Efficient search with WeldGalaxy chatbot

AeonX AI is one of the partners of the EU H2020 funded project WeldGalaxy. The core feature of the WeldGalaxy platform is to search for information in the Knowledge-Based Engineering (KBE) system and in the Dynamic Knowledge Management (DKM) platform. In a standard way, a user can search by text phrases and filter by choosing parameters using many dropdown menus in the KBE platform. WeldGalaxy chatbot helps the user to get to a specific result conversationally, making it an efficient process. The same interface can respond to frequently asked welding questions on welding technologies.


The WeldGalaxy chatbot (Chatbot) is a user-friendly channel to search for information in the KBE. The queries can be in form of unstructured natural language inputs. These unstructured user queries are then processed and parameterized by AeonX NLP Engine and finally, AeonX Bot Engine performs a parameterized search in the KBE system. In case any necessary parameter is missing from the user input, the chatbot engages in conversation with the user until all the parameters are provided correctly (Figure 1, Figure 2). The chatbot responses can be in the form of text and/or interactive UI cards that contain interactive buttons, images, and charts.



The chatbot is powered by a proprietary AeonX Natural Language Processing Engine (AeonX NLP Engine) that is customized and improved for the WeldGalaxy Project. AeonX NLP engine is supported by AeonX NLP Dashboard. Using this dashboard, a bot designer can easily prepare training data, train and deploy NLP models to production. AeonX NLP engine and NLP Dashboard supports English, French, German and Spanish. This allows users to query in any of the 4 languages and get a response from the chatbot. AeonX Bot Engine connects to the KBE to prepare responses (Figure 2). 

Figure 1: A high-level view of the WeldGalaxy chatbot usage together with KBE and DKM
Figure 2: WeldGalaxy Chatbot is fully integrated with the DKM platform and KBE system. The chatbot can be activated after the sign-in process DKM platform.

AeonX NLP model is regularly updated to understand new entries in the WeldGalaxy knowledge base. In WeldGalaxy, the AeonX NLP model handles over 1000 classes of information. To help the users phrase queries, the chatbot messenger provides suggestions during typing. Also, some of the query results, such as welding recommendations, can be saved in the DKM platform for later use.  The WeldGalaxy chatbot is deployed in Azure and has full SSO support through the DKM identity service.

Figure 3: The chatbot messenger suggests phrases when a user starts typing a query. Users can choose to select one of the phrases to get a quick response.

We aim to make it easy for the WeldGalaxy platform users to retrieve welding information from the welding knowledge base.