Advanced Infrastructure Projects Data Analysis with Entity Relationship Mapping

Electric Power Consumption (kWh per capita)

Dataset: Electric power consumption (kWh per capita)

Workflows

1. Extracting the Entities from the Dataset using tabular data.
2. Extracting the Entities from the Metadata using Language Model Babelscape/rebel-large
3. The code walkthrough is in the research folder.
4. The extracted entities and the defined SVO triplets are in the Entities Library.
5. Populating the Neo4j graph database with the extracted svo triplets.
6. Embeddings are not added because some of the extracted entities are not in the Wikidata or DBpedia Open Knowledge base.

Note: Spacy is used for the NER of the extracted entities. However given more resources this could be done using custom finetuned NER model.

Handy Cypher Commands

CALL apoc.import.json("https://raw.githubusercontent.com/ideepankarsharma2003/KnowledgeGraphs/main/json_files/svo_new_cat_ear_headphones_deduped.json") ;


MATCH (n:Node) 
CALL apoc.create.addLabels(n, n.node_labels) 
YIELD node 
RETURN node;

MATCH (n) RETURN n;

MATCH (n) DETACH DELETE n;

Graph Demo on Neo4j

- Labelled entities

- Extracted Relationships

- In-Depth Relationships

Directory Guide

• Code Walkthrough
• Entity Extraction using Language Model
• Final Generated Entities Guide
• Labelling the SVO triples

Data Sources

● World Bank Projects
● SAM.gov tenders
● Multi-modal data (images, videos, and textual descriptions)