Role-based Knowledge Graph Analysis of Mission-oriented Research Funding in Japan

Understanding the policy objectives under which mission-oriented research funding is allocated, and the targets and methods it supports, matters for interpreting research investment. Existing topic-based analyses mix words playing different roles (policy objective, purpose, target, method, and evaluation) within a single cluster, making it hard to read policy implications from funding. We propose the Role-based Research Knowledge Graph (R-RKG), which decomposes each project description into these five semantic roles. Using a large language model, we decomposed 20,854 projects supported by Japanese mission-oriented funding agencies and constructed the graph. We evaluated extraction quality with a large language model as a judge and two research-policy experts: mean scores for faithfulness, role appropriateness, and completeness ranged from 4.56 to 4.97 on a five-point scale, and judge and experts agreed within a mean absolute difference of 0.32. Comparing R-RKG against topic-based clustering and amount aggregation, we report four findings. First, role distributions differ systematically across agencies, which fall into three types: strategy-exploring, problem-solving, and infrastructure-building. Second, policy objectives change fastest over time while methods and evaluation stay stable overall, and the degree of change differs by agency. Third, the methods and targets of research that realize a given policy objective branch out or converge over time, appearing in both project counts and allocated amounts. Fourth, even the same technical term (for example , AI) shows a systematically different per-project allocation when used as a method versus adopted as a goal. These role-specific findings are not obtainable from topic-based analysis, and we release the dataset.This is a preprint of a manuscript that has not been peer reviewed.

Authors

Institutions

Publication Details

Journal
University of Tokyo Repository (University of Tokyo)
Published
2026-11-12
DOI
https://doi.org/10.82337/1000046
Primary Topic
Research, Science, and Academia
Type
preprint
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
preprint

Role-based Knowledge Graph Analysis of Mission-oriented Research Funding in Japan

Takahiro Kawamura, Shigenori Watanabe
University of Tokyo Repository (University of Tokyo)
Research, Science, and Academia
preprint

Role-based Knowledge Graph Analysis of Mission-oriented Research Funding in Japan

Takahiro Kawamura, Shigenori Watanabe
preprint en

Abstract

Understanding the policy objectives under which mission-oriented research funding is allocated, and the targets and methods it supports, matters for interpreting research investment. Existing topic-based analyses mix words playing different roles (policy objective, purpose, target, method, and evaluation) within a single cluster, making it hard to read policy implications from funding. We propose the Role-based Research Knowledge Graph (R-RKG), which decomposes each project description into these five semantic roles. Using a large language model, we decomposed 20,854 projects supported by Japanese mission-oriented funding agencies and constructed the graph. We evaluated extraction quality with a large language model as a judge and two research-policy experts: mean scores for faithfulness, role appropriateness, and completeness ranged from 4.56 to 4.97 on a five-point scale, and judge and experts agreed within a mean absolute difference of 0.32. Comparing R-RKG against topic-based clustering and amount aggregation, we report four findings. First, role distributions differ systematically across agencies, which fall into three types: strategy-exploring, problem-solving, and infrastructure-building. Second, policy objectives change fastest over time while methods and evaluation stay stable overall, and the degree of change differs by agency. Third, the methods and targets of research that realize a given policy objective branch out or converge over time, appearing in both project counts and allocated amounts. Fourth, even the same technical term (for example , AI) shows a systematically different per-project allocation when used as a method versus adopted as a goal. These role-specific findings are not obtainable from topic-based analysis, and we release the dataset.This is a preprint of a manuscript that has not been peer reviewed.

University of Tokyo Repository (University of Tokyo)
The University of Tokyo (JP)
Research, Science, and Academia
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon