. Leveraging speech recognition technology in call centers

Speech recognition applications are being used by companies for automatic conversion of speech into text or a machine-readable format. They provide customer support by handling incoming calls throughout since they can be used even when the call center agents are not available. These applications are being integrated into call centers to improve collections and conduct surveys.

2. Customer Sentiment analysis

Understanding customers' views of a product or service are key to the sales performance of any company. Natural Language Processing techniques are used to identify the sentiment, text content and intent in any customers' message. This helps the company adjust its products or services depending on the customer reviews and improve customer satisfaction.

3. Customer service chatbots

Natural Language processing is key to the development of retrieval-based chatbots. Research shows that by the end of 2020, 80% of businesses will have some form of chatbot integration available to their customers. This is because chatbots provide real-time self-service support.

4. Interactive Voice systems Call routing

Most of the times when a customer calls a call center is asked to select some options before being connected to the relevant department e.g. when calling a Safaricom customer care, you are asked to dial 1 or 2 for language and other keys depending on the service you are intending to get assistance on. With Interactive voice response call routing that utilizes Natural Language Processing you just mention the problem e.g. ‘send my Mpesa monthly statements ‘and the system automatically diverts your call to the relevant department. This saves the amount of time spent on attending to a customer.

5. Text classification

Text classification is among the most useful Natural Language Processing tasks and can be applied in a wide range of areas in customer service. This includes;

  • Text classification for customer feedback
  • Sorting incoming customer messages according to their languages
  • Analyze customer survey responses about a particular brand.

Text classification applications save time that a particular customer service representative would use to manually sort through documents, reviews or chats. 

NLP Case studies

1. Improving customer care with Natural Language Processing (Uber)

Uber receives millions of support tickets from riders, drivers and eateries per week all with thousands of different issues. An issue is a problem that the user sends to the customer service representative.  The complexity of the issues increases as the company increases the lines of business such as the introduction of bikes. From a single support ticket, a customer service representative has to figure out the type of issue, map the issue into a content type tree, determine whether an action has to be taken and provide a response to the user which is drawn from thousands of replies templates. This takes a lot of time and a large number of customer service representatives are required to handle the large volumes of support tickets.

How Uber is using Natural Language Processing to support their customer service representatives

Customer Obsession Ticket Assistant. (COTA)

This Natural Language Processing and Machine Learning application takes the incoming support ticket, understands the intent of the customer using the meta-information provided, provides suggested content types, actions and the 3 most relevant reply templates to the customer service representative. The customer service representative then chooses to use any of the reply templates.

One-click chat. (OCC)

One-click chat utilizes Natural Language Processing and Conversational AI to facilitate driver communication via a smart reply system. It effectively uses a small set of labeled examples to identify the intent in a customer’s message. It allows driver-partners to respond to incoming messages from users in one push of a button.

2. Improving the shopping experience for online store customers (Klevu)

Klevu provides an instant site search solution for e-commerce stores. Though text classification, the company provides relevant search results for shoppers and actionable insights for store owners. Its search technology is designed to provide an intuitive responsive and enjoyable shopping experience for store customers.

SignAll has developed a technology leveraging on computer vision machine learning and Natural Language Processing algorithms to help businesses and education institutions recognize and translate sign language into English text and display it on a chat dialogue. This has enabled brands to serve deaf customers better by making their services and products more accessible and understanding their needs better.

It is clear from the above use cases that Natural Language Processing will remain an important aspect of customer service for all businesses. The ever-growing field of Artificial Intelligence and an increase in unstructured data available to companies from social media and websites further stresses the relevance of Natural Language Processing in customer service. With most companies developing apps and websites to improve customer experience integration of Natural Language Processing in analyzing feedback will be inevitable.


Machine Learning at Uber (Natural Language Processing Use Cases)