Contains code for Yolo2 and Yolo3 used on text data to detect RDF triples in a sentence.
Project Summary:
• This project aims to design an embedding scheme for knowledge graphs using neural networks techniques.
• Used Word2Vec, ELMo and GloVe word embedding to convert text into 3D matrix and gave it as input to a modified YOLO algorithm to predict RDF triples in the sentence.
• The training set for the above algorithm was made by using OpenIE annotator of the Stanford CoreNLP software.
• Made a knowledge graph from predicted RDF triples to make a question-answering system from it.
-
Dataset : Contains dataset used for training.
-
Python scripts : Important python scripts used for working with dataset.
-
Yolo2-text : Yolo2 for text data.
-
Yolo3-text : Yolo3 for text data.