曝光臺(tái) 注意防騙
網(wǎng)曝天貓店富美金盛家居專(zhuān)營(yíng)店坑蒙拐騙欺詐消費(fèi)者
Empire (78) Roman Empire, British Empire, Ottoman Empire, Byzantine Empire, German Empire, Mughal Empire
Flower (59) Rose, Lotus, Iris, Lily, Violet, Daisy, Lavender, Magnolia, Tulip, Orchid
Holiday (82) Christmas, Halloween, Easter, Thanksgiving, Labor Day, Independence Day, Lent, Yule, Hanukkah
Hurricane (74) Hurricane Katrina, Hurricane Ivan, Hurricane Dennis, Hurricane Wilma, Hurricane Frances
Mountain (245) K2, Everest, Mont Blanc, Table Mountain, Etna, Mount Fuji, Mount Hood, Annapurna, Kilimanjaro
Movie (626) Star Wars, Die Hard, Air Force One, The Matrix, Lord of the Rings, Lost in Translation, Oce Space
NationalPark (59) Great Smoky Mountains National Park, Grand Canyon National Park, Joshua Tree National Park
NbaTeam (30) Utah Jazz, Sacramento Kings, Chicago Bulls, Milwaukee Bucks, San Antonio Spurs, New Jersey Nets
Newspaper (599) New York Times, Le Monde, Washington Post, The Independent, Die Welt, Wall Street Journal
Painter (1011) Leonardo da Vinci, Rembrandt, Andy Warhol, Vincent van Gogh, Marcel Duchamp, Frida Kahlo
ProgLanguage (101) A++, PHP, C, C++, BASIC, JavaScript, Java, Forth, Perl, Ada
Religion (128) Islam, Christianity, Voodoo, Buddhism, Judaism, Baptism, Taoism, Confucianism, Hinduism, Wicca
River (167) Nile, Mississippi River, Hudson River, Colorado River, Danube, Amazon River, Volga, Snake River
SearchEngine (25) Google, Lycos, Excite, AltaVista, Baidu, HotBot, Dogpile, WebCrawler, AlltheWeb, Clusty
SkyBody (97) Earth, Mercury, Saturn, Vega, Sirius, Polaris, Pluto, Uranus, Antares, Canopus
Skyscraper (172) Empire State Building, Sears Tower, Chrysler Building, Taipei 101, Burj Al Arab, Chase Tower
SoccerClub (116) Chelsea, Real Madrid, Juventus, FC Barcelona, AC Milan, Aston Villa, Real Sociedad, Bayern Munich
SportEvent (143) Tour de France, Super Bowl, US Open, Champions League, Bundesliga, Stanley Cup, FA Cup
Stadium (190) Stade de France, Olympic Stadium, Wembley Stadium, Soldier Field, Old Tra ord, Camp Nou
TerroristGroup (74) Hezbollah, Khmer Rouge, Irish Republican Army, Shining Path, Tupac Amaru, Sendero Luminoso
Treaty (202) North Atlantic Treaty, Kyoto Protocol, Louisiana Purchase, Montreal Protocol, Berne Convention
University (501) University of Oslo, Stanford, CMU, Columbia University, Tsing Hua University, Cornell University
VideoGame (450) Half Life, Final Fantasy, Grand Theft Auto, Warcraft, Need for Speed, Metal Gear, Gran Turismo
Wine (60) Port, Champagne, Bordeaux, Rioja, Chardonnay, Merlot, Chianti, Cabernet Sauvignon, Pinot Noir
WorldWarBattle (127) D-Day, Battle of Britain, Battle of the Bulge, Battle of Midway, Battle of the Somme, Battle of Crete
Table 1: Target classes with sizes of instance sets and examples of instances
most frequently in text processing and described in [6]: Co-
sine (the ubiquitous dot product); Jaccard (Jaccard's coef-
cient); Jensen-Shannon (the Jensen-Shannon divergence);
L1-Norm; and Skew-Divergence.
Evaluation Procedure: Multiple lists of attributes are
evaluated for each class, corresponding to the combination
of the use of a particular vector similarity function with a
particular choice of number of input instances per class, in-
put seed attributes per class, and other system settings. To
remove any undesirable psychological bias towards higher-
ranked attributes during the assessment, the elements of
each list to be evaluated are sorted alphabetically into a
merged list. Each attribute of the merged list is manually
assigned a correctness label within its respective class. Sim-
ilarly to methodology previously proposed to evaluate an-
swers to De nition questions [21], an attribute is vital if it
must be present in an ideal list of attributes of the target
class; okay if it provides useful but non-essential informa-
tion; and wrong if it is incorrect. Thus, a correctness label
is manually assigned to a total of 18,608 attributes extracted
for the 40 target classes, in a process that once again con-
rms that evaluation of information extraction methods can
be quite time consuming.
Label Value Examples of Attributes
vital 1.0 ProgLanguage: portability, Wine: taste
okay 0.5 Company: vision, NationalPark: reptiles
wrong 0.0 BasicFood: low carb, CarModel: driver
Table 2: Labels for assessing attribute correctness
To compute the overall precision score over a ranked list of
extracted attributes, the correctness labels are converted to
numeric values as shown in Table 2. Precision at some rank
N in the list is thus measured as the sum of the assigned
values of the rst N candidate attributes, divided by N.
3.2 Quality of the Extracted Attributes
Figure 4 plots precision values for ranks 1 through 50,
for each of the ve similarity functions (Cosine vs. Jaccard
中國(guó)航空網(wǎng) www.k6050.com
航空翻譯 www.aviation.cn
本文鏈接地址：
航空資料36(19)
