Taqi Haider

Novosibirsk State University

I am an AI researcher and a graduate of the Master’s program in Artificial Intelligence and Big Data Analytics at Novosibirsk State University. I work on extracting and comparing evidence from scientific literature and on practical machine learning tools. I am seeking PhD opportunities in NLP and related areas of AI.

My master’s research led to MedSETNER, a benchmark built from 691 medical research articles with 9,376 annotated dataset-name mentions. In ClaimGraph-BT, I trained context-aware claim detectors for brain tumor AI literature; PubMedBERT Context reached 0.8391 macro F1 on manually verified test data. I also build applied NLP systems, including multilingual question answering and customer support tools.

Explore my projects · Read about my experience · Download my CV

Research updates

MedSETNER and ClaimGraph-BT were accepted at AIST 2026, CCIS track.
BrainTumorClaims, a dataset for structured claim extraction, is listed among my scholarly outputs.
My research interests include NLP, multimodal language models, and interpretability and controllability of large language models.

Selected publications

MedSETNER: A Benchmark Corpus for Extracting Dataset Names from Medical Scientific Literature

Taqi Haider, Nawazish Hussain, Syed Muhammad Irtiza Zaidi, and Evgeniy Nikolaevich Pavlovskiy

2026

Accepted at the 13th International Conference on Analysis of Images, Social Networks and Texts (AIST 2026), CCIS track

Abs Bib

A benchmark for medical dataset-name extraction built from 691 research articles and 9,376 annotated mentions. BiomedBERT reached 60.7 percent mean exact-span F1 on same-pool data and 60.8 percent on an out-of-distribution test set.
@unpublished{haider2026medsetner, title = {MedSETNER: A Benchmark Corpus for Extracting Dataset Names from Medical Scientific Literature}, author = {Haider, Taqi and Hussain, Nawazish and Zaidi, Syed Muhammad Irtiza and Pavlovskiy, Evgeniy Nikolaevich}, year = {2026}, note = {Accepted at the 13th International Conference on Analysis of Images, Social Networks and Texts (AIST 2026), CCIS track}, }
ClaimGraph-BT: Automated Claim Extraction and Cross-Paper Comparison for Brain Tumor AI Literature

Nawazish Hussain, Taqi Haider, and Evgeniy Nikolaevich Pavlovskiy

2026

Accepted at AIST 2026, CCIS track

Abs Bib

Context-enhanced biomedical claim detection reached 0.8391 macro F1 and 0.9301 AUC on manually verified test data. Deterministic repair and grounding improved macro field-presence F1 from 0.1668 to 0.4850.
@unpublished{hussain2026claimgraph, title = {ClaimGraph-BT: Automated Claim Extraction and Cross-Paper Comparison for Brain Tumor AI Literature}, author = {Hussain, Nawazish and Haider, Taqi and Pavlovskiy, Evgeniy Nikolaevich}, year = {2026}, note = {Accepted at AIST 2026, CCIS track}, }

BrainTumorClaims: A Dataset for Scientific Claim Detection and Structured Information Extraction in Brain Tumor Literature

Nawazish Hussain and Taqi Haider

2026

DOI Bib

@misc{hussain2026braintumorclaims,
  title = {BrainTumorClaims: A Dataset for Scientific Claim Detection and Structured Information Extraction in Brain Tumor Literature},
  author = {Hussain, Nawazish and Haider, Taqi},
  year = {2026},
  howpublished = {Dataset, Novosibirsk State University},
  doi = {10.5281/zenodo.20364566},
}