| {"cut":0.41500412441215667,"best_f1_cut":0.5849553369198143,"weights":{"classifier":0.9306403554233327,"basal":0.9179535217920844,"judge":0.8554876739272002},"counts":{"terms":62066,"dropped":5328,"definitions":103476,"concepts":34916,"papers":6115,"seeds":{"terms":35807,"papers":3715},"closure":{"terms":26259,"papers":2400}},"shard":1500,"concepts":{"concepts":34916,"single_term_concepts":28210},"topics":[{"id":0,"name":"social media · social bots · memes","size":3330,"top":["social media","social bots","memes","hashtags"]},{"id":1,"name":"metaphor · conceptual metaphors · presupposition","size":3276,"top":["metaphor","conceptual metaphors","presupposition","modality"]},{"id":2,"name":"populism · polarization · public sphere","size":2872,"top":["populism","polarization","public sphere","affective polarization"]},{"id":3,"name":"ideology · neoliberalism · nationalism","size":2817,"top":["ideology","neoliberalism","nationalism","hegemony"]},{"id":4,"name":"framing · frames · framing theory","size":2735,"top":["framing","frames","framing theory","media framing"]},{"id":5,"name":"heuristics · Elaboration Likelihood Model (ELM) · empathy","size":2605,"top":["heuristics","Elaboration Likelihood Model (ELM)","empathy","need for cognition"]},{"id":6,"name":"journalism · gatekeepers · peace journalism","size":2441,"top":["journalism","gatekeepers","peace journalism","gatekeeping"]},{"id":7,"name":"homophily · stereotypes · identity","size":2416,"top":["homophily","stereotypes","identity","social influence"]},{"id":8,"name":"othering · representational harms · discrimination","size":2382,"top":["othering","representational harms","discrimination","racism"]},{"id":9,"name":"media · mediatization · media logic","size":2248,"top":["media","mediatization","media logic","mass media"]},{"id":10,"name":"cognitive dissonance · cognitive dissonance theory · third-person effect","size":2070,"top":["cognitive dissonance","cognitive dissonance theory","third-person effect","vaccine hesitancy"]},{"id":11,"name":"news · news values · hard news","size":2028,"top":["news","news values","hard news","soft news"]},{"id":12,"name":"morality · ethics · freedom of expression","size":1979,"top":["morality","ethics","freedom of expression","agency"]},{"id":13,"name":"clickbait · satire · sarcasm","size":1967,"top":["clickbait","satire","sarcasm","whataboutism"]},{"id":14,"name":"misinformation · disinformation · infodemic","size":1804,"top":["misinformation","disinformation","infodemic","malinformation"]},{"id":15,"name":"credibility · trust · source credibility","size":1769,"top":["credibility","trust","source credibility","trustworthiness"]},{"id":16,"name":"hate speech · cyberbullying · toxicity","size":1641,"top":["hate speech","cyberbullying","toxicity","offensive language"]},{"id":17,"name":"propaganda · computational propaganda · astroturfing","size":1624,"top":["propaganda","computational propaganda","astroturfing","manipulation"]},{"id":18,"name":"fact-checking · fake news detection · rumor detection","size":1603,"top":["fact-checking","fake news detection","rumor detection","fact-checkers"]},{"id":19,"name":"discourse · Critical Discourse Analysis (CDA) · discourse analysis","size":1524,"top":["discourse","Critical Discourse Analysis (CDA)","discourse analysis","content analysis"]},{"id":20,"name":"infodemic · information · information overload","size":1451,"top":["infodemic","information","information overload","transparency"]},{"id":21,"name":"narrative · event · storytelling","size":1371,"top":["narrative","event","storytelling","story"]},{"id":22,"name":"bias · gender bias · implicit bias","size":1349,"top":["bias","gender bias","implicit bias","algorithmic bias"]},{"id":23,"name":"deepfakes · hallucinations · cheapfakes","size":1330,"top":["deepfakes","hallucinations","cheapfakes","manipulated content"]},{"id":24,"name":"fake news · hoaxes · false news","size":1255,"top":["fake news","hoaxes","false news","clickbait"]},{"id":25,"name":"sentiment analysis · sentiment · attitude","size":1217,"top":["sentiment analysis","sentiment","attitude","affect"]},{"id":26,"name":"gender · intersectionality · gender stereotypes","size":1193,"top":["gender","intersectionality","gender stereotypes","hegemonic masculinity"]},{"id":27,"name":"rumor · conspiracy theories · post-truth","size":1190,"top":["rumor","conspiracy theories","post-truth","conspiracy"]},{"id":28,"name":"prebunking · backfire effect · inoculation theory","size":1117,"top":["prebunking","backfire effect","inoculation theory","illusory truth effect"]},{"id":29,"name":"fairness · individual fairness · group fairness","size":1000,"top":["fairness","individual fairness","group fairness","Counterfactual Data Augmentation (CDA)"]},{"id":30,"name":"confirmation bias · selective exposure · motivated reasoning","size":874,"top":["confirmation bias","selective exposure","motivated reasoning","biased assimilation"]},{"id":31,"name":"stance detection · stance · stance classification","size":871,"top":["stance detection","stance","stance classification","opinion"]},{"id":32,"name":"agenda setting · agenda-setting theory · second-level agenda-setting","size":776,"top":["agenda setting","agenda-setting theory","second-level agenda-setting","intermedia agenda-setting"]},{"id":33,"name":"media bias · political bias · bias","size":718,"top":["media bias","political bias","bias","informational bias"]},{"id":34,"name":"media literacy · information literacy · media and information literacy","size":676,"top":["media literacy","information literacy","media and information literacy","digital literacy"]},{"id":35,"name":"echo chambers · filter bubbles · echo chamber effect","size":547,"top":["echo chambers","filter bubbles","echo chamber effect","homophily"]}],"baseline":{"random_pair_cos_quantiles":{"0.01":0.1453825980424881,"0.05":0.17864347994327545,"0.25":0.23080956935882568,"0.5":0.27249979972839355,"0.75":0.3213024139404297,"0.95":0.413118839263916,"0.99":0.5104117393493652},"random_pair_cos_mean":0.2815283536911011,"nearest_neighbour_cos_quantiles":{"0.01":0.5605762004852295,"0.05":0.6157150268554688,"0.25":0.7185686826705933,"0.5":0.8100404739379883,"0.75":0.8959345817565918,"0.95":0.9714548587799072,"0.99":0.99299556016922},"nearest_neighbour_cos_mean":0.8041006326675415},"method":{"vector":"the L2-normalised mean of the Qwen3-Embedding-8B vectors of a mention's definitions, each embedded as \"term: definition\"","umap":"a UMAP of the vectors of the released mentions (PCA to 50 dimensions, then UMAP with cosine distance, 30 neighbours and minimum distance 0.15, to 3D and 2D); 36 topics by k-means (k = 36) on the 50 PCA dimensions, named by their most frequent concepts","spread":"the mean cosine of a label group's mentions to their normalised centroid; the mean pairwise cosine","neighbours":"the 8 most similar other mentions, by exact cosine on the full vectors","rule":{"name":"the label-match rule","source":"src/mbdef/postprocess/label_match.py"},"concept_neighbours":"the 8 most similar kept concepts, by the cosine of the normalised mean of their kept mentions' vectors"},"fields":{"codes":{"description":"de","synonyms":"sy","related_terms":"rt","subtypes":"sb","supertypes":"sp","examples":"ex"},"names":["sy","rt","sb","sp"],"counts":{"de":[41762,21314],"sy":[6434,4495],"rt":[21269,12277],"sb":[12682,3941],"sp":[9105,8011],"ex":[10190,5095]},"terms_shard":2000,"relations":27695,"note":"items as the dataset reads them; the merged view is the dataset's terms table: names merged by the label-match rule and linked to the concept they name, sentences merged when they read the same once case, punctuation and spacing are set aside"}} |