PhD Student in Computer Science · UESTC, Chengdu, China
Event Extraction · Natural Language Processing · Machine Learning

I am a PhD student in Computer Science at the University of Electronic Science and Technology of China (UESTC), Chengdu, with a background in applied data science and natural language processing.

My research centres on Event Extraction - detecting event triggers and classifying event types in unstructured text - and on information extraction more broadly. I work with transformer architectures (BERT and variants) across both high-resource and low-resource languages, including original corpus construction for Pashto.

Alongside research, I build end-to-end machine learning systems: data acquisition and scraping, preprocessing pipelines, model training and evaluation, and deployment behind production APIs.

• Research interests: Event Extraction, Information Extraction, NLP, Low-Resource Language Processing
• Technical interests: Transformer architectures, distributed data processing, production ML deployment
• Languages: English (fluent), Pashto, Urdu

Programming

Machine Learning & Deep Learning

NLP & LLMs

Named Entity Recognition · Event & Trigger Detection · Sentiment Analysis · Text Classification

Big Data

PySpark · Spark MLlib · HDFS

Web & Backend

Cloud & DevOps

• Researching event extraction and trigger detection with transformer models as part of my PhD at UESTC
• Building industrial text intelligence systems - BERT-based NER over maintenance and operations logs
• Extending NLP methods to low-resource languages, particularly Pashto
• Strengthening production ML practice: containerised deployment, testing, and low-latency inference

BERT-based Named Entity Recognition for industrial maintenance logs, extracting Faults, Components, Actions, and Equipment from unstructured text. Ships with a bilingual (English / Chinese) Streamlit interface for single and batch analysis, a rule-based spaCy bootstrapping pipeline for weak labelling, a full BERT training pipeline, and Docker deployment.

Python PyTorch Transformers spaCy Streamlit Docker

Submitted to the Shijiazhuang Innovation & Entrepreneurship Competition (International Track).

Event trigger detection and event-type classification using BERT on the RAMS dataset (Johns Hopkins). Core to my doctoral research on information extraction.

Python BERT Transformers RAMS

Delivery-time prediction and late-delivery classification built on PySpark ML. Covers the full lifecycle across four sequential notebooks - data preparation, training, evaluation, and inference - and exports a dependency-free scoring function served through a Flask API, avoiding per-request Spark JVM overhead for real-time use.

PySpark Spark MLlib Flask Gunicorn Hadoop HDFS

Sentiment analysis for Pashto, a low-resource language. Includes an original corpus collected from social media and annotated by two native speakers across politics, sports, and entertainment, benchmarked with SVM, Naive Bayes, Decision Tree, Random Forest, and AdaBoost.

Python scikit-learn Django Kaggle Dataset

Interests

Event Extraction · Information Extraction · Named Entity Recognition · Low-Resource Language NLP · Applied Transformer Models

Datasets & Benchmarks

• RAMS (Roles Across Multiple Sentences) - event trigger detection and event-type classification
• Pashto sentiment corpus - original, double-annotated, published on Kaggle
• News AI labor coverage corpus - 22,000+ articles, 2014-2025

Publications

Google Scholar: Farhad Khan

Full publication list in preparation.

Open to research collaborations in NLP and information extraction.