Farrukh Ahmed

Academic

Career

Verbatim text

Biomedical Text Annotation Using Knowledge Graphs 
This project will create a web-based semantic biomedical text annotator using NLP (Natural 
Language Processing) and knowledge graphs/semantic web ontologies. This annotator focuses on 
two domains of the biomedical literature: diseases and genes. To implement our biomedical 
semantic text annotator, we needed to perform the following three major steps: (1) Named Entity 
Recognition (NER) to extract the entities of types genes and diseases, (2) Entity Linking (EL) to link 
the extracted entities to relevant entities from the selected ontologies (3) Relation Extraction (RE) 
to extract relations between genes and disease entities. For NER and RE, we used BioBERT which 
gave good accuracies greater than 90%. For relation extraction, we used Personalised Pagerank 
(PPR) algorithm supplemented with Gene-Disease relations extracted from text. The user interface 
is developed using python-based django web framework. The interface consists of only a single 
view.  Using the interface, the user inputs a text. The application will then display the text with the 
extracted entities highlighted by their entity types. It will also display a document graph showing 
entities as nodes and links between the entities. The application also displays a brief description of 
the highlighted entity, when clicked.  
 
 
 
 
 
 
 
 
 
 
 
Technology Used: 
Python, PyTorch, NetworkX, Owlready2,  
Transformer, Django Web Framework 
Supervisor Name: 
Dr. Amna Basharat 
Group Members:   
Ahmad Wali Bin Saeed (i17 - 0106) 
Ahmad Ali Bin Saeed (i17 - 0105)                    
Farrukh Ahmed (i17 - 0100)

AI enrichment