hitesh96db / Community-Detection

Community Detection from Research Articles

The task is to detect research papers which belong to a common field of research.

The project contains following folders:

• aan :

  • This folder consists of the AAN (ACL Anthology Network) dataset. We worked on the 2013 release of the dataset.

• aan_small :

  • This folder consists of a subset of the AAN dataset. It contains approximately 2000 nodes.

• algorithms:

  • This folder contains the code for the different algorithms used.
    • Cosine-kmeans
    • Jaccard-kmeans
    • Louvain
    • Newman-girvan
    • Newman-girvan-v2 (not used to generate the outputs)

• metrics:

  • This folder contains code for different metrics used to construct the network.
    • authorCitation
      • This metric uses the author citation network to detect communities. We have used louvain and newman-girvan algorithms for this.
      • How to run: Refer Community-Detection/metrics/authorCitation/README.md
    • paperCitaion
      • This metric uses the paper citation network to detect communities. We have used louvain and newman-girvan algorithms for this.
      • How to run: Refer Community-Detection/metrics/paperCitation/README.md
    • title
      • This folder contains the code for running K-means algorithm (using Jaccard and Cosine) on the title metric.
      • How to run: Refer Community-Detection/metrics/title/README.md
    • year
      • This folder contains the code for running K-means algorithm (using Jaccard and Cosine) on the year metric.
      • How to run: Refer Community-Detection/metrics/year/README.md

• outputs

  • This folder contains the outputs for various algorithms run on different metrics along with their respective graphs.

• CommunityDetectionReport.pdf: Project report