This research revolves around utilising Named Entity Recognition (NER) to analyse and categorise data from English translations of pharmacopeias from the Abbasid era, noted for its valuable contributions to science and medicine. The main goal of this work, along with publishing this resource freely, is to assess cross-manuscript NER performance by evaluating the NER model’s performance on unseen corpora and translation styles, as well as demonstrating the transferability of the NER task on such corpora. Two distinct experiments were conducted, focusing on F1-scores differences from mixing source translators and varying training dataset sizes. In experiments involving mixing translator styles, training on a mix of all available styles while accounting for dataset size yielded the best F1-scores compared to even training on the same style as the testing data, while experiments with dataset sizes show diminishing returns of scaling training datasets compared to varying translation styles. This work attempts to enhance the exploration of the medical knowledge embodied in these texts to facilitate their analysis for knowledge extraction relevant to modern medical practices. Furthermore, this research demonstrates strategies to optimise NER results in this context, forming a juncture between digitising historical information and enabling further explorations in pharmacopeia-related Natural Language Processing research.

Reuse of learnt knowledge is of critical importance in the majority of knowledge-intensive application areas, particularly because the operating context can be expected to vary from training to deployment. Dataset shift is a crucial example of this where training and testing datasets follow different distributions. However, most of the existing dataset shift solving algorithms need costly retraining operation and are not suitable to use the existing model. In this paper, we propose a new approach called reframing to handle dataset shift. The main objective of reframing is to build a model once and make it workable without retraining. We propose two efficient reframing algorithms to learn the optimal shift parameter values using only a small amount of labelled data available in the deployment. Thus, they can transform the shifted input attributes with the optimal parameter values and use the same existing model in several deployment environments without retraining. We have addressed supervised learning tasks both for classification and regression. Extensive experimental results demonstrate the efficiency and effectiveness of our approach compared to the existing solutions. In particular, we report the existence of dataset shift in two real-life datasets. These real-life unknown shifts can also be accurately modeled by our algorithms.

In this paper, we present a new approach to enhance the performance of Bayesian classifiers. Our method relies on the combination of two ideas: pairwise classification on the one hand, and threshold optimization on the other hand. Introducing one threshold per pair of classes increases the expressivity of the model, therefore its performance on complex problems such as cost-sensitive problems increases as well. Indeed a comparison of our algorithm to other cost-sensitive approaches shows that it reduces the total misclassification cost.