Semantic Data Management - Knowledge Graphs

This repository contains the implementation of a Knowledge Graph for academic publication data from Tsinghua University's computer science field (2018-2023), sourced from the Scopus database. The project demonstrates the construction and querying of a semantic data model using RDF/RDFS technologies.

Project Overview

The project builds a comprehensive knowledge graph that models the academic publication ecosystem, including authors, papers, conferences, journals, reviews, and their interconnected relationships. The implementation follows semantic web standards using RDF Schema (RDFS) for ontology definition and RDFLib for graph construction and querying.

Repository Structure

Core Files

• 12E_Isakova_Mahmudova_Report.pdf - Complete project report with methodology and results
• tbox.ttl - Terminological Box (TBOX) defining the ontological schema
• abox.ttl - Assertional Box (ABOX) containing instance data
• B3_TBOX_ABOX_Connection.ttl - Final combined ontology with inferred relationships

Implementation Notebooks

• B1_TBOXCreation.ipynb - TBOX creation with class and property definitions
• B2_ABOXCreation.ipynb - ABOX population from CSV data
• preprocessing-sdm-lab2.ipynb - Data preprocessing and cleaning pipeline

Data & Results

• data/ - Directory containing processed CSV files with academic publication data
• queries_result/ - Directory with SPARQL query results and outputs
• .ipynb_checkpoints/ - Jupyter notebook checkpoints
• .DS_Store - System file (macOS)

Query Files

• B3_statisticsQueries.txt - Statistical queries for ontology analysis
• B4_Queries.txt - Complex SPARQL queries for data retrieval

Visualizations

TBOX Schema

Class Hierarchy

Class Relationships

• B1_TBOXCreation_VisualRepresentation.png - Visual schema representation
• B1_TBOX_GraphDB_ClassHierarchy.png - Class hierarchy visualization
• B1_TBOX_GraphDB_ClassRelationships.png - Class relationship diagram

Key Features

Ontological Model

• Classes: Person, Author, Paper, Publication, Conference, Journal, Topic, Review, etc.
• Properties: writes, publishedAs, reviewedBy, relatedTo, isIn, isOf
• Hierarchical Structure: Supports inheritance and complex relationships

Data Statistics

• Total Statements: 57,785,594 (28,766,866 explicit + 29,018,728 inferred)
• Authors: 105,871 instances
• Papers: 500 instances
• Publications: 20,729 instances
• Topics: 487 instances

Query Capabilities

• Author publication analysis
• Conference and journal statistics
• Topic-based paper classification
• Citation and review networks
• Publication venue analysis

Technical Implementation

• Language: Python with RDFLib
• Schema: RDF Schema (RDFS) with RDFS-Plus entailment
• Query Language: SPARQL
• Inference: Automatic relationship derivation through RDFS reasoning

Usage

1. Data Preprocessing: Run preprocessing-sdm-lab2.ipynb to clean and prepare raw data
2. TBOX Creation: Execute B1_TBOXCreation.ipynb to define the ontological schema
3. ABOX Population: Run B2_ABOXCreation.ipynb to populate the graph with instances
4. Query Execution: Use the provided SPARQL queries in B4_Queries.txt for data analysis

Sample Queries

• Find all papers by a specific author in database conferences
• Calculate average paper length by journal
• Identify most prolific conferences by publication count
• Retrieve properties by domain (Author, Conference, Journal)

Authors

• Dilbar Isakova - dilbar.isakova@student-cs.fr
• Narmina Mahmudova - narmina.mahmudova@estudiantat.upc.edu

Course Information

Course: Semantic Data Management
Institution: Facultat d'Informàtica de Barcelona, UPC
Instructors: Anna Queralt, Oscar Romero
Date: May 29, 2024

License

This project is for academic purposes as part of the Semantic Data Management course curriculum.