direkt zum Inhalt springen

direkt zum Hauptnavigationsmenü

Logo der TU Berlin

Conversational Assistant for Home Appliances

Bachelor Thesis

Smart devices have been taking place their places since more than 10 years in our life in order to simplify and automate the regular tasks. Nowadays, the technological improvements have brought the point that the smart devices should not only carry the automated tasks, but also intelligently infer a result or an objective through the observations collected from the environments, the interactions with users, and many other contributors. The aim of this project is to reflect the aforementioned features by developing a conversational assistant that establishes the bridge between users and home appliances.


A typical use-case: A daily life is filled by a number of events for people and the time is devoted to the dish preparation is only mostly limited. This case leads to either ordering dishes in the online platforms or eating in a nearby restaurant. Indeed, most of the people are unaware about the existence of the recipes that can be prepared in a short time. The objective of this use-case is to shed light on how a conversational assistant can help the users to cook a recipe recommended by using the given user preferences and dynamic constraints, which is time in this use-case.


In order to meet the aforementioned challenge, beside the development of a recommendation system, the development of a context-aware chatbot is a fundamental task to manage the dialogues between user and home appliance.


The core work packages of this thesis are listed below:

  1. User Speech Recognition
  2. User Input Analyzer & Intent Prediction
  3. Dialogue Management 
  4. Feedback Analyzer

User Speech Recognition & Web-based Text Interface

This work package requires the development of two user interfaces: speech & textual interfaces. The textual interfaces enable simply entering the text inside the web application and the text will be then directly transmitted to the input analyzer component for the further processes. The most challenging task in this work package is to apply one of the state of art approach to figure out the user identity through the user voice pattern. The voice pattern should be matched to each single user in the environment where the speech interface is used. Developer has to show clearly to what extent the voice pattern detection can be detected.


User Input Analyzer & Intent Prediction

In the previous section, the data will be received from both interfaces in a text format, which has to be analyzed and analyzed correctly to comprehend the true intention of the user. The sentences can be wrong formulated or a syntax or grammatical problem can arise. After correcting the given sentence and shedding the light on the user intention, a query that calls the relevant action must be created.

Dialogue Management 

The core work package of the conversational assistant is based on the reliable dialogue management. Dialogues are composed of many parts such as greetings, content request, feedback collection, etc. How dialog management categorizes correctly what user says and how the conversation fluently is tracked can be counted the essential problems. 

Feedback Collector & Analyzer

The purpose of this task is to provide a way to user in order to give feedback on the offered content. Feedback collection will be realized directly from the user inputs.



-       Programming Languages: Python, Javascript

-       Technological Enablers: Rasa, Webhook



Supervisor: Cem Akpolat
E-mail: Cem.Akpolat@gt-arc.com - Contact me for your questions and further details if you are interested.