Master Thesis: Implementing a Chatbot hos IKEA Sweden

Category: Software, Machine Learning, Artificial Intelligence
Scope: 2 students completing 30 credits (20 weeks) each

Background

Building chatbots using NLP and AI/ML

Today, the expectations on a website are high. They are expected to be i.e. responsive, fast and intuitive to use. When a webpage grows larger, with more functionality and capabilities, the user can easily get lost on the way. To facilitate the demand for understandability, one approach is to introduce an automated Chatbot. The purpose of this is, by using artificial intelligence and natural language processing, to create a tool for the user to quickly get responses to its questions regarding the website as well as performing automated tasks based on the users input. It will act as an automated support with which the user can directly interact.

 

Currently there are three main types of Chatbots:

  • Rule based chatbots - Bots which simply uses predefined options that the user chooses from. This is great for easier tasks, but when a user request gets more complex, these bots are not sufficient. 
  • Intellectually independent chatbots - These bots use Machine Learning (ML) which helps the chatbot learn from users’ inputs and requests. These bots learns and improve over time using key words and phrases the users' inputs. 
  • AI-powered chatbots - AI-powered bots combines the two variants above, by using NLP and NLG ( natural language generation) to create an interactive experience for the user, while still learning over time.

The prospect and scope of this master thesis is to implement either a Intellectually independent or AI-powered chatbot. A simple chatbot is implemented at the moment, but it's not very intelligent. This needs to either be enhanced or a new bot could be implemented. However, if the existing bot is improved, there's already UI/graphics implemented which could be helpful.

 

Our organization

The overall vision of our organization within IKEA, Customer Engagement, is to establish a customer data and marketing software foundation that enables a world-class, personalized meeting with the Customer.

The Customer Engagement area covers capabilities in areas such as Customer Information, Customer Analytics, Marketing and Loyalty. From a business software architecture perspective, the Customer Engagement area is quite complex since it consists of real time collaboration between global IKEA components and a large number of external partner components.

Customer Engagement is part of the “Reclaim IKEAs data and build new products in the cloud” initiative. In short - we are reclaiming our customer data - including all IKEA Family accounts - which has, up until now, often been managed by local vendors, and are building the capabilities to personalize the customer journey across IKEA touchpoints (digital and physical). Until now the majority of our components are built on GCP (Google Cloud Platform), but other cloud vendors are to be expected when rolling out software globally.

Goals

Investigate the benefits, impacts and challenges with using a NLP-Chatbot.

The main steps of the thesis shall be:

  • Do a literature study of best practices and research around NLP-ChatBots.
  • Implementation of a ChatBot and deploying into a test environment. 
  • Identify key metrics in order to enhance the performance over time.
  • Get feedback from stakeholders.

Keywords: Natural Language Processing, Artificial Intelligence, Software Development.

Observera: De examensarbete och projekt som du hittar i SH Karriär är inte på förväg godkända av ditt universitet. Du måste själv se till att de eventuella samarbeten som du ingår med organisationer för examensarbete och projekt blir godkända av din handledare eller kursansvarig.