NEWTS: A Corpus for News Topic-Focused Summarization

Text summarization models are approaching human levels of fidelity. Existing benchmarking corpora provide concordant pairs of full and abridged versions of Web, news or, professional content. To date, all summarization datasets operate under a one-size-fits-all paradigm that may not reflect the full range of organic summarization needs. Several recently proposed models (e.g., plug and play language models) have the capacity to condition the generated summaries on a desired range of themes. These capacities remain largely unused and unevaluated as there is no dedicated dataset that would support the task of topic-focused summarization.

This paper introduces the first topical summarization corpus NEWTS, based on the wellknown CNN/Dailymail dataset, and annotated via online crowd-sourcing. Each source article is paired with two reference summaries, each focusing on a different theme of the source document. We evaluate a representative range of existing techniques and analyze the effectiveness of different prompting methods.

Topic Id Topic Words 62 island, beach, sea, gaal, navy 32 water, river, lake, bridge, walker 78 court, judge, case, appeal, justice 46 law, legal, state, marriage, rights 12 islamic, terror, terrorist, al, threat 229 hotel, guests, bar, glass, wine 105 charged, allegedly, charges, arrested, alleged 72 health, virus, cases, people, bird 153 fire, residents, san, wood, firefighters 97 visit, pope, peace, catholic, roman 134 air, plane, aircraft, flight, flying 13 price, cost, products, market, prices 187 website, disease, spread, ill, contact 152 united, manchester, liverpool, chelsea, league 195 court, trial, guilty, prison, heard 64 group, forces, fighters, killed, fighting 113 campaign, clinton, governor, presidential 163 airport, passengers, flight, travel, airlines 162 president, obama, white, house, barack 199 cup, real, madrid, brazil, ronaldo 129 attack, attacks, killed, attacked, bomb 175 house, committee, congress, senate, republican 211 london, british, uk, britain, royal 227 music, singer, song, band, bruce 194 russian, russia, european, europe, ukraine 217 club, team, season, players, england 61 match, murray, won, title, round 90 arsenal, ball, alex, wenger, villa 115 family, wife, daughter, husband, couple 236 film, movie, character, films, viewers 89 weight, pounds, fat, diet, body 39 war, military, defence, army, iraq 180 goal, win, side, scored, minutes 247 tax, average, benefits, people, rate 110 billion, figures, economy, global, growth 85 coast, miles, storm, east, map 196 school, schools, teacher, high, education 248 hospital, medical, doctors, patients, care 205 art, museum, display, century, history, 83 road, driver, driving, traffic, speed 48 food, restaurant, eat, eating, babies 144 online, users, internet, site, device 100 earth, sun, climate, planet, change 200 children, child, parents, birth, born 198 study, researchers, google, scientists, university 245 facebook, mobile, phone, network, samsung 128 money, pay, paid, card, credit 55 energy, power, heat, plant, fuel 101 crown, grand, race, hamilton, team 218 snow, weather, cold, winter, temperatures