Nikhil Agrawal

Indian Institute of Tecnology, Kharagpur

CGPA: 8.93

M.Tech Thesis:- Improving aspect based ranking in clinical trials

-Information Retrieval, Natural Language Processing
Clinical Trials are crucial for the practice of evidence-based medicine. It provides updated and important health-related information for the patients. Different stakeholders like trial volunteers, trial investigators, and meta-analyses researchers often need to search for trials having similar eligibility features. Clinical trials are the first source of information available to users in terms of new drugs and new treatments for the disease. Apart from clinical trials, there are also other sources of information eg: PubMed, adverse events that can be leveraged to provide ranked results of clinical trials. In this research work, we develop an automated method that can be applied across all classes of disease to retrieve relevant trials on the basis of UMLS concepts overlapped between the disease information provided by the user as a query and clinical trials. We propose and compare the different relevancy based approaches and found out that the inclusion of synonyms words provided better results in terms of precision value. We also ranked the retrieved clinical trials on different aspects like relevancy, adversity, recency, and popularity. Also, we found out that there’s a high negative Spearman rank’s correlation coefficient existing between popularity and recency which is expected as the paper published earlier has high chances of more citations. Also at the end, we provided the possible limitations of the methods.

Finding the influential users in the Twitter network

-Complex Network
The goal of the project was to use temporal pattern of retweets combined with structural information of the network to identify the best set of influential users that can be targeted for viral diffusion in the Twitter network.

Automatic Concept Map Generation from Text-based Learning Material

-Natural Language Processing, Information Retrieval
Generated a Concept Map from a document by first converting text to simple language, identify important entities, finding the weighted relationship between entities, and then finally obtain a visual representation of the relations between entities.

Towards an Aspect-based Ranking Model for Clinical Trial Search

Clinical Trials are crucial for the practice of evidence-based medicine. It provides updated and important health-related information for the patients. Different stakeholders like trial volunteers, trial investigators, and meta-analyses researchers often need to search for trials having similar eligibility features. Clinical trials are the first source of information available to users in terms of new drugs and new treatments for the disease. Apart from clinical trials, there are also other sources of information eg: PubMed, adverse events that can be leveraged to provide ranked results of clinical trials. In this research work, we develop an automated method that can be applied across all classes of disease to retrieve relevant trials on the basis of UMLS concepts overlapped between the disease information provided by the user as a query and clinical trials. We propose and compare the different relevancy based approaches and found out that the inclusion of synonyms words provided better results in terms of precision value. We also ranked the retrieved clinical trials on different aspects like relevancy, adversity, recency, and popularity. Also, we found out that there’s a high negative Spearman rank’s correlation coefficient existing between popularity and recency which is expected as the paper published earlier has high chances of more citations. Also at the end, we provided the possible limitations of the methods.