Sentiment and Position Taking of Witness Testimonies

Dr. Mark Lokanan received a New Frontiers in Research Fund – Rapid Response 2021 Grant to study witness sentiments in the Cullen Commission inquiring into money laundering activities in BC.

Objectives
In 2019, the British Columbia’s (B.C.) government appointed Austin Cullen, a B.C. Supreme Court justice, to conduct an independent public inquiry into money laundering in the province’s real estate, luxury vehicle, and gaming sectors. After listening to the testimonies of 200 witnesses, the inquiry will wrap up in December of 2021. The objectives of the proposed research are twofold: (1) to understand the sentiments of the witnesses towards anti-money laundering practices in B.C. real estate, luxury vehicle, and gaming sectors; and (2) to analyze the polarity and factuality of Twitter data regarding the witness testimonies. Given the rapidly changing COVID-19 pandemic, this research is time-sensitive, and I will endeavour to pursue the project from January–December, 2022.

Research Approach
Initially, the plan was to conduct in situ discussions with key personnel from the real estate industry and field research in the casinos and luxury car dealerships in B.C. However, COVID-19 put an abrupt end to the planned research activities. To substitute field research, I plan to employ Deep Learning Methodology (DLM) to conduct emotional and sentiment analyses of the witnesses' testimonies in the Cullen Commission's ("Commission") inquiry using natural language processing (NLP). NLP techniques will be used to build a Convolutional Neural Network classification model to analyze discrete emotions of the witnesses’ testimonies to the Commission. To examine the polarity and factuality of the Twitter data, I plan to use machine learning algorithms to explore the sentiments the public shares towards the witnesses' testimonies.

Novelty and Significance
DLM is a machine learning approach that performs classification tasks directly from images, sounds, and texts. DLM success relies heavily on computational resources to process large amounts of data. As the input data increases, the amount of computational resources (time, memory, storage space, and costs) to clean, process, and analyze the unstructured data increases. The increase in model complexity amplifies the risks in DLM and can lead to model failure owing to flawed methodology and the statistical methods used. On the other hand, DLM can exploit large volumes of multimodal data and, if successful, can lead to high rewards for qualitative researchers. Complementary DLM and subject expertise can advance current research methods and enhance the usefulness of NLP in Qualitative research.