Jonas Rieger
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Interests
Jonas Rieger has a total of 33 co-authors. Here are the top co-authors by number of shared publications:
Carsten Jentsch
Department of Statistics, TU Dortmund University, Dortmund, Germany
21 shared publicationsJörg Rahnenführer
Department of Statistics, TU Dortmund University, Dortmund, Germany
20 shared publications- Carsten Jentsch
TU Dortmund University
5 shared publications - Gerret von Nordheim
DER SPIEGEL, Dokumentation, SPIEGEL-Verlag Rudolf Augstein GmbH & Co. KG, Hamburg, Deutschland
2 shared publications - Jörg Rahnenführer
Department of Statistics, TU Dortmund University, Dortmund, Germany
2 shared publications - Andreas Fischer
Forschungsinstitut Betriebliche Bildung (f-bb)
1 shared publication
Latest Publications
- Open Access
- 20 August 2026
- Published by Springer Nature in Empirical Economics
- Vol. 71 (3) , 30
Changing inflation perception and the propagation of news shocks
Based on a previously established text data-based indicator, the Inflation Perception Indicator (IPI), we analyze the information content of inflation-related media coverage. It is argued that media do not report on inflation in a constant or systematic manner but instead react to specific events, which means that reporting is state-dependent. Using a dynamic extension of the Latent Dirichlet Allocation (LDA) topic modeling approach we examine the salience and framing of inflation in a corpus of 50,000 articles published by major German newspapers between 2001 and 2024. This approach allows to decompose the IPI into dynamic topic-specific subindicators. To analyze the information content of these newly constructed indicators, we estimate a threshold Bayesian Vector AutoRegressive (BVAR) model for variables from a New Keynesian macroeconomic model with two inflation regimes. Our results suggest that the topic-specific IPI subindicators contain different information about price developments. Moreover, we find some evidence that households make different use of this information in high- and low-inflation regimes when adjusting their inflation expectations.
- 20 August 2026
- Published by Springer Nature in Empirical Economics
- Vol. 71 (3) , 30
Open AccessChanging inflation perception and the propagation of news shocks
Based on a previously established text data-based indicator, the Inflation Perception Indicator (IPI), we analyze the information content of inflation-related media coverage. It is argued that media do not report on inflation in a constant or systematic manner but instead react to specific events, which means that reporting is state-dependent. Using a dynamic extension of the Latent Dirichlet Allocation (LDA) topic modeling approach we examine the salience and framing of inflation in a corpus of 50,000 articles published by major German newspapers between 2001 and 2024. This approach allows to decompose the IPI into dynamic topic-specific subindicators. To analyze the information content of these newly constructed indicators, we estimate a threshold Bayesian Vector AutoRegressive (BVAR) model for variables from a New Keynesian macroeconomic model with two inflation regimes. Our results suggest that the topic-specific IPI subindicators contain different information about price developments. Moreover, we find some evidence that households make different use of this information in high- and low-inflation regimes when adjusting their inflation expectations.
- Open Access
- 17 June 2026
- Published by Center for Open Science in SocArXiv
Following Socio-Environmental Conflict Narratives About Energy Transition in Chile: A Spatio-Temporal Analysis Using Dynamic Topic Modeling
Understanding the construction of socio-environmental narratives at a national scale is a complex challenge, particularly when research remains fragmented across disconnected case studies. In Chile, the energy transition has generated territorial disputes as extractive industries and renewable energy projects expand, yet large-scale systematic analyses of how these conflicts are represented in public discourse remain scarce. This paper addresses this gap by applying a spatio-temporal topic modelling framework to a corpus of 1,996 validated news articles covering conflicts related to the energy transition in Chile from 2011 to 2025. Using RollingLDA, a dynamic adaptation of latent Dirichlet allocation that prevents information leakage from future documents, we identify twelve topics that provide insights into the public narratives surrounding socio-environmental conflicts. Our analysis reveals how specific conflicts, such as the HidroAysén dam project, the Dominga mining controversy, and pollution in sacrifice zones such as Quintero-Puchuncaví, have evolved over time, with some narratives declining while others, including green hydrogen development and lithium extraction, have emerged as central concerns. We complement this temporal analysis with a spatial dimension by mapping the prevalence of topics across Chilean regions through an interactive dashboard. By combining established methods, our work offers a reproducible framework that can be adapted to topic modelling results incorporating spatial and temporal dimensions, enabling the tracking of how socio-environmental narratives emerge, evolve, and fade over time. Please also refer to the GitHub repository at https://github.com/JonasRieger/t2s2026.
- 17 June 2026
- Published by Center for Open Science in SocArXiv
Open AccessFollowing Socio-Environmental Conflict Narratives About Energy Transition in Chile: A Spatio-Temporal Analysis Using Dynamic Topic Modeling
Understanding the construction of socio-environmental narratives at a national scale is a complex challenge, particularly when research remains fragmented across disconnected case studies. In Chile, the energy transition has generated territorial disputes as extractive industries and renewable energy projects expand, yet large-scale systematic analyses of how these conflicts are represented in public discourse remain scarce. This paper addresses this gap by applying a spatio-temporal topic modelling framework to a corpus of 1,996 validated news articles covering conflicts related to the energy transition in Chile from 2011 to 2025. Using RollingLDA, a dynamic adaptation of latent Dirichlet allocation that prevents information leakage from future documents, we identify twelve topics that provide insights into the public narratives surrounding socio-environmental conflicts. Our analysis reveals how specific conflicts, such as the HidroAysén dam project, the Dominga mining controversy, and pollution in sacrifice zones such as Quintero-Puchuncaví, have evolved over time, with some narratives declining while others, including green hydrogen development and lithium extraction, have emerged as central concerns. We complement this temporal analysis with a spatial dimension by mapping the prevalence of topics across Chilean regions through an interactive dashboard. By combining established methods, our work offers a reproducible framework that can be adapted to topic modelling results incorporating spatial and temporal dimensions, enabling the tracking of how socio-environmental narratives emerge, evolve, and fade over time. Please also refer to the GitHub repository at https://github.com/JonasRieger/t2s2026.
- Open Access
Zeitenwenden: Detecting changes in the German political discourse
From a monarchy to a democracy, to a dictatorship and back to a democracy -- the German political landscape has been constantly changing ever since the first German national state was formed in 1871. After World War II, the Federal Republic of Germany was formed in 1949. Since then every plenary session of the German Bundestag was logged and even has been digitized over the course of the last few years. We analyze these texts using a time series variant of the topic model LDA to investigate which events had a lasting effect on the political discourse and how the political topics changed over time. This allows us to detect changes in word frequency (and thus key discussion points) in political discourse.
Open AccessZeitenwenden: Detecting changes in the German political discourse
From a monarchy to a democracy, to a dictatorship and back to a democracy -- the German political landscape has been constantly changing ever since the first German national state was formed in 1871. After World War II, the Federal Republic of Germany was formed in 1949. Since then every plenary session of the German Bundestag was logged and even has been digitized over the course of the last few years. We analyze these texts using a time series variant of the topic model LDA to investigate which events had a lasting effect on the political discourse and how the political topics changed over time. This allows us to detect changes in word frequency (and thus key discussion points) in political discourse.
- Open Access
- 20 September 2024
- Published by PeerJ in PeerJ Computer Science
- Vol. 10, e2279
LDAPrototype: a model selection algorithm to improve reliability of latent Dirichlet allocation
Latent Dirichlet allocation (LDA) is a popular method for analyzing large text corpora, but it suffers from instability due to its reliance on random initialization. This results in different outcomes for replicated runs, hindering reproducibility. To address this, we introduce LDAPrototype, a new approach for selecting the most representative LDA run from multiple replications on the same dataset. LDAPrototype enhances the reliability of LDA conclusions by ensuring greater similarity between replications compared to traditional LDA runs or models chosen based on perplexity or NPMI. A key feature of LDAPrototype is its use of a novel model similarity measure called S-CLOP (Similarity of multiple sets by Clustering with LOcal Pruning). It is based on topic similarities, for which we compare the usage of measures like the thresholded Jaccard coefficient, cosine similarity, Jensen-Shannon divergence, and rank-biased overlap. The effectiveness of LDAPrototype is demonstrated through its application to six real datasets, including newspaper articles and tweets. The results show improved reproducibility and reliability in topic modeling outcomes. LDAPrototype’s approach is noteworthy for its practical applicability, comprehensibility, ease of implementation, and computational efficiency. Furthermore, the algorithm’s concept can be generalized to other topic modeling procedures that characterize topics through word distributions, making it a versatile tool in text data analysis.
- 20 September 2024
- Published by PeerJ in PeerJ Computer Science
- Vol. 10, e2279
Open AccessLDAPrototype: a model selection algorithm to improve reliability of latent Dirichlet allocation
Latent Dirichlet allocation (LDA) is a popular method for analyzing large text corpora, but it suffers from instability due to its reliance on random initialization. This results in different outcomes for replicated runs, hindering reproducibility. To address this, we introduce LDAPrototype, a new approach for selecting the most representative LDA run from multiple replications on the same dataset. LDAPrototype enhances the reliability of LDA conclusions by ensuring greater similarity between replications compared to traditional LDA runs or models chosen based on perplexity or NPMI. A key feature of LDAPrototype is its use of a novel model similarity measure called S-CLOP (Similarity of multiple sets by Clustering with LOcal Pruning). It is based on topic similarities, for which we compare the usage of measures like the thresholded Jaccard coefficient, cosine similarity, Jensen-Shannon divergence, and rank-biased overlap. The effectiveness of LDAPrototype is demonstrated through its application to six real datasets, including newspaper articles and tweets. The results show improved reproducibility and reliability in topic modeling outcomes. LDAPrototype’s approach is noteworthy for its practical applicability, comprehensibility, ease of implementation, and computational efficiency. Furthermore, the algorithm’s concept can be generalized to other topic modeling procedures that characterize topics through word distributions, making it a versatile tool in text data analysis.
- Open Access
- 20 September 2024
- Published by PeerJ
- Vol. 10, e2279
Figure 9: Increase of reliability for the reuters dataset for R = 50, 100, 200, 500 and K = 5, …, 15.
Latent Dirichlet allocation (LDA) is a popular method for analyzing large text corpora, but it suffers from instability due to its reliance on random initialization. This results in different outcomes for replicated runs, hindering reproducibility. To address this, we introduce LDAPrototype, a new approach for selecting the most representative LDA run from multiple replications on the same dataset. LDAPrototype enhances the reliability of LDA conclusions by ensuring greater similarity between replications compared to traditional LDA runs or models chosen based on perplexity or NPMI. A key feature of LDAPrototype is its use of a novel model similarity measure called S-CLOP (Similarity of multiple sets by Clustering with LOcal Pruning). It is based on topic similarities, for which we compare the usage of measures like the thresholded Jaccard coefficient, cosine similarity, Jensen-Shannon divergence, and rank-biased overlap. The effectiveness of LDAPrototype is demonstrated through its application to six real datasets, including newspaper articles and tweets. The results show improved reproducibility and reliability in topic modeling outcomes. LDAPrototype’s approach is noteworthy for its practical applicability, comprehensibility, ease of implementation, and computational efficiency. Furthermore, the algorithm’s concept can be generalized to other topic modeling procedures that characterize topics through word distributions, making it a versatile tool in text data analysis.
- 20 September 2024
- Published by PeerJ
- Vol. 10, e2279
Open AccessFigure 9: Increase of reliability for the reuters dataset for R = 50, 100, 200, 500 and K = 5, …, 15.
Latent Dirichlet allocation (LDA) is a popular method for analyzing large text corpora, but it suffers from instability due to its reliance on random initialization. This results in different outcomes for replicated runs, hindering reproducibility. To address this, we introduce LDAPrototype, a new approach for selecting the most representative LDA run from multiple replications on the same dataset. LDAPrototype enhances the reliability of LDA conclusions by ensuring greater similarity between replications compared to traditional LDA runs or models chosen based on perplexity or NPMI. A key feature of LDAPrototype is its use of a novel model similarity measure called S-CLOP (Similarity of multiple sets by Clustering with LOcal Pruning). It is based on topic similarities, for which we compare the usage of measures like the thresholded Jaccard coefficient, cosine similarity, Jensen-Shannon divergence, and rank-biased overlap. The effectiveness of LDAPrototype is demonstrated through its application to six real datasets, including newspaper articles and tweets. The results show improved reproducibility and reliability in topic modeling outcomes. LDAPrototype’s approach is noteworthy for its practical applicability, comprehensibility, ease of implementation, and computational efficiency. Furthermore, the algorithm’s concept can be generalized to other topic modeling procedures that characterize topics through word distributions, making it a versatile tool in text data analysis.
- Open Access
- 20 September 2024
- Published by PeerJ
- Vol. 10, e2279
Figure 1: Schematic representation of the determination of a prototype based on a set of LDA models.
Latent Dirichlet allocation (LDA) is a popular method for analyzing large text corpora, but it suffers from instability due to its reliance on random initialization. This results in different outcomes for replicated runs, hindering reproducibility. To address this, we introduce LDAPrototype, a new approach for selecting the most representative LDA run from multiple replications on the same dataset. LDAPrototype enhances the reliability of LDA conclusions by ensuring greater similarity between replications compared to traditional LDA runs or models chosen based on perplexity or NPMI. A key feature of LDAPrototype is its use of a novel model similarity measure called S-CLOP (Similarity of multiple sets by Clustering with LOcal Pruning). It is based on topic similarities, for which we compare the usage of measures like the thresholded Jaccard coefficient, cosine similarity, Jensen-Shannon divergence, and rank-biased overlap. The effectiveness of LDAPrototype is demonstrated through its application to six real datasets, including newspaper articles and tweets. The results show improved reproducibility and reliability in topic modeling outcomes. LDAPrototype’s approach is noteworthy for its practical applicability, comprehensibility, ease of implementation, and computational efficiency. Furthermore, the algorithm’s concept can be generalized to other topic modeling procedures that characterize topics through word distributions, making it a versatile tool in text data analysis.
- 20 September 2024
- Published by PeerJ
- Vol. 10, e2279
Open AccessFigure 1: Schematic representation of the determination of a prototype based on a set of LDA models.
Latent Dirichlet allocation (LDA) is a popular method for analyzing large text corpora, but it suffers from instability due to its reliance on random initialization. This results in different outcomes for replicated runs, hindering reproducibility. To address this, we introduce LDAPrototype, a new approach for selecting the most representative LDA run from multiple replications on the same dataset. LDAPrototype enhances the reliability of LDA conclusions by ensuring greater similarity between replications compared to traditional LDA runs or models chosen based on perplexity or NPMI. A key feature of LDAPrototype is its use of a novel model similarity measure called S-CLOP (Similarity of multiple sets by Clustering with LOcal Pruning). It is based on topic similarities, for which we compare the usage of measures like the thresholded Jaccard coefficient, cosine similarity, Jensen-Shannon divergence, and rank-biased overlap. The effectiveness of LDAPrototype is demonstrated through its application to six real datasets, including newspaper articles and tweets. The results show improved reproducibility and reliability in topic modeling outcomes. LDAPrototype’s approach is noteworthy for its practical applicability, comprehensibility, ease of implementation, and computational efficiency. Furthermore, the algorithm’s concept can be generalized to other topic modeling procedures that characterize topics through word distributions, making it a versatile tool in text data analysis.
- Open Access
- 20 September 2024
- Published by PeerJ
- Vol. 10, e2279
Table 4: Reliability scores according to Fig. 6.
Latent Dirichlet allocation (LDA) is a popular method for analyzing large text corpora, but it suffers from instability due to its reliance on random initialization. This results in different outcomes for replicated runs, hindering reproducibility. To address this, we introduce LDAPrototype, a new approach for selecting the most representative LDA run from multiple replications on the same dataset. LDAPrototype enhances the reliability of LDA conclusions by ensuring greater similarity between replications compared to traditional LDA runs or models chosen based on perplexity or NPMI. A key feature of LDAPrototype is its use of a novel model similarity measure called S-CLOP (Similarity of multiple sets by Clustering with LOcal Pruning). It is based on topic similarities, for which we compare the usage of measures like the thresholded Jaccard coefficient, cosine similarity, Jensen-Shannon divergence, and rank-biased overlap. The effectiveness of LDAPrototype is demonstrated through its application to six real datasets, including newspaper articles and tweets. The results show improved reproducibility and reliability in topic modeling outcomes. LDAPrototype’s approach is noteworthy for its practical applicability, comprehensibility, ease of implementation, and computational efficiency. Furthermore, the algorithm’s concept can be generalized to other topic modeling procedures that characterize topics through word distributions, making it a versatile tool in text data analysis.
- 20 September 2024
- Published by PeerJ
- Vol. 10, e2279
Open AccessTable 4: Reliability scores according to Fig. 6.
Latent Dirichlet allocation (LDA) is a popular method for analyzing large text corpora, but it suffers from instability due to its reliance on random initialization. This results in different outcomes for replicated runs, hindering reproducibility. To address this, we introduce LDAPrototype, a new approach for selecting the most representative LDA run from multiple replications on the same dataset. LDAPrototype enhances the reliability of LDA conclusions by ensuring greater similarity between replications compared to traditional LDA runs or models chosen based on perplexity or NPMI. A key feature of LDAPrototype is its use of a novel model similarity measure called S-CLOP (Similarity of multiple sets by Clustering with LOcal Pruning). It is based on topic similarities, for which we compare the usage of measures like the thresholded Jaccard coefficient, cosine similarity, Jensen-Shannon divergence, and rank-biased overlap. The effectiveness of LDAPrototype is demonstrated through its application to six real datasets, including newspaper articles and tweets. The results show improved reproducibility and reliability in topic modeling outcomes. LDAPrototype’s approach is noteworthy for its practical applicability, comprehensibility, ease of implementation, and computational efficiency. Furthermore, the algorithm’s concept can be generalized to other topic modeling procedures that characterize topics through word distributions, making it a versatile tool in text data analysis.
- Open Access
- 20 September 2024
- Published by PeerJ
- Vol. 10, e2279
Table 6: Reliability scores according to Fig. 8.
Latent Dirichlet allocation (LDA) is a popular method for analyzing large text corpora, but it suffers from instability due to its reliance on random initialization. This results in different outcomes for replicated runs, hindering reproducibility. To address this, we introduce LDAPrototype, a new approach for selecting the most representative LDA run from multiple replications on the same dataset. LDAPrototype enhances the reliability of LDA conclusions by ensuring greater similarity between replications compared to traditional LDA runs or models chosen based on perplexity or NPMI. A key feature of LDAPrototype is its use of a novel model similarity measure called S-CLOP (Similarity of multiple sets by Clustering with LOcal Pruning). It is based on topic similarities, for which we compare the usage of measures like the thresholded Jaccard coefficient, cosine similarity, Jensen-Shannon divergence, and rank-biased overlap. The effectiveness of LDAPrototype is demonstrated through its application to six real datasets, including newspaper articles and tweets. The results show improved reproducibility and reliability in topic modeling outcomes. LDAPrototype’s approach is noteworthy for its practical applicability, comprehensibility, ease of implementation, and computational efficiency. Furthermore, the algorithm’s concept can be generalized to other topic modeling procedures that characterize topics through word distributions, making it a versatile tool in text data analysis.
- 20 September 2024
- Published by PeerJ
- Vol. 10, e2279
Open AccessTable 6: Reliability scores according to Fig. 8.
Latent Dirichlet allocation (LDA) is a popular method for analyzing large text corpora, but it suffers from instability due to its reliance on random initialization. This results in different outcomes for replicated runs, hindering reproducibility. To address this, we introduce LDAPrototype, a new approach for selecting the most representative LDA run from multiple replications on the same dataset. LDAPrototype enhances the reliability of LDA conclusions by ensuring greater similarity between replications compared to traditional LDA runs or models chosen based on perplexity or NPMI. A key feature of LDAPrototype is its use of a novel model similarity measure called S-CLOP (Similarity of multiple sets by Clustering with LOcal Pruning). It is based on topic similarities, for which we compare the usage of measures like the thresholded Jaccard coefficient, cosine similarity, Jensen-Shannon divergence, and rank-biased overlap. The effectiveness of LDAPrototype is demonstrated through its application to six real datasets, including newspaper articles and tweets. The results show improved reproducibility and reliability in topic modeling outcomes. LDAPrototype’s approach is noteworthy for its practical applicability, comprehensibility, ease of implementation, and computational efficiency. Furthermore, the algorithm’s concept can be generalized to other topic modeling procedures that characterize topics through word distributions, making it a versatile tool in text data analysis.
- Open Access
- 20 September 2024
- Published by PeerJ
- Vol. 10, e2279
Figure 2: Sets of tuples consisting of LDA models and their mean LDA similarity values to all other LDAs from the same set.
Latent Dirichlet allocation (LDA) is a popular method for analyzing large text corpora, but it suffers from instability due to its reliance on random initialization. This results in different outcomes for replicated runs, hindering reproducibility. To address this, we introduce LDAPrototype, a new approach for selecting the most representative LDA run from multiple replications on the same dataset. LDAPrototype enhances the reliability of LDA conclusions by ensuring greater similarity between replications compared to traditional LDA runs or models chosen based on perplexity or NPMI. A key feature of LDAPrototype is its use of a novel model similarity measure called S-CLOP (Similarity of multiple sets by Clustering with LOcal Pruning). It is based on topic similarities, for which we compare the usage of measures like the thresholded Jaccard coefficient, cosine similarity, Jensen-Shannon divergence, and rank-biased overlap. The effectiveness of LDAPrototype is demonstrated through its application to six real datasets, including newspaper articles and tweets. The results show improved reproducibility and reliability in topic modeling outcomes. LDAPrototype’s approach is noteworthy for its practical applicability, comprehensibility, ease of implementation, and computational efficiency. Furthermore, the algorithm’s concept can be generalized to other topic modeling procedures that characterize topics through word distributions, making it a versatile tool in text data analysis.
- 20 September 2024
- Published by PeerJ
- Vol. 10, e2279
Open AccessFigure 2: Sets of tuples consisting of LDA models and their mean LDA similarity values to all other LDAs from the same set.
Latent Dirichlet allocation (LDA) is a popular method for analyzing large text corpora, but it suffers from instability due to its reliance on random initialization. This results in different outcomes for replicated runs, hindering reproducibility. To address this, we introduce LDAPrototype, a new approach for selecting the most representative LDA run from multiple replications on the same dataset. LDAPrototype enhances the reliability of LDA conclusions by ensuring greater similarity between replications compared to traditional LDA runs or models chosen based on perplexity or NPMI. A key feature of LDAPrototype is its use of a novel model similarity measure called S-CLOP (Similarity of multiple sets by Clustering with LOcal Pruning). It is based on topic similarities, for which we compare the usage of measures like the thresholded Jaccard coefficient, cosine similarity, Jensen-Shannon divergence, and rank-biased overlap. The effectiveness of LDAPrototype is demonstrated through its application to six real datasets, including newspaper articles and tweets. The results show improved reproducibility and reliability in topic modeling outcomes. LDAPrototype’s approach is noteworthy for its practical applicability, comprehensibility, ease of implementation, and computational efficiency. Furthermore, the algorithm’s concept can be generalized to other topic modeling procedures that characterize topics through word distributions, making it a versatile tool in text data analysis.
- Open Access
- 20 September 2024
- Published by PeerJ
- Vol. 10, e2279
Table 8: Reliability scores according to Figs. 4 and 10.
Latent Dirichlet allocation (LDA) is a popular method for analyzing large text corpora, but it suffers from instability due to its reliance on random initialization. This results in different outcomes for replicated runs, hindering reproducibility. To address this, we introduce LDAPrototype, a new approach for selecting the most representative LDA run from multiple replications on the same dataset. LDAPrototype enhances the reliability of LDA conclusions by ensuring greater similarity between replications compared to traditional LDA runs or models chosen based on perplexity or NPMI. A key feature of LDAPrototype is its use of a novel model similarity measure called S-CLOP (Similarity of multiple sets by Clustering with LOcal Pruning). It is based on topic similarities, for which we compare the usage of measures like the thresholded Jaccard coefficient, cosine similarity, Jensen-Shannon divergence, and rank-biased overlap. The effectiveness of LDAPrototype is demonstrated through its application to six real datasets, including newspaper articles and tweets. The results show improved reproducibility and reliability in topic modeling outcomes. LDAPrototype’s approach is noteworthy for its practical applicability, comprehensibility, ease of implementation, and computational efficiency. Furthermore, the algorithm’s concept can be generalized to other topic modeling procedures that characterize topics through word distributions, making it a versatile tool in text data analysis.
- 20 September 2024
- Published by PeerJ
- Vol. 10, e2279
Open AccessTable 8: Reliability scores according to Figs. 4 and 10.
Latent Dirichlet allocation (LDA) is a popular method for analyzing large text corpora, but it suffers from instability due to its reliance on random initialization. This results in different outcomes for replicated runs, hindering reproducibility. To address this, we introduce LDAPrototype, a new approach for selecting the most representative LDA run from multiple replications on the same dataset. LDAPrototype enhances the reliability of LDA conclusions by ensuring greater similarity between replications compared to traditional LDA runs or models chosen based on perplexity or NPMI. A key feature of LDAPrototype is its use of a novel model similarity measure called S-CLOP (Similarity of multiple sets by Clustering with LOcal Pruning). It is based on topic similarities, for which we compare the usage of measures like the thresholded Jaccard coefficient, cosine similarity, Jensen-Shannon divergence, and rank-biased overlap. The effectiveness of LDAPrototype is demonstrated through its application to six real datasets, including newspaper articles and tweets. The results show improved reproducibility and reliability in topic modeling outcomes. LDAPrototype’s approach is noteworthy for its practical applicability, comprehensibility, ease of implementation, and computational efficiency. Furthermore, the algorithm’s concept can be generalized to other topic modeling procedures that characterize topics through word distributions, making it a versatile tool in text data analysis.