Semantic Search and Visualisation using Machine Learning and Natural Language Processing
This project tackles the issue of knowledge incompleteness and lack of domain coverage in resume and job posting matching caused by the exploitation of domain-general resources. A variety of co-operative semantic/ontological resources will be used to filter out irrelevant resumes. A two-way (candidate to job and job to candidate) semantic-based automatic suitability ranking is proposed. The suitability is determined by the semantic distance of resumes and job postings, evaluated by their word embeddings. An efficient semantic space created through the Convolutional and Recurrent Neural Networks will be utilized with different word embedding mechanisms along with different classification methods. The project also investigates the potential of knowledge graphs in illustrating inconsistencies between the resumes and job postings. This study develops an automatic system capable of precisely detecting, extracting, and visualizing the resume and job posting’s relevant skills as well as the implicitly encoded semantic dimensions of applicant resumes.