• Eilon Tsadok
• Zvi Mints
  Managed by Or Anidjar and Amit Dvir

In this project, we will explore how to benefit the knowledge that can be extracted from a Knowledge Graph, in order to detect a community of pedophilles with some charachteristics. As such, we will represent in this Knowledge Graph an IM dataset containing users that served as pedophilles impersonators and their textual content shared over the Internet. For the task of*"understanding"*the content, we will take advantage of the recently based language modelling developed by Google corpora, which is called BERT (Bidirectional Encoder Representations from Transformers). After the knowledge representation of the data set in the embedded knowledge graph, we will apply a spectral clustering algorithm in order to detect communities of users that are sharing close enough behavioral patterns.

• Dataset Source
  • https://pan.webis.de/clef12/pan12-web/sexual-predator-identification.html
  • .json format training dataset: https://f2h.io/8b3xzuy6rnq4
  • .json format test dataset: https://f2h.io/5n7mcc7cx3b6
• Knowledge Graph Vertices Embedding
  • https://github.com/jimmywangheng/knowledge_representation_pytorch
  • https://radimrehurek.com/gensim/models/word2vec.html
• PCA (Principal Component Analysis)
  • https://en.wikipedia.org/wiki/Principal_component_analysis
  • https://scikit-learn.org/stable/modules/generated/sklearn.decomposition.PCA.html
• Spectral Clustering
  • https://en.wikipedia.org/wiki/Spectral_clustering
  • https://scikitlearn.org/stable/modules/generated/sklearn.cluster.SpectralClustering.html
  • https://en.wikipedia.org/wiki/Elbow_method_(clustering)
• Topic Extraction
  • https://en.wikipedia.org/wiki/Topic_model
  • https://towardsdatascience.com/the-complete-guide-for-topicsextraction-in-python-a6aaa6cedbbc
• Google BERT Model Utilization in Python
  • https://towardsdatascience.com/word-embedding-using-bert-inpython-dd5a86c003423

• Start Date: 24/03/2020
• End Date: Start Date + 12 week