Research

My primary research interest lies in the creativity of Large Language Models (LLMs). I explore how LLMs can generate creative and attractive text across various domains, including ad text generation, summarization, and data-to-text applications.

I also work on understanding what makes text attractive or appealing, particularly in advertising contexts, as well as computational approaches to humor understanding and generation.

Research Areas

Natural Language Generation

Text generation with LLMs: ads, summaries, and data-to-text

Computational Advertising

Analysis and generation of advertising texts

Humor Understanding

Computational understanding and generation of humor

Projects

Ad Text Generation

Research on automatically generating attractive ad texts using LLMs. Extracting and utilizing effective appeal expressions from landing page information.

Attractive Expression Analysis

Building and analyzing datasets of ad text paraphrase pairs to study the characteristics of attractive expressions.

Humor Understanding

Benchmark construction for humor understanding through Oogiri tasks and analysis of user preferences.

Data-to-Text Generation

Generating natural language text from numerical or structured data, such as weather forecasts and stock market commentary.

Research Keywords

Natural Language GenerationLarge Language ModelsComputational AdvertisingAd Text GenerationData-to-TextParaphrase GenerationText EvaluationHumor Understanding

Collaboration

At CyberAgent AI Lab, I conduct applied research in collaboration with advertising delivery platforms. Through joint research with Okumura Laboratory at Institute of Science Tokyo, I engage in a wide range of work from fundamental research to practical applications.