Keyword extraction is an automated process that collects a set of terms, illustrating an overview of the document. The term is defined how the keyword identifies the core information of a particular document. Analyzing huge number of documents to find out the relevant information, keyword extraction will be the key approach. This approach will help us to understand the depth of it even before we read it. In this paper, we have given an overview of different approaches and algorithms that have been used in keyword extraction technique and compare them to find out the better approach to work in the future. We have studied various algorithms like support vector machine (SVM), conditional random fields (CRF), NP-chunk, n-grams, multiple linear regression, and logistic regression to find out important keywords in a document. We have figured out that SVM and CRF give better results where CRF accuracy is greater than SVM based on F1 score (The balance between precision and recall). According to precision, SVM shows a better result than CRF. But, in case of the recall, logit shows the greater result. Also, we have found out that, there are two more approaches that have been used in keyword extraction technique. One is statistical approach and another is machine learning approach. Statistical approaches show good result with statistical data. Machine learning approaches provide better result than the statistical approaches using training data. Some specimens of statistical approaches are Expectation-Maximization, K-Nearest Neighbor and Bayesian. Extractor and GenEx are the example of machine learning approaches in keyword extraction fields. Apart from these two approaches, semantic relation between words is another key feature in keyword extraction techniques.