Bolun (Namir) Xia

Bolun (Namir) Xia

Auto-generated profile, based on publicly available information

Bolun (Namir) Xia

Computer Science, Rensselaer Polytechnic Institute

Research Focus & Metrics

Academic Career Length

3Years (2024-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
22 November 2024
Verified
Conference Paper
Open access
LYLing Yue
Ling Yue
SXSixue Xing
Sixue Xing
JLJonathan Li
Jonathan Li
ZIZabirul Islam
Zabirul Islam
BXBolun Xia
Bolun Xia
TFTianfan Fu
Tianfan Fu
The clinical trial process is crucial for ensuring the safety and efficacy of new pharmaceuticals but is often hindered by lengthy durations, high costs, and labor-intensive processes. These challenges delay the availability of life-saving treatments and increase financial burdens on pharmaceutical companies, with average costs exceeding 2 billion dollars per trial. Efficient management of trial timelines is essential to control budgets and improve the economic feasibility of research. To address these challenges, we introduce TrialDura, an innovative machine learning-based framework designed to predict the duration of clinical trials. TrialDura utilizes a multimodal approach, incorporating diverse data sources such as disease names, drug molecules, trial phases, and detailed eligibility criteria. These elements are encoded using Bio-BERT embeddings, which are specifically tuned for biomedical contexts to provide a comprehensive semantic understanding of the trial data. The model employs a hierarchical attention mechanism that effectively integrates these embeddings, capturing complex interactions between different data modalities to enhance prediction accuracy. Our extensive experiments demonstrate that TrialDura significantly outperforms existing models, achieving a mean absolute error (MAE) of 1.04 years and a root mean square error (RMSE) of 1.39 years. This performance underscores the model's ability to deliver more precise predictions of trial durations, which can facilitate better planning and resource allocation in clinical research. By providing insights into expected trial timelines, TrialDura aids in optimizing trial management strategies, potentially reducing costs and accelerating the development of new treatments. Furthermore, TrialDura's interpretability, enabled by its hierarchical attention mechanism, offers valuable insights into the factors influencing trial durations, assisting clinicians and researchers in making informed decisions. The model's ability to highlight key features contributing to duration predictions enhances its utility as a decision-support tool in the pharmaceutical industry. In addition to its predictive capabilities, TrialDura marks a significant advancement in applying artificial intelligence to clinical trials. By leveraging state-of-the-art natural language processing techniques and integrating them with structured clinical data, TrialDura not only enhances the accuracy of duration predictions but also deepens the understanding of factors driving trial timelines. This comprehensive approach allows for a more nuanced analysis of clinical trial data, providing stakeholders with actionable insights to inform strategic decision-making and improve the overall efficiency of the drug development process. The implications of TrialDura extend beyond individual trial predictions, offering a framework that can be adapted and scaled to address various challenges in clinical research. As the pharmaceutical industry continues to evolve, the ability to accurately predict and manage trial durations will become increasingly important, making tools like TrialDura indispensable for researchers and decision-makers. By fostering a deeper understanding of trial dynamics and enabling more effective resource allocation, TrialDura has the potential to transform the landscape of clinical trials, ultimately leading to faster and more cost-effective development of new therapies. The code for TrialDura is publicly available at https://github.com/LeoYML/TrialDura.

19 June 2024 in ACM Transactions on Knowledge Discovery From Data
Verified
Journal Article
Open access
BXBolun (Namir) Xia
Bolun (Namir) Xia
VRVipula Rawte
Vipula Rawte
AGAparna Gupta
Aparna Gupta
MZMohammed Zaki
Mohammed Zaki
In the financial sphere, there is a wealth of accumulated unstructured financial data, such as the textual disclosure documents that companies submit on a regular basis to regulatory agencies, such as the Securities and Exchange Commission. These documents are typically very long and tend to contain valuable soft information about a company’s performance that is not present in quantitative predictors. It is therefore of great interest to learn predictive models from these long textual documents, especially for forecasting numerical key performance indicators. In recent years, there has been great progress in natural language processing via pre-trained language models (LMs) learned from large corpora of textual data. This prompts the important question of whether they can be used effectively to produce representations for long documents, as well as how we can evaluate the effectiveness of representations produced by various LMs. Our work focuses on answering this critical question, namely, the evaluation of the efficacy of various LMs in extracting useful soft information from long textual documents for prediction tasks. In this article, we propose and implement a deep learning evaluation framework that utilizes a sequential chunking approach combined with an attention mechanism. We perform an extensive set of experiments on a collection of 10-K reports submitted annually by U.S. banks, and another dataset of reports submitted by U.S. companies, to investigate thoroughly the performance of different types of language models. Overall, our framework using LMs outperforms strong baseline methods for textual modeling as well as for numerical regression. Our work provides better insights into how utilizing pre-trained domain-specific and fine-tuned long-input LMs for representing long documents can improve the quality of representation of textual data and, therefore, help in improving predictive analyses.

10 June 2026 in ACM Transactions on Management Information Systems
Verified
Journal Article
Open access
BXBolun (Namir) Xia
Bolun (Namir) Xia
AGAparna Gupta
Aparna Gupta
MZMohammed J Zaki
Mohammed J Zaki
Language models (LMs) have shown great promise in generative and predictive tasks. However, they do not explicitly incorporate semantic relations, and despite the progress in increasing the context size, they still struggle with long documents. Since abstract meaning representation (AMR), which is a graph-based representation of text to preserve its semantic relations, can encode semantic relationships at a deeper level, it can be beneficially utilized by graph neural networks (GNNs) for constructing effective document-level graph representations built upon LM embeddings for predictive tasks. We propose FLAG, an AMR-based framework to generate document-level embeddings via GNNs for long document classification tasks. We construct document-level graphs from sentence-level AMR graphs, endow them with finance-specific LM word embeddings, apply a GNN-based deep learning mechanism, and examine the efficacy of our AMR-based approach in predicting trends from financial documents. Extensive experiments on several different tasks are conducted on two large datasets of quarterly earnings call transcripts. We find that FLAG outperforms fine-tuning LMs directly on text in predicting stock price movement trends, as well as previous work utilizing document graphs and GNNs for text classification. Finally, we demonstrate our AMR-graph-based approach’s potential for explainability via a case study.