Literature Map: A Lineage of Local Assembly Speech Data Research — Centered on the Work of Haruka Watanabe, Yasutomo Kimura, and Kenjiro Higashi

This note organises the prior research that machikarte(external site, opens in a new tab) (the nationwide local assembly speech search platform) builds upon, arranged as three lineages of accumulated work. It introduces the lineage of local assembly minutes analysis, the lineage of large-scale corpora and NLP analysis, and the lineage of citizen-participation platforms, each anchored by a central researcher. The note also serves to publish, in literature-map form, the body of work behind the researchers under consideration as invitees to the machikarte editorial committee.

Overview — Three Lineages That Accumulated Separately

Local assembly minutes are published across all 1,788 councils in Japan. Technically, the path to cross-cutting analysis is open. Even so, research that treats minutes nationally has accumulated as three independent lineages, each forming its own research field.

One traces topic and issue extraction from assembly minutes, reading the accountability of cultural policy and other specific domains. Another builds large-scale corpora as objects of natural language processing, advancing quantitative measurement of councillor activity and stance classification. A third is the civic tech lineage that returns assembly data to citizens as material for participation.

Each stage of the research process, including data acquisition, structural extraction, and citizen feedback, has its own independent body of work. machikarte is positioned to connect all three stages on a single platform. This article introduces the prior research that grounds that position, presented as the work of representative researchers in each of the three lineages.

The researchers introduced here are Haruka Watanabe(external site, opens in a new tab) (Assistant Professor, Faculty of Global Informatics, Chuo University), Yasutomo Kimura(external site, opens in a new tab) (Professor, Faculty of Commerce, Otaru University of Commerce), and Kenjiro Higashi(external site, opens in a new tab) (Project Manager at Code for Japan and Policy Advisor to Hino Town, Shiga Prefecture).

A Timeline of the Principal Works

Laid out by year, the order of accumulation becomes visible. The methodology for treating minutes as material came first (2020); corpora and quantification followed (2023-2024); and the return of data to citizens reached the institutional arena (2024).

Four items are left off the timeline. The Local Assembly Proceedings Corpus Project and the NTCIR series are ongoing operations that cannot be placed in a single year, and Code for Japan and machikarte are organisations and platforms rather than works.

Seen as a timeline, the methodology for treating minutes as observable material came first, and the technology to run it at scale, along with the implementations that return it to citizens, caught up afterwards. machikarte sits after all three arrived, in the position of connecting them on a single platform.

The Three Lineages

The sections below follow the order the timeline revealed, one lineage at a time. They correspond to the "Lineage" column in the table above.

The Lineage of Assembly Minutes Analysis — Haruka Watanabe's Research

Haruka Watanabe's research has focused on building a methodology for reading the accountability of the policy process through local assembly minutes. As the debate around EBPM (Evidence-Based Policy Making) moved from the stage of institutional adoption to the stage of empirical data, her work laid the groundwork for how minutes should be treated academically as a research material.

Doctoral thesis "A Study on Accountability in Local Government Cultural Policy" (Graduate School of Interdisciplinary Information Studies, University of Tokyo, 2020) reconstructs the governance of public cultural facilities from local assembly minutes. It established a methodology for treating the discourse accumulated in the public layer of local assemblies as empirical material for policy-making processes. It sits closest to the academic lineage that grounds machikarte's basic stance of "reading minutes as observations."

Book Governance Theory of Public Cultural Facilities — The Practice and Transformation of Accountability by Local Governments (Waseda University Press, 2023, 312 pages) is the published version of the doctoral thesis. In 2023 it received the Society for Socio-Informatics Outstanding Publication Award and the Emerging Researcher Award simultaneously. Drawing on empirical analysis of assembly minutes from the six prefectural assemblies of Shizuoka, Sapporo, Sendai, Niigata, Yokohama, and Kyoto, it traces how the accountability of public cultural facilities has been practised and transformed in the public layer of the assembly. The book serves as a foundational reference for machikarte's cultural policy topic analysis.

Paper "An Attempt at Topic Extraction from Local Assembly Minutes" (Socio-Informatics, Vol. 9, No. 1, 2020, DOI: 10.14836/ssi.9.1_1) is a methodological paper that systematises an approach to extracting topics from minutes texts. It presents a framework that classifies debates on cultural policy into five common topics: financial burden, facility maintenance, management and operation, programmes, and the structural promotion of cultural policy. This is a methodological pillar for machikarte's municipality-level topic classification and topic search features.

Paper "Methodological Considerations for the Use of Local Assembly Minutes Data in Policy Research" (Global Informatics Studies, Vol. 5, No. 5, 2025, co-authored with Kimitake Asatani) organises the challenges of data quality assurance, preprocessing, and analytical design when treating local assembly minutes as a research material. It discusses where, across the workflow from data acquisition to analysis, the risks to academic verifiability lie. This paper shares almost the same problem awareness as machikarte's commitment to "publishing data quality assurance and verification scripts" as a core differentiator.

KAKEN Young Researcher Grant 23K12068 "Discourse Analysis of Policy Decisions on Public Theaters and Cultural Facilities — A Comparative Study of Prefectural Assembly Minutes" (FY2023-2026, research budget 2.47 million yen) is a project advancing comparative discourse analysis across the minutes of 47 prefectural assemblies. It can be read as a preparatory step for the cross-cutting analysis of all 1,788 assemblies that machikarte aims for, building up comparative analysis at the prefectural level.

These five works by Watanabe cover the three layers of methodology, empirical study, and a KAKEN-funded project, and together form the academic foundation on which the machikarte editorial committee should stand in the domains of cultural policy and accountability.

The Lineage of Large-Scale Corpora and NLP Analysis — Yasutomo Kimura's Research

Yasutomo Kimura's research has focused on building local assembly minutes into large-scale corpora and on systematising methods for quantifying councillor activity as objects of natural language processing. He is one of the central figures in Japanese political informatics and computational social science working on local assembly data.

Paper "Building a Large Corpus and Pre-trained Language Models from National and Local Assembly Minutes" (Journal of Natural Language Processing, Vol. 31, No. 2, 2024, DOI: 10.5715/jnlp.31.707) reports the construction of a large-scale corpus from 104,969 National Diet documents and 553,823 minutes from 710 local assemblies nationwide (approximately 58 GB), and the training of a political-domain-adapted BERT. It stands as one of the closest precedents to the infrastructure that machikarte operates, holding 247,063,315 assembly speeches (aggregated 18 August 2026). The paper is also a direct reference for design decisions concerning scrapers and corpus infrastructure.

Paper "Quantifying the Activities of Local Assembly Members in Japan: Recent Advances and an Approach Using the BERTopic Model" (Interdisciplinary Information Sciences, Vol. 30, No. 1, 2024, DOI: 10.4036/iis.2024.R.03, Open Access) presents a method that applies the BERTopic model to Fukushima Prefectural Assembly minutes to quantitatively visualise councillor activity. It is an important reference for understanding the academic grounding that machikarte's councillor performance indicators can rest upon.

Paper "Silence is Not Golden: Legislative Participation and Re-Election of Japanese Local Assembly Members" (Asian Journal of Law and Economics, 2026, co-authored with Akihiko Kawaura) demonstrates a positive correlation between councillor speech volume and re-election rates, drawing on 24 years of local assembly data from 2000 to 2023. As the title suggests, councillors who speak less tend to be re-elected less often. This paper provides core academic grounding for understanding how machikarte's councillor activity visualisation features may connect to citizens' voting behaviour.

Paper "Deeds, not words? Speech and re-election of Japan's local legislators" (Constitutional Political Economy, 2023, DOI: 10.1007/s10602-023-09402-3, co-authored with Akihiko Kawaura) is an earlier paper examining the causal relationship between councillor speech volume and re-election. It uses local assembly data to ask empirically whether speech is evaluated by voters as an "act." It supports the political-science argument that "speech is a meaningful observation," on which machikarte's councillor activity visualisation can build.

Local Assembly Proceedings Corpus Project (http://local-politics.jp/(external site, opens in a new tab)) is a cross-search system for all 47 prefectural assemblies directed by Kimura. It is one of the technical predecessors that machikarte references, and the work that first implemented the idea of cross-cutting search across local assemblies as an academic project.

NTCIR QA Lab-PoliInfo Series (NTCIR-14, 2019 / NTCIR-15, 2020 / NTCIR-16, 2022 / NTCIR-17, 2023 / NTCIR-18 U4, 2025) is an international evaluation workshop designing and operating tasks in political-information question answering, summarisation, stance classification, and budget argument mining; Kimura is one of its principal organisers. The stance classification tasks at NTCIR-15 and NTCIR-17 form the direct methodological grounding for machikarte's stance classification feature. The Budget Argument Mining task at NTCIR-16 is the reference point for budget-debate analysis features.

KAKEN Grant-in-Aid for Scientific Research (B) 23K21841 "Organic Linking of Unstructured and Structured Data in Local Assemblies" (FY2024 onwards, research budget 16.38 million yen) is a project advancing corpus construction that links the unstructured data of minutes with the structured data of budgets, ordinances, and elections. It aligns with the direction machikarte aims for in the future: connecting minutes data with budget and ordinance data.

Book Promoting Local Government DX and the Use of Open Data (Nippon Hyoron Sha, 2022, edited volume, with Masami Honda, Kazunori Kawamura, Keiichi Takamaru, Yuzu Uchida, Hokuto Otobe and others) organises the practice of local government DX and open data policy as an edited volume. It is a valuable reference for the academic framing of the institutional background on which machikarte stands (local government open data policy, the Digital Agency, the Ministry of Internal Affairs and Communications' Local Government DX Promotion Plan, and similar).

These seven works by Kimura cover multiple layers, including corpus infrastructure, councillor activity quantification, stance classification, argument mining, KAKEN projects, and edited volumes, and together cover almost the full range of academic foundation that machikarte's NLP analysis features need to stand upon.

The Lineage of Citizen-Participation Platforms and Civic Tech — Kenjiro Higashi's Work

Kenjiro Higashi's accumulated work focuses on the implementation and institutional design of civic tech that returns assembly speech data and administrative information back to citizens as material for participation. After 13 years of internal administrative practice as an official of the Kyoto Prefectural Government (2007-2020), he now occupies the position of deploying citizen-participation platforms to local governments across the country at Code for Japan.

Book Digital Society and Local Governments — The Future of Local Autonomy and Urban Management (Japan Center for Cities, 2024, contributing author) is a collection covering the institutional framing of local autonomy in a digital society. Higashi contributed two chapters: Chapter 6 "Resident Self-Governance in a Digital Society" and Chapter 7 "The Shape of Co-Creation between Residents and Government — Ten Years of Civic Tech." It can be positioned as one of the most important references for the institutional framing of civic tech in Japan. It is a touchstone for how machikarte should connect, as a "platform that returns assembly data to citizens," to the lineage of civic tech.

Book Transition from the Local Toward Sustainable Development (Kankyo Shimbunsha, 2023, contributing author, edited by Nobuo Shirai and Hideaki Kurishima) is a collection on local transitions. Higashi contributed Chapter 4 "Digital Democracy Using a Participatory Consensus-Building Platform." It is written from the standpoint of someone who actually operates participatory platforms including Decidim(external site, opens in a new tab). It suggests how machikarte, when returning assembly data to citizens, can connect with the layer of consensus building.

Contribution "The Practice of Participatory Democracy Platforms in the Pandemic Era — Decidim as an Example" (Gihyo, Software Design, October 2021 issue) is a contribution covering the technical explanation of Decidim and its domestic implementation in Japan. It has become one of the standard introductory references for the diffusion of Decidim in Japan. It is a reference for the technology and operations when machikarte connects assembly data, as a starting point, to citizen participation.

Witness, House of Councillors Committee on General Affairs (June 11, 2024, 213th Session of the Diet, deliberation on the Local Autonomy Act amendment bill): Higashi appeared as one of four witnesses alongside Izuru Makihara (University of Tokyo), Takaharu Ohara (Waseda University), and Takio Honda (Ryukoku University). His title was "Code for Japan / Policy Advisor to Hino Town, Shiga Prefecture." Speaking as a civic tech practitioner in a forum on institutional design at the level of the Local Autonomy Act amendment is a case that suggests where machikarte stands institutionally as a "platform that returns assembly data to citizens."

Code for Japan as Decidim Official Partner (April 2024, first in Japan): As the practical lead for this move, Higashi has driven the domestic deployment ever since Kakogawa City (October 2020, Japan's first Decidim adoption). He has led deployment to Nishi-Aizu Town, Yosano Town, Kamaishi City, Hyogo Prefecture, Shiga Prefecture and others. When machikarte deploys to small-scale and basic municipalities, his work represents the closest practical foundation as a connection point with local governments that already have Decidim adoption experience.

These five items in Higashi's work cover multiple layers, including books, article contributions, Diet witness testimony, and practical leadership, and together form the practitioner foundation on which the machikarte editorial committee should stand in the lineages of civic tech and citizen participation.

Key Concepts — Three Stages of the Research Process

Surveying the three lineages, the entire research process around local assembly speech data is almost completely covered.

machikarte can be positioned as a platform design that aims to integrate all three of these stages. It draws on Kimura's methodology in corpus infrastructure and NLP analysis, anticipates connections with Watanabe's methodology in minutes analysis and cultural policy accountability, and envisions linking to Higashi's civic tech practice and lineage of citizen participation.

A Map Linking Research Lineages and machikarte Features

The correspondences between the three lineages and the features that machikarte implements or plans to implement are organised below on a feature-by-feature basis.

Lined up side by side, the three lineages reveal a design in which each feature is connected to prior research on a unit-by-unit basis. Rather than re-inventing the accumulated work of the three lineages, machikarte is designed to inherit each methodology and to stand as a platform that integrates the three lineages.

Who Carries This Lineage Forward

Building on the prior research of the three lineages, machikarte's intended new contributions can be organised under three points.

The first is the application of cross-cutting analysis to all 1,788 assemblies. Prior research has been bounded either at the level of 47 prefectural assemblies (Watanabe's six prefectures, Kimura's prefectural corpus) or at a subset of municipalities (Kimura's 710 local governments, Fukushima Prefectural Assembly). Treating the full data of all 1,788 assemblies as the object of cross-cutting analysis is a natural extension of prior research, but it cannot stand without a continuously operated infrastructure for data acquisition and quality assurance.

The second is reproducibility through the publication of data quality assurance and verification scripts. The methodological papers of prior research (Watanabe 2025, Kimura 2024) have organised data-quality challenges academically. machikarte places at the core of its differentiation the publication of that quality-assurance logic on GitHub, so that third parties can reach the same conclusions. It is positioned to extend the academic debate around reproducibility to reproducibility as an operated platform.

The third is automation of analysis through AI agent integration. As the EBPM debate enters the AI era, publishing assembly data in a form that AI agents can handle (such as MCP Tool exposure) becomes less an extension of a research field and more an attempt to redesign the boundary between research and practice. This is a Phase β area for consideration onwards.

The Editorial Committee as a Way of Carrying It

machikarte is taking shape inside the research community as a platform built upon the accumulated work of three lineages. It is natural to position its editorial committee, too, as a forum for dialogue between researchers and practitioners.

Inviting one representative researcher from each of the three lineages to take part in reviewing feature design and editorial judgment is also a design choice that declares the platform's intent to inherit and operate the accumulated work of each lineage rather than to launch its own methodology from scratch. In forming the editorial committee for Phase α, the goal is a composition that balances the three lineages.

By having three researchers from distinct fields look at the same platform, the aim is to have machikarte's editorial judgment and feature design verified from three mutually complementary perspectives: the methodology of minutes analysis, the design of NLP corpora, and the implementation of civic tech.

Limits — Where the Prior Research Stops, and Where machikarte Stops

A literature map is finished not only by what has accumulated, but by what has not.

Where the prior research stops. The objects of analysis are six prefectural assemblies for Watanabe, and 710 municipal assemblies plus the Fukushima Prefectural Assembly for Kimura. No study in any of the three lineages treats all 1,788 assemblies as a single object. Stance classification and argument mining are designed as tasks within NTCIR, but no implementation on a nationwide platform is evident. On the citizen-participation side, Higashi's work centres on consensus-building implementations; the connection to minutes data remains at the level of written argument.

Where machikarte stops. In the aggregation of 18 August 2026, of the 1,788 assemblies in the reporting denominator, 31 have not a single speech on record. A further 600 or so have speeches but no list of meetings (a figure that falls as generation proceeds), meaning that for roughly a third of all assemblies we cannot show at which meeting something was said. Topic aggregation covers 1,529 assemblies; the rest are outside it. machikarte does not, in fact, handle the national scale that prior research could not reach. On scale, it is partway there.

These two are not the same problem. The prior research had a limited range because its studies were designed to narrow their object. machikarte's gaps exist because the acquisition and generation pipeline has not yet reached every assembly. The first is a question of methodology; the second is a question of operations.

Trying It on machikarte

Analyses similar to those handled in the works above can be tried on machikarte. What is open today and what does not yet exist are listed separately.

• The nationwide distribution of assembly topics comes close to the cultural policy debates covered by Watanabe's doctoral thesis and book, the distribution of each topic across the country can be followed from Topic distribution(external site, opens in a new tab). There is no page that isolates cultural policy on its own
• Descriptive statistics per councillor run in line with Kimura's councillor activity quantification research, descriptive statistics for each councillor can be reached from the councillor index(external site, opens in a new tab). Comparison of speech volume itself is not implemented
• On stance classification of agenda items, a feature inheriting the methodology of Kimura's NTCIR stance classification tasks does not yet exist

How the aggregates are built is set out in Methodology(external site, opens in a new tab), and what has been collected so far is in Coverage(external site, opens in a new tab). The work is positioned to inherit the methodology of prior research while extending the application to all 1,788 assemblies.

References

1. 1.
   A Study on Accountability in Local Government Cultural Policy↗(external site, opens in a new tab)

   Watanabe, H. (2020)

   Doctoral thesis, Graduate School of Interdisciplinary Information Studies, University of Tokyo

2. 2.
   公立文化施設のガバナンス論: 自治体によるアカウンタビリティの実践と変容↗(external site, opens in a new tab)

   Watanabe, H. (2023)

   Waseda University Press (ISBN 978-4-657-23803-0, 312 pages)

3. 3.
   An Attempt at Topic Extraction from Local Assembly Minutes↗(external site, opens in a new tab)

   Watanabe, H. (2020)

   Socio-Informatics 9(1), DOI: 10.14836/ssi.9.1_1

4. 4.
   Methodological Considerations for the Use of Local Assembly Minutes Data in Policy Research↗(external site, opens in a new tab)

   Watanabe, H. & Asatani, K. (2025)

   Global Informatics Studies 5(5)

5. 5.
   Discourse Analysis of Policy Decisions on Public Theaters and Cultural Facilities — A Comparative Study of Prefectural Assembly Minutes (KAKEN 23K12068)↗(external site, opens in a new tab)

   Watanabe, H. (2023)

   Japan Society for the Promotion of Science, Young Researcher Grant (FY2023-2026, 2.47 million yen)

6. 6.
   Building a Large Corpus and Pre-trained Language Models from National and Local Assembly Minutes↗(external site, opens in a new tab)

   Nagafuchi, Kimura, Kadowaki, Araki (2024)

   Journal of Natural Language Processing 31(2), DOI: 10.5715/jnlp.31.707

7. 7.
   Quantifying the Activities of Local Assembly Members in Japan: Recent Advances and an Approach Using the BERTopic Model↗(external site, opens in a new tab)

   Li, Kawamura, Kimura (2024)

   Interdisciplinary Information Sciences 30(1), DOI: 10.4036/iis.2024.R.03

8. 8.
   Silence is Not Golden: Legislative Participation and Re-Election of Japanese Local Assembly Members↗(external site, opens in a new tab)

   Kimura, Y. & Kawaura, A. (2026)

   Asian Journal of Law and Economics

9. 9.
   Deeds, not words? Speech and re-election of Japan's local legislators↗(external site, opens in a new tab)

   Kimura, Y. & Kawaura, A. (2023)

   Constitutional Political Economy, DOI: 10.1007/s10602-023-09402-3

10. 10.
    Local Assembly Proceedings Corpus Project↗(external site, opens in a new tab)

    Kimura, Y. et al. (2026)

    local-politics.jp

11. 11.
    NTCIR QA Lab-PoliInfo Series (NTCIR-14, 15, 16, 17, 18 U4)↗(external site, opens in a new tab)

    Kimura, Y. et al. (2025)

    NII Testbeds and Community for Information access Research

12. 12.
    Organic Linking of Unstructured and Structured Data in Local Assemblies (KAKEN 23K21841)↗(external site, opens in a new tab)

    Kimura, Y. (2024)

    Japan Society for the Promotion of Science, Grant-in-Aid for Scientific Research (B) (FY2024 onwards, 16.38 million yen)

13. 13.
    自治体DX推進とオープンデータの活用↗(external site, opens in a new tab)

    Kimura, Y. (ed.), Honda, M., Kawamura, K., Takamaru, K., Uchida, Y., Otobe, H. et al. (2022)

    Nippon Hyoron Sha

14. 14.
    デジタル社会と自治体: 地方自治と都市経営の未来 (Chapters 6 and 7, 'Ten Years of Civic Tech')↗(external site, opens in a new tab)

    Higashi, K. et al. (2024)

    Japan Center for Cities

15. 15.
    持続可能な発展に向けた地域からのトランジション (Chapter 4, 'Digital Democracy Using a Participatory Consensus-Building Platform')↗(external site, opens in a new tab)

    Higashi, K. / Shirai, N. & Kurishima, H. (eds.) (2023)

    Kankyo Shimbunsha

16. 16.
    The Practice of Participatory Democracy Platforms in the Pandemic Era — Decidim as an Example↗(external site, opens in a new tab)

    Higashi, K. (2021)

    Gihyo, Software Design, October 2021 issue

17. 17.
    213th Session of the Diet, House of Councillors Committee on General Affairs — Witness Statement (Deliberation on the Local Autonomy Act Amendment Bill)↗(external site, opens in a new tab)

    Higashi, K. (Code for Japan / Policy Advisor to Hino Town, Shiga Prefecture) (2024)

    House of Councillors (Sangiin)

18. 18.
    Code for Japan (Decidim Official Partner)↗(external site, opens in a new tab)

    Code for Japan (2024)

    Code for Japan

19. 19.
    machikarte — Nationwide Local Assembly Speech Search Platform (Beta)↗(external site, opens in a new tab)

    Institute for Social Vision Design (ISVD) (2026)

    ISVD

Related Content

Labs June 1, 2026

National Distribution of Deferral Phrasing in Assembly Responses — A Structural Analysis of 18.97 Million Records from 870 Municipalities

Labs June 1, 2026

Local Assembly Civic Data Lab — Hypotheses and Overview: Reading Assembly Speeches as Observations

Labs June 9, 2026

Seven Years of 'Care' Mentions in Japanese Local Assemblies — The Time Structure of Long-Term Care Insurance Act Revisions and Policy Lag

Labs June 20, 2026

Regional Distribution of 'Foreign Workforce' Debate in Japanese Local Assemblies — A Seven-Year Structural Analysis Across 870 Municipalities