Current Organizations / Institutes

The organizations/institutes from the scholar's latest article

Previous Organizations / Institutes

The organizations / institutes from the scholar's other articles in the past ten years


Wonseong Kim has a total of 6 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Open Access

    Evidence-Grounded Persona Synthesis: Constructing AI Personas from Four Classes of Heterogeneous Person Data

    by Wonseong Kim

    Large language models can turn scattered records about a person into fluent characterizations, but fluency and fidelity pull in opposite directions: the more readable the synthesis, the more easily unsupported claims enter it. I propose a framework in which the persona-a persistent, versioned intermediate object-rather than any generated text, is the unit of construction, verification, and consistency. The framework classifies person-describing data into four classes with distinct epistemic roles and error structures: performance measurements, third-party observations, self-reported dispositions, and population reference statistics. Because the classes fail differently, they enter synthesis asymmetrically: observations require incident-level anchoring; self-reports may suggest but never confirm; measurements acquire meaning only against versioned norms; and cross-class disagreement is preserved as a first-class contradiction record rather than averaged away. Every persona trait carries machine-checkable links to its evidence, every generated statement links to a trait, and independent verifier models excise statements that exceed their trait's scope. Update discipline-differential updates, version freezing, and holds on unsupported abrupt change-gives characterizations the inertia that descriptions of persons warrant. I instantiate the framework in a virtual survey: evidence-grounded synthetic respondents, constructed from the records of a fictitious household panel, answer a new questionnaire item by item, abstaining wherever their evidence provides no basis for an answer, so that simulated results carry an explicit map of what the underlying records can and cannot support.

  • Open Access

    Is Small Really Beautiful for Central Bank Communication? Evaluating Language Models for Finance: Llama-3-70B, GPT-4, FinBERT-FOMC, FinBERT, and VADER

    by Wonseong Kim, Jan Spörer, Choong Lyol Lee, Siegfried Handschuh

    This study compares the sentiment detection capabilities of language models for the domain of central bank communication, particularly the official statements released by the U.S. Federal Open Market Committee (FOMC). While previous studies have explored FOMC communication, this work is one of the few studies that use a natural language processing-based approach. The analysis employs VADER, FinBERT, a fine-tuned FinBERT model (FinBERT-FOMC), GPT-4, and Llama-3-70B. Within the scope of our labeled dataset on FOMC minutes, Llama 3 is the most accurate model, followed by GPT-4, FinBERT-FOMC, FinBERT, and VADER. The FinBERT-FOMC model, which was fine-tuned on central bank communication and utilizes a text simplification pipeline, performs better than the original FinBERT model. Llama 3 and GPT-4 outperform at the expense of large model sizes. Unlike GPT-4, FinBERT and FinBERT-FOMC are open-source and can be deployed on consumer-grade hardware. Llama 3 requires substantial hardware investments to deploy. The work thus finds that there is still a between model size and performance, and that the notion that “small is beautiful” can still hold for use cases where maximum accuracy is a lesser concern than inference speed and cost. Human performance is still significantly above all models, indicating that further improvements in language models and FOMC-specific prompting are possible. The labeled dataset for central bank communication we present in this paper is thus a challenging benchmark for future research. The dataset is freely available for download.1


  • Open Access

    Bounded Rationality in Central Bank Communication

    by Wonseong Kim, Choong Lyol Lee

    This study explores the influence of FOMC sentiment on market expectations, focusing on cognitive differences between experts and non-experts. Using sentiment analysis of FOMC minutes, we integrate these insights into a bounded rationality model to examine the impact on inflation expectations. Results show that experts form more conservative expectations, anticipating FOMC stabilization actions, while non-experts react more directly to inflation concerns. A lead-lag analysis indicates that institutions adjust faster, though the gap with individual investors narrows in the short term. These findings highlight the need for tailored communication strategies to better align public expectations with policy goals.

  • Open Access

    Words that Matter: The Impact of Negative Words on News Sentiment and Stock Market Index

    by Wonseong Kim

    This study investigates the impact of negative words on sentiment analysis and its effect on the South Korean stock market index, KOSPI200. The research analyzes a dataset of 45,723 South Korean daily economic news articles using Word2Vec, cosine similarity, and an expanded lexicon. The findings suggest that incorporating negative words significantly increases sentiment scores' negativity in news titles, which can affect the stock market index. The study reveals that an augmented sentiment lexicon (Sent1000), including the top 1,000 negative words with high cosine similarity to 'Crisis,' more effectively captures the impact of news sentiment on the stock market index than the original sentiment lexicon (Sent0). The results underscore the importance of considering negative nuances and context when analyzing news content and its potential impact on market dynamics and public opinion.


  • Open Access

    Analyzing FOMC Minutes: Accuracy and Constraints of Language Models

    by Wonseong Kim, Jan Frederic Spörer, Siegfried Handschuh

    This research article analyzes the language used in the official statements released by the Federal Open Market Committee (FOMC) after its scheduled meetings to gain insights into the impact of FOMC official statements on financial markets and economic forecasting. The study reveals that the FOMC is careful to avoid expressing emotion in their sentences and follows a set of templates to cover economic situations. The analysis employs advanced language modeling techniques such as VADER and FinBERT, and a trial test with GPT-4. The results show that FinBERT outperforms other techniques in predicting negative sentiment accurately. However, the study also highlights the challenges and limitations of using current NLP techniques to analyze FOMC texts and suggests the potential for enhancing language models and exploring alternative approaches.


  • Open Access

    The impact of African swine fever news sentiment on the Korean meat market

    by Byung Min Soon, Wonseong Kim

    Our study analyzed the impact of African swine fever (ASF) news on the Korean meat market using sentiment analysis. We applied a neural network language model (NNLM) to generate a sentiment index indicating whether the news had a positive or negative impact on consumer expectations. We analyzed 24,143 news articles to estimate the impulse responses of meat price variables to sentiment shocks. Our study contributes significantly to agricultural economics as it applies NNLM to generate a sentiment index. The empirical results indicated that ASF news sentiment has a substantial impact on meat prices in Korea, and there is evidence of substitution effects among different types of meat. ASF news has a positive impact on the price of pork, negative effects on beef and chicken prices, and a greater impact on the price of chicken than beef. The findings imply that the effect of ASF news on demand outweighs its impact on supply in the pork market, whereas the effect on supply surpasses the effect on demand in the beef and chicken market. We believe our methods and results will inspire discussions among applied economists studying consumer behavior in this specific market and could encourage the application of big data analysis to the agricultural economy.


  • Open Access

    Advancing Agricultural Predictions: A Deep Learning Approach to Estimating Bulb Weight Using Neural Prophet Model

    by Wonseong Kim, Byung Min Soon

    A deep learning methodology was utilized to predict the bulb weights of garlic and onions in the Jeolla Province of Korea. The Korea Rural Economic Institute (KREI) operates the Outlook & Agricultural Statistics Information System (OASIS) platform, which provides actual measurements of garlic and onions. We trained the Neural Prophet (NP) lagged time-series model using this data. The NP model effectively handles lagged variables and their covariates by inserting a hidden layer. Our results indicate that the NP model performed with around 5% mean absolute error in predicting bulb weights, with a gap of 3.3 g and 4.7 g with average weights of 63.7 g and 129.9 g for garlic and onions, respectively. This experimental research was based on only three years of measurement data. Hence, the gap between observed and predicted data can be reduced by accumulating more measurement data in the future.