Skip to main content Skip to main navigation


Cases without borders: Automating Knowledge Acquisition Approach using Deep Autoencoders and Siamese Networks in Case-based Reasoning

Kareem Amin; Stelios Kapetanakis; Klaus-Dieter Althoff; Andreas Dengel; Miltos Petridis
In: 2019 IEEE 31st International Conference on Tools with Artificial Intelligence. International Conference on Tools with Artificial Intelligence (AAAI SSS-2019), November 4-6, Portland, OR, USA, Pages 133-140, ISBN 978-1-7281-3798-8, IEEE Digital Library, 11/2019.


Finding an ideal text case-base representation for a new domain is important for a CBR [1] system. To do this, the choice of an ideal representation is guided by the domain characteristics and the complexity of its cases. Recently, the explosion of deep learning techniques and other forms of vectorised representations, has provided a new source for case insights. Richer text features can be extracted and used for each case if required. In this paper, we build on recent work in this area and generate richer case representation by automatically acquiring domain knowledge from unstructured sentences. We describe how the Deep Knowledge Acquisition Framework obtains its representation vectors from stemmed words and improving these vectors iteratively, suggesting high quality outputs and relevance to domain experts based on either explicit queries or their past intentions. We evaluate these ideas using two, not related, datasets from the automotive and legal domains respectively. The results show the benefits of combining Autoencoders and Siamese Networks in CBR while achieving better textual data dimensionality reduction, data de-noising and similarity measures.

Weitere Links