Continuous Active Learning Using Pretrained Transformers

Technology Assisted Review (TAR) was first introduced The main objective of High recall information retrieval by Grossman and Cormack (2011) and it is to retrieve virtually all relevant documents revolutionized how legal e-discovery is being done to an information need, while minimizing the number around the globe. The most successful implementations of non-relevant documents returned. This is of TAR is a human-in-the-loop setting where particularly useful is high-stake settings. For example, the relevance feedback from user is used to improve a lawyer might need to find all the relevant the model. This implementation is referred information to a specific trademark infringement, to as Continuous Active Learning (CAL). The current or an epidemiologist might need to find all the state-of-art is an implementation by Grossman relevant documents to inner workings of certain et al. (2016), introduced as part of the High Recall protein structures.