Main Article Content
Abstract
The process of training model in supervised machine learning is more difficult than expected due to the challenges in labeling and annotating data. It has observed that majority of the organization dealing in Artificial Intelligence projects have run in to problems with labeling data to train models. Labeling is commonly done manually by domain experts, which is time consuming task. Many authors have given different approaches to reduce the burden of manual labeling. However, all approaches have been facing different challenges due to increasing volume and shape of the data which further degrades the performance of automation. In order to produce quality in AI projects, the training data must be correctly labelled. The paper presents various challenges and opportunities occur in dealing with unstructured textual data for labeling to produce training data at the expected quality. The paper would also help the readers or scholars to purse their research projects in the area of text analytics or natural language processing.