MAS a scalable framework for research effort evaluation by unsupervised machine learning-Hybrid plagiarism model

In the era of web new information is upcoming day by day. Researches add their work for their research domains. Detecting of originality of research work is in hype. In Academic sector students researchers bring in innovative ideas, algorithms stating that their work outperforms prior research. They may implement NULL Hypothesis or alternative Hypothesis, detecting their effort is a challenge. By means of plagiarism detectors such academic efforts can be evaluated or graded. This reflects the essence of research in the field of Plagiarized content detection and grading. Some of our research issue highlights to technical scenario to design an algorithm which is adaptable to changing nature of dataset. The dataset grows, as new research work is added in due course of time. Data extraction from unstructured information is challenging, as no standard pattern is yet defined. Such patterns vary from research to research and are domain specific. A document in question i.e plagiarized or not? Is a join of one or more sentences that originate by the authors research or referenced from previous publications. Authors to prove originality use paraphrasing which may have semantic similarity, also some of the contents act as metaphor for upcoming research work. It is complex task point out such an activity. Methodology states that a document in question is a join of sentences, whereas each sentence is a join of terms. Thus we conclude by fork and join operations; plagiarism detection is possible in effective way. Document in question is split to produce a sentence vector. A term vector is generated by forking sentence to terms for each sentence in sentence vector. Mapper is implemented that maps term to sentence and sentence to source document. To enhance the accuracy of the model a Multi Agent Based System MAS frame is recommended to adapt varying similarity functions. Achieve parallelism in system and adaptability of new similarity measures as well remove one which are not suitable any more to the task.