Combination of Neural Networks and Conditional Random Fields for Efficient Resume Parsing

摘要

Resume parsing is a technique to extract useful information from resumes for further processing such as resume ranking and selection. Different companies process thousands of resumes during their recruitment process using traditional methods like manual processing and by providing unique resume templates to applicants. The current job recruitment horizon demands better approaches for efficient resume parsing technologies and methods. Even though there are many elementary techniques for parsing the structured documents, they are not suitable for parsing unstructured documents like resumes. The ongoing approaches for resume parsing mainly use regular expressions, chunking, keyword based models and entity recognition models. Relevant to this context, this paper proposes a system for resume parsing using deep learning models such as the convolutional neural network (CNN), Bi-LSTM (Bidirectional Long Short-Term Memory) and Conditional Random Field (CRF). CNN Model is used for classifying different segments in a resume. CRF and Bi-LSTM-CNN models were used for sequence labeling inorder to tag different entities. Pre-trained Glove model is used for word embedding. The proposed system could classify a resume into three segments and extract 23 fields.