[1910.09532v2] Building Dynamic Knowledge Graphs from Text-based Games
To better understand the behavior of our proposed models on TextWorld KG, we conduct an error analysis, which we show in Appendix
Abstract We are interested in learning how to update Knowledge Graphs (KG) from text. In this preliminary work, we propose a novel Sequence-to-Sequence (Seq2Seq) architecture to generate elementary KG operations. Furthermore, we introduce a new dataset for KG extraction built upon text-based game transitions (over 300k data points). We conduct experiments and discuss the results.
‹Figure 1: Illustration of an example in TextWorld KG. By issuing an action At−1 at game step t − 1, the environment returns a new observation, Ot. Given the KG at step t − 1, a model is required to predict the new KG given the text observation. (Introduction)Figure 2: Graph update operation generation model. (Model Architecture)Figure 3: Average TF-F1 scores grouped by verbs. (Error Analysis)›