Springe direkt zu Inhalt

Comparison of Chatbot Development Frontends for Rasa


  • Solid understanding of software development
  • Knowledge or interest in Chatbots, Rasa, NLP
Academic Advisor
Chatbots, Conversational Interfaces, Software Development, Distributed Systems, NLP
Bachelor of Science (B.Sc.) or Master of Science (M.Sc.)



Chatbots have recently gained popularity, due to advances in artificial intelligence. In contrast to graphical user interfaces, where interaction is precisely defined in terms of action and reaction (i.e. 'click here to do this'), the interaction between a human and a chatbot is more fuzzily defined, because it takes place in natural language. Chatbots apply data-driven methods (NLU: Natural Language Understanding) to decode (or rather guess) the user's intent and the content from their messages. Based on that, appropriate actions are initiated and a natural language response is generated (Figure 1). [1] [2] [3] This process is modeled in modern Chatbot frameworks like the Rasa Framework [4]. In an upcoming project this framework will be used to build Chatbots that are deployed on the messaging platform Mattermost [5].

[Caption] Figure 1: Chatbot Architecture [1]


Current Chatbots cannot handle any type of conversation automatically. Instead, dialogs have to be created manually by a developer. Furthermore, the data-driven methods for Natural Language Understanding have to be supplied appropriate data, as well as trained and evaluated.

Several Chatbot Development Frontends exist, that aim to facilitate the creation of dialogs and context-specific datasets (i.e. intents, entities, synonyms, actions etc.) for the Rasa Framework. [6][7][8] Evaluating these solutions is easiest with hands-on experience, because the intricate differences between them become apparent during development and the importance of their individual advantages and disadvantages is dependent upon the specific application context.


The objective is to compare three different Development Frontends [6][7][8] for the Rasa Framework. This will be achieved by building a proof-of-concept Chatbot for a specific use-case, that will be deployed on the messaging platform Mattermost [10].

Possible Procedure

  1. Literature Review
  2. Definition of use case
  3. Setup of infrastructure (Mattermost, Rasa, Chatbot Development Frontends, Connection to APIs)
  4. Comparison of different Rasa Frontends
  5. Creation of context-specific Dataset
  6. Dialog creation
  7. Evaluation


[1] https://towardsdatascience.com/architecture-overview-of-a-conversational-ai-chat-bot-4ef3dfefd52e

[2] Følstad, A., & Brandtzæg, P. B., (2017). Chatbots and the new World of HCI. Interactions 24 (4), 38 - 42. doi:10.1145/3085558

[3] Radziwill, N. M., & Benton, M. C. (2017, April 15). Evaluating Quality of Chatbots and Intelligent Conversational Agents.

[4] https://rasa.com/

[5] https://github.com/paschmann/rasa-ui

[6] https://github.com/jackdh/RasaTalk

[7] https://github.com/samtecspg/articulate

[8] https://mattermost.com/