Interrogatives in news discourse have been examined in linguistics and conversation analysis, but mostly in broadcast interviews and relatively small, often English-language corpora, while large-scale computational studies of news rarely distinguish interrogatives from declaratives or differentiate their functions. This paper brings these strands together through a mixed-methods study of the "Politics of Questions" in contemporary French-language digital news. Using over one million articles published between January 2023 and June 2024, we automatically detect interrogative stances, approximate their functional types, and locate textual answers when present, linking these quantitative measures to a qualitatively annotated subcorpus grounded in semantic and pragmatic theories of questions. Interrogatives are sparse but systematically patterned: they mainly introduce or organize issues, with most remaining cases being information-seeking or echo-like, while explicitly leading or tag questions are rare. Although their density and mix vary across outlets and topics, our heuristic suggests that questions are overwhelmingly taken up within the same article and usually linked to a subsequent answer-like span, most often in the journalist's narrative voice and less often through quoted speech. Interrogative contexts are densely populated with named individuals, organizations, and places, whereas publics and broad social groups are mentioned much less frequently, suggesting that interrogative discourse tends to foreground already prominent actors and places and thus exhibits strong personalization. We show how interrogative stance, textual uptake, and voice can be operationalized at corpus scale, and argue that combining computational methods with pragmatic and sociological perspectives can help account for how questioning practices structure contemporary news discourse.
Figures & tables
Figure 1: Overview of the mixed-methods pipeline for interrogative stance analysis. Starting from a corpus of 1.2M French-language news articles from 24 outlets (2023-2024), the upper branch applies a teacher-student NLP pipeline to detect interrogative stances and derive measures of interrogativity, answerability, dialogicity, and actor/addressivity profiles. In parallel, a stratified qualitative subcorpus is manually annotated for stance, form, addressee, answerhood, and macro-functions. The two branches are then triangulated to evaluate the models and interpret corpus-level patterns of prevalence, answerability, voice, and power.
Stance
NQ
% of interrogatives
echo-clarification
70,281
9.2
framing-procedural
396,298
52.1
information-seeking
153,404
20.2
leading
16,221
2.1
rhetorical
107,862
14.2
tag
16,116
2.1
Table 1: Global distribution of predicted interrogative stance types in the corpus.
Figure 2: Variation in interrogative density across (a) regions and (b) editorial scales in French-language news (2023–2024).
Figure 3: Mean answerability and external dialogicity by editorial scale. Values are shares of all interrogative sentences in each scale.
Appendix figures & tables9 assets
Supplementary material from the paper’s appendix.
Appendix
Source
Articles
Country/region
Scale
Type
arcinfo.ch
11,204
Switzerland
Hyper-local
general
footmercato.net
35,579
France
Thematic
sports
fr.allafrica.com
116,680
Africa (pan-African)
Transnational
general
francetvinfo.fr
39,709
France
National
general
gala.fr
32,556
France
Thematic
celebrity
ici.radio-canada.ca
50,976
Canada
National
general
Appendix
Table 2: Outlets in the corpus with total article counts (2023-mid-2024) and outlet ontologies.
Meta-topic
Articles
Mean interrog. index
% Q with ORG
% with LOC/EVENT
Local news
39,662
0.0298
38.3
36.3 / 28.4
Professional sports
181,056
0.0290
51.8
33.4 / 33.4
Lifestyle, entertainment & people
44,533
0.0271
27.8
25.8 / 26.1
Faits divers
29,932
0.0220
37.7
39.6 / 21.5
National / local politics
118,167
0.0214
47.1
35.7 / 23.8
Technology
12,104
0.0184
44.7
17.7 / 20.6
Appendix
Table 3: Meta-topic overview: article volume, interrogative density, and institutional, geographic, and event anchoring of questions (Q).
Stance
NQ
% answered
information-seeking
153,404
97.2
echo-clarification
70,281
97.4
rhetorical
107,862
95.5
leading
16,221
95.4
tag
16,116
95.1
framing-procedural
396,298
94.8
Appendix
Table 4: Answerability and dialogicity of interrogative stances.
Confidence
NQ
% of sentences
Mean IDa
0.6
785,014
3.34
0.0261
0.7
760,182
3.23
0.0252
0.8
714,023
3.03
0.0236
Appendix
Table 5: Sensitivity of interrogative prevalence estimates to the confidence threshold used for the interrogative predictions.
Similarity
answered
internal
via quotes
0.05
95.6%
80.3%
15.4%
0.40
95.6%
80.3%
15.4%
0.80
95.6%
80.3%
15.4%
0.95
94.9%
79.6%
15.3%
0.975
82.5%
69.1%
13.5%
Appendix
Table 6: Sensitivity of answerability and dialogicity estimates to the cosine-similarity threshold used for answer matching.
Pred. answered
Local
National
Total
Clear answer
11 (55%)
14 (70%)
25 (62.5%)
Partial answer
5 (25%)
4 (20%)
9 (22.5%)
Answer elsewhere
1 (5%)
2 (10%)
3 (7.5%)
No genuine answer
3 (15%)
0 (0%)
3 (7.5%)
Appendix
Table 7: Manual spot check of the answer heuristic on 50 randomly sampled questions from 50 different articles.
Binary interrogative detector
Metric
Value
Evaluation sentences
8,399
Accuracy
0.97
Precision (interrogative)
0.76
Recall (interrogative)
0.80
F1 (interrogative)
0.78
Appendix
Table 8: Summary of model performance and inter-annotator agreement.
Stance
Precision
Recall
F1
Support
Fram.-proc.
0.45
0.59
0.51
129
Inform.-seek.
0.80
0.37
0.51
171
Rhetorical
0.60
0.39
0.47
113
Leading
0.68
0.38
0.49
39
Tag
0.59
0.50
0.54
32
Echo-clarif.
0.48
0.62
0.54
32
Appendix
Table 9: Per-class performance of the six-way stance classifier on the gold-standard evaluation set.
Figure 4: Row-normalized conditional confusion matrix for the six-way stance classifier.
Questions scraped from the web are used across academia and industry as a proxy for what people want to know. Across QA training data, retrieval benchmarks, and content strategy, questions on a page are assumed to reflect human intent. We test this assumption at scale by extracting 13.4B question occurrences across 110 FineWeb snapshots (2013-2025), and report three findings. First, you can tell who is asking: provenance (the host/page of questions) leaves a signal in question form, and a logistic model can separate genuine user questions from templated/manufactured ones at AUC 0.725 via length and surrounding context rather than question type, though only 0.554 against commerce FAQ writing. Second, question frequency does not measure demand: the most-frequent questions are boilerplate/templated (over 70% of the top thousand), so occurrence counts measure how often a string was published and not how often it was asked. Third, over twelve years the genuine share of occurrences fell by 79% (42-56% after controlling for crawl composition), with question length and context decreasing. We present the first diachronic, occurrence-level measurement of web question provenance, and find the crawlable web's questions have shifted from being asked by humans toward manufactured for machines to read.
Calvin Zhou, Vincent McCloskey, Krishna Srinivasan
University of Pennsylvania · Bodhium Labs · Work done during an internship at Bodhium Labs.
We present HistoriQA-ThirdRepublic: a French-language dataset of multi-hop historical questions derived from parliamentary debates and newspapers of the French Third Republic. Designed in collaboration with a historian, the corpus captures complex reasoning patterns typical of historical inquiry, including cross-source synthesis, temporal reasoning, and the integration of sparse evidence. The dataset is made of 1782 questions and emphasizes multi-hop connections across heterogeneous historical documents, providing a resource for evaluating retrieval-augmented and large language model systems in domain-specific contexts. We describe the methodology for constructing the corpus, including the selection and alignment of sources, question validation, and metadata integration. While the dataset focuses on French historical documents, our methodology can be readily adapted to other languages and national corpora. Finally, we demonstrate how the corpus can support realistic evaluation scenarios for multi-hop question answering, bridging the gap between NLP benchmarks and the needs of historical scholarship.
News headlines frame public issues both by what they select and by how they word it, yet computational framing work typically collapses these operations into a single score. We introduce a two-dimensional framework that separates salience framing, measured through four wording devices (loaded vocabulary, blame attribution, threat framing, rhetorical question), from selection framing, measured through outlet-level story-form and high-charge distributions. We build a 10,000-headline French supervision set using three LLM annotators with majority-vote resolution and human arbitration, validate the labels against two annotator-independent blind human studies, and apply the strongest classifier to 902,111 deduplicated headlines from 25 French outlets (2022-2025). Three main findings emerge. First, salience and selection divergence are positively correlated yet leave nearly half of outlet-level variance unexplained, populating interpretively distinct off-diagonal cells in a four-cell outlet typology. Second, default classification thresholds systematically inflate corpus-level salience estimates; a precision-floor recalibration protocol corrects this distortion. Third, group-mention analysis reveals sharply unequal salience contexts: headlines mentioning Jews, the Far-right, and Muslims carry the highest detected salience rates, which broad event-context composition does not fully explain (residuals are descriptive, not same-event causal estimates; per-group lexicon precision is reported alongside). To our knowledge, this is the largest framing-focused French headline audit to date; we release the supervision set, lexicons, and analysis code.