Dr. Vardan Baghdasaryan
, , ,

Afeyan Research Grant Recipient: Dr. Vardan Baghdasaryan

4 min read

YEREVAN, Armenia — The American University of Armenia (AUA) is pleased to announce the three recipients of a one-year research grant, which was made possible thanks to a generous donation by the Afeyan Family Foundation. The University is grateful to the Afeyan Family Foundation for supporting these research activities, selected to bring positive impact to Armenian research and development.  

The supported projects are in three disciplines that have been deemed priorities for the development of Armenia, namely economics, law, and medicine. AUA will continue to support faculty research through internal seed funding for the continuation of these projects beyond the first year, as well as to promote similar research among other faculty members. 

The research awards were announced on July 18, 2023, with immediate start dates. They are expected to be conducted during the academic year 2023-24, with each principal investigator delivering a mid-year status report as well as a final report upon completion of the project by August 30, 2024.

In a series of three articles, beginning with this one, we will profile each of the principal investigators’ projects. In this first article, we will focus on the research Dr. Vardan Baghdasaryan is conducting with the grant he has received from the Afeyan Family Foundation.

An associate professor in the Manoogian Simone College of Business and Economics, Dr. Baghdasaryan received a grant to lead a research project titled “Taxpayer Clustering and Fraud Identification with Non-Structured Transaction Level Data Using Natural Language Processing.”

Dr. Baghdasaryan’s team aims to develop an alternative approach for the identification of tax fraud by combining natural language processing (NLP) and unsupervised learning approaches for taxpayers clustering. The core of the approach is the application of NLP to non-structured transaction-level big data of the taxpayers with the purpose of understanding the “economic” and “business” proximity based on actual operations rather than the announced sector of operations. The ultimate goal is to demonstrate that recent advances in NLP and ability to analyze textual data can significantly improve the understanding of the taxpayers’ operations and business nature and contribute to the task of tax fraud identification.

Dr. Baghdasaryan was interviewed by AUA Communications Specialist Serena Hajjar Bakunts to obtain a first-hand account of his project:

  • What inspired your research project?

This is not our first experience of applying machine learning and AI tools to tax data, but we have always been interested in unlocking the potential hidden within unstructured textual data found in invoices and tax receipts. Currently, the State Revenue Committee (SRC) has limited means of analyzing that data, as it is hindered by noisy and mixed formats. Our goal is to develop effective algorithms that transform this unstructured data into a usable and coherent format, ultimately facilitating product-level analysis and unlocking insights that were previously inaccessible.

  • Where does your research process currently stand?

We are working with a specific subset of unstructured data buyer invoices. Our objective is to understand what companies acquire to produce their output and cluster them on that basis. Later, this clustering information will be used in conjunction with other data to uncover unusual reporting patterns. We have just finalized the preparatory stage of the project, which included acquisition and preliminary cleaning of the data and exploratory data analysis. We already see some patterns in terms of goods and services used and to what extent classifying this information will make it easier to understand the business processes of the taxpayers. 

  • Are students engaged in your project? If so, how many and in what capacity?

The core team of data scientists on our project includes two alumni of the Master of Science in Management (MSM) program: Zaruhi Navasardyan (MSM ’20) and Arsine Sarikyan (MSM ’20), who is also an adjunct lecturer for the “Advanced Topics in Data Analysis” course in the MSM program. Our project team is mostly comprised of women, which is rare for a research project in STEM. Our collaboration has endured throughout the last few years, and we have already published a couple of our interesting outcomes in peer-reviewed journals. On top of that, one of our Industrial Engineering and Systems Management program alumni is working with us as a research assistant. Ani Saribekyan (MEIESM ’22) is employed by the SRC, and she helps us with obtaining correct data in a timely manner. 

  • How do you see your research benefitting Armenia’s development?

Unveiling tax evasion is becoming increasingly difficult, because there are no more low-hanging fruit as was the case years ago. Additionally, businesses and technologies are becoming more sophisticated. In this regard, applying machine learning (ML) and artificial intelligence (AI) is a must. Indeed, a recent study by the Organisation for Economic Co-operation and Development shows that more than two-thirds of tax administrations around the globe are actively using these tools to improve tax collection and decrease tax evasion. We believe we are not only conducting a research project, but also contributing to the capacity of the SRC to use these ML/AI tools in their daily operations. This is extremely motivating for us. 

Founded in 1991, the American University of Armenia (AUA) is a private, independent university located in Yerevan, Armenia, affiliated with the University of California, and accredited by the WASC Senior College and University Commission in the United States. AUA provides local and international students with Western-style education through top-quality undergraduate and graduate degree and certificate programs, promotes research and innovation, encourages civic engagement and community service, and fosters democratic values.