Text entry in Virtual Reality (VR) is becoming an increasingly important task as the availability of hardware increases and the range of VR applications widens. This is especially true for VR industrial applications where users need to input data frequently. Large-scale industrial adoption of VR is still hampered by the productivity gap between entering data via a physical keyboard and VR data entry methods. Data entry needs to be efficient, easy-to-use and to learn and not frustrating. In this paper, we present a new data entry method based on handwriting recognition (HWR). Users can input text by simply writing on a virtual surface. We conduct a user study to determine the best writing conditions when it comes to surface orientation and sensory feedback. This feedback consists of visual, haptic, and auditory cues. We find that using a slanted board with sensory feedback is best to maximize writing speeds and minimize physical demand. We also evaluate the performance of our method in terms of text entry speed, error rate, usability and workload. The results show that handwriting in VR has high entry speed, usability with little training compared to other controller-based virtual text entry techniques. The system could be further improved by reducing high error rates through the use of more efficient handwriting recognition tools. In fact, the total error rate is 9.28% in the best condition. After 40 phrases of training, participants reach an average of 14.5 WPM, while a group with high VR familiarity reach 16.16 WPM after the same training. The highest observed textual data entry speed is 21.11 WPM.

Gated Recurrent Neural Networks (GRNNs) are important models that continue to push the state-of-the-art solutions across different machine learning problems. However, they are composed of intricate components that are generally not well understood. We increase GRNN interpretability by linking the canonical Gated Recurrent Unit (GRU) design to the well-studied Hopfield network. This connection allowed us to identify network redundancies, which we simplified with an Input Residual Connection (IRC). We tested GRNNs against their IRC counterparts on language modelling. In addition, we proposed an Input Highway Connection (IHC) as an advance application of the IRC and then evaluated the most widely applied GRNN of the Long Short-Term Memory (LSTM) and IHC-LSTM on tasks of i) image generation and ii) learning to learn to update another learner-network. Despite parameter reductions, all IRC-GRNNs showed either comparative or superior generalisation than their baseline models. Furthermore, compared to LSTM, the IHC-LSTM removed 85.4% parameters on image generation. In conclusion, the IRC is applicable, but not limited, to the GRNN designs of GRUs and LSTMs but also to FastGRNNs, Simple Recurrent Units (SRUs), and Strongly-Typed Recurrent Neural Networks (T-RNNs).