APPENDIX A:
EVALUATION TASK DESCRIPTION
1 .0 WHAT CONSTITUTES A "RELEVANT" INCIDENT ?
Relevant incidents are, in general, violent acts perpetrated with political aim s
and a motive of intimidation . These are acts of terrorism . Here is an official ,
"schoolbook" definition of terrorism :
"a sudden, unexpected act of shocking, calculated, and unlawful violence, o r
the plausible threat of such violence, by an illegal, subnational, clandestin e
group -- usually carried out in a peaceful, civilian environment, be i t
urban, rural, in the air or on a body of water -- against certai n
noncombatants or targets that represent or symbolize a certain country, bu t
sometimes indiscrimately against bystanders or passersby at a particula r
location, with the intention of garnering publicity, propagandizing a cause ,
and intimating as many people as possible in order to attain social, political ,
or strategic objectives . "
-Rex A . Hudson, "Dealing with Internationa l
Hostage-Taking : Alternatives to Reactive
Counterterrorist Assaults," Internationa l
Terrorism, Vol . 12, 1990.
This has been turned into more of an operational definition by widening th e
schoolbook definition somewhat and making some rather arbitrary distinctions .
Your advice on how to improve on this operational definition is welcome .
1)
	
Terrorist acts may be perpetrated by an "illegal, subnational, clandestin e
group," which includes known guerrilla and drug-traffickin g
organizations . Their targets may be just about anything/anybody except (a )
another such group or member of such a group or (b) a military/polic e
installation or force (with the same nationality as the perpetrator), or a
member of such a force, in which case the incident is presumed to be a
purely guerrilla act and is not relevant to the MUC-3 terrorist inciden t
database . Thus, killings of drug traffickers will not be included in th e
database (by exception a, above), nor will "clashes" between guerrill a
groups and the military (by exception b, above) .
However, if a guerrilla warfare incident happens to affect civilia n
personnel or property, whether intentionally or accidentally, the inciden t
becomes relevant and should be included in the database . In these cases, th e
database should contain information on both the military and the
nonmilitary targets.
Similarly, the database will include incidents of sabotage on a nation' s
infrastructure. It will also include incidents perpetrated against individual s
who are former members of the military, e .g ., a murder perpetrated against
a retired general .
A—1
Sometimes an article gives mixed reports, e .g., a "terrorist attack" on a
bridge guarded by the national (military) police . In a case such as this ,
which could be interpreted as a case of terrorism (sabotage of the nation' s
infrastructure) or of guerrilla warfare (attack on a physical target o f
military importance), we will consider the physical target to be nonmilitar y
and will include the incident in the database .
2) Under certain circumstances, the perpetrator may be a member of th e
government, including the military . This is the case when the target i s
civilian .
3) Sometimes the perpetrator is not identified and no language is used t o
indicate that the act is *not* being attributed to "terrorists," "extremists, "
"subversives," "guerrillas," etc . In these cases, as long as the act itself is o f
one of the expected types and the target(s) is/are not exclusively military ,
the incident will be presumed to be a possible terrorist act and is to b e
included in the database .
4) The articles in the MUC-3 corpus originate in many different parts of th e
world, but they all emphasize something about one of the nine Lati n
American countries of interest (El Salvador, Guatemala, Honduras ,
Argentina, Bolivia, Chile, Colombia, Ecuador, Peru) . However, only incident s
whose perpetrator(s) or target(s) "belong to" one of the nine countries o f
interest belong in the MUC-3 database . Thus, an incident perpetrated agains t
a human target who is a citizen of one of those countries may belong in th e
database even if that person is attacked while he's in a different country . Or
an incident perpetrated against the embassy of one of those nine countrie s
may belong in the database, no matter where that embassy is located .
5) Only terrorist acts (and attempted acts) that are reported as factual are to b e
included in the database. An incident is *not* to be recorded in the database
if the article discredits the story or cites a source that discredits the story .
(However, if the article also cites a source that *supports* the story, tha t
version of the story should go in the database .) Rumors are to be excluded .
6) An incident is to be recorded in the database only if it is "recent ." The intent
is to include only information that is new or updated . We define "recent" t o
mean that either (a) it occurred within two months of the date of the articl e
or (b) new information is being provided in the article on an incident tha t
occured more than two months prior, such as new suspects being brough t
forth as the perpetrators of an old crime . However, there are a number of
articles that fail to mention the date of the incidents, some of which are, i n
fact, recent. Therefore, if there is no indication that the information is no t
recent or if there is no indication that the information is being given onl y
to provide historical context (as indicated by phrases such as "Let us recal l
that.. ."), the incident should go in the database .
7) An incident is to be recorded in the database only if the article's descriptio n
of it is "specific ." Generic descriptions of incidents like "terrorist attacks o n
the Cano Limon-Covenas oil pipeline" and summary descriptions like "th e
total persons kidnapped in Colombia over the last 6 weeks was 85" are to b e
excluded .
A—2
The articles sometimes give useful information about a specific incident bu t
provide very few details . Therefore, as long as all the other criteria are met, w e
will include any incident that at least indicates the nature of the target or th e
nature of the attack . Thus, an article whose only reference to an incident is "a n
attack perpetrated by a subversive group" or "the tragic incident at Uchiza" woul d
be excluded, but an article that says "Last night there were fewer attacks on store s
and electric towers in the capital" or "a UCR district headquarters in Buenos Aires
province was completely destroyed by a bomb explosion" would be included .
Here are two further examples of texts that do *not* warrant generation o f
templates : "As of 0500 GMT today, the police had received reports of two othe r
explosions in two La Paz neighborhoods" (the nature of the explosion isn' t
mentioned, and the target isn't mentioned either) ; "the bombing on San Salvador' s
neighborhoods" (the prepositional phrase may be interpreted as conveying the
target as well as the location, but as a target, it's too general -- any location name
that is on the set list cannot serve as a target ID, with the exception of farm names) .
2.0 DEFINITIONS OF SLOT TYPE S
0 . MESSAGE ID -- The first line of the message, e .g., DEV-MUC3-0001 (NOSC).
This slot serves as an index and is not scored in its own right .
1. TEMPLATE ID -- A number that distinguishes the templates for a give n
message . On the answer sheet, the word OPTIONAL in parentheses afte r
the template number indicates that there is significant doubt whether th e
incident belongs in the database .
2. DATE OF INCIDENT -- The date of incident (according to local time, no t
Greenwich Mean Time) .
3. TYPE OF INCIDENT -- A terrorist act reported on in the message .
4. CATEGORY OF INCIDENT -- The subcategory of terrorism that the inciden t
fits into .
5. PERPETRATOR : ID OF INDIV(S) -- The person(s) responsible for th e
incident .
6. PERPETRATOR : ID OF ORG(S) --The organization responsible for th e
inciden t
7. PERPETRATOR : CONFIDENCE
	
The way the perpetrator organization i s
viewed in the message .
8. PHYSICAL TARGET: ID(S) -- The thing(s) (inanimate object) that was/wer e
attacked .
9. PHYSICAL TARGET : TOTAL NUM -- The total number of physical targets .
10. PHYSICAL TARGET : TYPE(S) -- The category that the physical target fit s
into .
11. HUMAN TARGET : ID(S) -- The person(s) who was/were the obvious or
apparent target(s) of the attack or who became victim(s) of the attack .
12. HUMAN TARGET: TOTAL NUM -- The total number of human targets .
13. HUMAN TARGET : TYPE(S) -- The category that the human target fits into .
14. TARGET : FOREIGN NATION(S) -- The nationality of the interests that were
targeted, if identified in the article and if different from country wher e
incident occurred .
15. INSTRUMENT : TYPE(S) -- The device used by the perpetrator(s) i n
carrying out the terrorist act .
16. LOCATION OF INCIDENT -- The name of the country and specifi c
geographic location where the incident occurred .
A—3
17. EFFECT ON PHYSICAL TARGET(S) --The impact of the incident on th e
physical target(s) .
18. EFFECT ON HUMAN TARGET(S) -- The impact of the incident on the huma n
target(s) .
3.0 DEFINITIONS OF FILL TYPES
Set fill : An item from a finite list of possibilities enumerated in th e
documentation provided to participants .
Numeric : A number.
Text conversion : A canonical form extracted or derived from the text . The
documentation provided to participants specifies the expected format .
Text string : A string extracted from the text and placed inside double quote s
in the slot . ,Note that this string may contain misspellings, if found in the original
text.
Null : A hyphen (-) indicating that the system has no response .
4.0 GENERAL REQUIREMENTS FOR FILLS
4.1 FORMAT FOR SYSTEM RESPONSES
System responses should look like plain text, e .g., they cannot be enclosed i n
parentheses, multi-word fillers cannot be connected by underscore characters, etc .
Systems are not permitted to use the slash notation that is used in the key t o
separate alternative forms of a single filler.
When multiple fillers for a single slot are generated, each should appear on a
separate line, arranged in a column, with space padding in front (no tabs) .
Some slots require that their fillers be cross-referenced to the referent (i .e., the
filler of some other slot) . To match the answer sheet, this should be done b y
appending the referent, separating the two data items with a colon .
Example fill for the EFFECT ON HUMAN TARGET slot, whose fillers must b e
cross-referenced to the corresponding fillers in the HUMAN TARGET : ID(S) slot.
(Note that the persons' roles have been omitted from the cross-reference .):
HUMAN TARGET ID(S)
	
"HORACIO FERNANDEZ COTIELLO" ("LAWYER" )
"JORGE SAMUDIO" ("JUDGE" )
EFFECT ON HUMAN TARGET(S) DEATH : "HORACIO FERNANDEZ COTIELLO "
INJURY : "JORGE SAMUDIO "
One template form serves for all incident types . However, depending on th e
incident type, one or more slots do not apply . These slots should appear with a n
asterisk as the fill, and they will not figure in the scoring . For irrelevant articles ,
all slots except the MESSAGE ID slot should have an asterisk .
A—4
4.2 SPECIAL FORMATS FOR ANSWER KEY S
All fills in the answer sheet will be in upper case . If there is more than on e
"correct" answer for any given slot filler, the options will all appear on one line i n
the answer sheet, separated by a slash . The word OPTIONAL may appear i n
parentheses after the filler for the TEMPLATE ID, indicating that generation of thi s
template is optional . A question mark preceding a filler indicates that generation
of this filler is optional, i .e ., that null (hyphen) is also a correct system response .
5 .0 GENERAL GUIDELINES FOR GENERATING MORE THAN ONE TEMPLAT E
PER ARTICL E
If an article discusses more than one relevant type of terrorist incident, eac h
such incident should be captured in a separate template . If an article discusse s
more than one instance of the same relevant type of terrorist incident, each suc h
incident should be captured in a separate template . A "separate instance" is on e
which has a different filler for the location, date, category, or perpetrator slot . If
no distinctions of these kinds are made in the text, the instances should b e
incorporated into a single template, e .g ., if the article describes "attacks" b y
"rebels" on "ten power stations" and does not give varying information on th e
location of those targets, etc . Note that the level of granularity is defined by wha t
can go in a slot ; thus, an article that describes two bombings on targets on
different streets in the same neighborhood should be captured in one templat e
rather than two, since the location slot cannot have a filler that is at the level o f
granularity of a street .
6.0 DATA TYPES FOR FILLER S
0 .
	
MESSAGE ID
1. TEMPLATE ID
2. DATE OF INCIDENT
3. TYPE OF INCIDENT
4. CATEGORY OF INCIDEN T
5. PERPETRATOR: ID OF INDIV(S)
6. PERPETRATOR : ID OF ORG(S )
7. PERPETRATOR : CONFIDENCE
8. PHYSICAL TARGET : ID(S)
9. PHYSICAL TARGET : TOTAL NU M
10. PHYSICAL TARGET : TYPE(S )
11
	
HUMAN TARGET : ID(S)
12. HUMAN TARGET : TOTAL NUM
13. HUMAN TARGET : TYPE(S )
14. TARGET: FOREIGN NATION(S )
15. INSTRUMENT : TYPE(S )
16. LOCATION OF INCIDENT
17. EFFECT ON PHYSICAL TARGET(S )
18. EFFECT ON HUMAN TARGET(S)
Set fill (alphanumeric)
Numeric (integer)
Text conversion (alphanumeric) or nul l
Set fill (alpha)
Set fill (alpha) or nul l
Text string (alphanumeric) or nul l
Text string (alphanumeric) or nul l
Set fill (alpha) or nul l
Text string (alphanumeric) or nul l
Text conversion (alphanumeric) or nul l
Set fill (alpha) or nul l
Text string (alphanumeric) or nul l
Text conversion (alphanumeric) or nul l
Set fill (alpha) or nul l
Set fill (alpha) or nul l
Set fill (alpha) or nul l
Set fill (alpha)
Set fill (alpha) or nul l
Set fill (alpha) or nul l
A—5
7 .0 SPECIFICATION OF SLOT FILLERS
7.1 SLOT 0 -- MESSAGE ID
Number of fillers : Always one filler per template .
General method for filling : To be copied by the system from first line of th e
article .
Set list : This list is the set of all message IDs in the corpus .
Note : If an article is irrelevant, this is the only slot that should be filled in .
7.2 SLOT 1 -- TEMPLATE I D
Number of fillers : Always one filler per templat e
General method for filling : A number (in digits) generated by the system ,
starting with 1 and incremented for each new template .
Notes :
1) Ordering need not match answer sheet -- the scoring program will mak e
best match.
2) On the answer sheet, the word OPTIONAL in parentheses after the templat e
number indicates that there is significant doubt whether the incident
belongs in the database .
7.3 SLOT 2 -- DATE OF INCIDENT
Number of fillers : Always one filler per template, unless NO DATE appears in th e
message header and a date cannot be determined from the body of the article, i n
which case NULL (hyphen) is appropriate .
General method for filling : A form generated by the system from the text ,
expressed in one of the following forms (whichever is appropriate for the article) :
1)
	
DD MMM YY (e .g. 21 JAN 89, meaning "on 21 JAN 89")
2). DD MMM YY - DD MMM YY (e .g. 15 JAN 89 - 21 JAN 89, meaning "on or
between 15 JAN 89 and 21 JAN 89" )
3). - DD MMM YY (e .g. - 21 JAN 89, meaning "on or before 21 JAN 89" )
Notes :
1) If the article contains the name of a day of the week, e .g. "last Wednesday" ,
rather than a reference to the date, use form 3, above, with the best know n
absolute date as the anchor, e .g., the date of the article.
2) If the article uses an easily calculable expression such as "yesterday," us e
form 1, above, with the date calculated from the best known absolute date a s
A—6
the anchor, e.g., the date of the article . An "easily calculable" expression is
one that indicates how many days to add or subtract from the anchor .
3) Whenever the date for an even t
message header,' the answer key
as the anchor.
7 .4 SLOT 3 -- TYPE OF INCIDEN T
Number of fillers : Always one filler per templat e
Set list :
ARSON ATTEMPTED ARSON ARSON THREAT
MURDER ATTEMPTED MURDER DEATH THREAT
BOMBING ATTEMPTED BOMBING BOMB THREAT
KIDNAPPING ATTEMPTED KIDNAPPING KIDNAPPING THREAT
HIJACKING ATTEMPTED HIJACKING HIJACKING THREAT
ROBBERY ATTEMPTED ROBBERY ROBBERY THREAT
FORCED WORK ATTEMPTED FORCED WORK FORCED WORK STOPPAGE
STOPPAGE STOPPAGE THREAT
ATTACK ATTEMPTED ATTACK ATTACK THREAT
Notes :
1)
	
If a non-MURDER incident results in the death of a target (e .g. via a
BOMBING), a MURDER template should be generated as well .
2) The ATTACK incident type should be used only when a terrorist inciden t
clearly does not fall into one of the other categories . Thus, a MURDER
template should be generated *instead of* an ATTACK template if the onl y
target is a person who dies as a result of the incident, even if the incident i s
described as an "attack . "
3)
	
If a bomb is planted (by possible terrorists) but is detonated intentionally b y
authorities, an ATTEMPTED BOMBING template should be generated .
7.5 SLOT 4 -- CATEGORY OF INCIDEN T
Applicability : Applicable to all types of incident s
Number of fillers : One filler per template, or NULL (hyphen) if nature o f
perpetrator cannot be determined .
Set list :
TERRORIST ACT
STATE-SPONSORED VIOLENC E
Notes :
	
1)
	
TERRORIST ACT should be used when the perpetrator is *not* a member o f
the
	
government/military/police .
is derived by default from the date in in th e
will use form 3, above, with the header dat e
A— 7
2)
	
The subcategory of STATE-SPONSORED VIOLENCE should be used when th e
perpetrator is a member of the government/military/police .
3) This slot may be filled even if the perpetrators are not identified, if the tex t
gives strong evidence for the category without providing any strings to fil l
the perpetrator slots; however, it will generally be the case that this slot i s
null if the perpetrator slots are null .
7.6 SLOT 5 -- PERPETRATOR : ID OF INDIV(S )
Applicability : Applicable to all types of incident s
Number of fillers : Null, or one or more fillers
7.7 SLOT 6 -- PERPETRATOR : ID OF ORG(S )
Applicability : Applicable to all types of incident s
Number of fillers : Null, or one or more fillers
Note : Include only if mentioned in the text, Le ., do not use information from the
system's general lexicon to fill this in (such as information linking known
terrorists to organizations) .
7.8 SLOT 7 -- PERPETRATOR : CONFIDENC E
Applicability : Applicable to all types of incident s
Number of fillers : Null if PERPETRATOR : ID OF ORG(S) slot is null, or one fille r
per filler in the PERPETRATOR : ID OF ORG(S) slot .
Cross-referencing : Include cross-reference to PERPETRATOR : ID OF ORG(S) .
Set list:
REPORTED AS FACT
ACQUITTED
CLAIMED OR ADMITTED
SUSPECTED OR ACCUSE D
SUSPECTED OR ACCUSED BY AUTHORITIES
POSSIBLE
Notes :
1)
	
Use REPORTED AS FACT as a default if the identity of the perpetrator org an d
its responsibility for the incident appear not to be in question .
2) Use SUSPECTED OR ACCUSED if the article explicitly expresses some degree o f
uncertainty about the identity of the perpetrator org, e .g ., by use of
qualifying words. This fill should also be used if the article's reporting rest s
on suspicions or accusations expressed by some source cited by the article' s
originator or by the article's originator himself .
A—8
3) Use SUSPECTED OR ACCUSED BY AUTHORITIES instead of SUSPECTED O R
ACCUSED for cases where the article is conveying the information that th e
government, military, police, or other authority figure is expressing th e
suspicion/accusation or is accepting/admitting guilt on behalf of anothe r
authority that has been identified as the perpetrator org . (In the latter case ,
where the perpetrator is himself an authority figure, use CLAIMED O R
ADMITTED if he admits/claims his own org's guilt .)
4) Use POSSIBLE if an article mentions an organization in the context of th e
incident without actually accusing that organization of being responsibl e
for it.
7.9 SLOT 8
	
PHYSICAL TARGET :--
	
ID(S)
Applicability :
	
Not
	
applicable
	
to
	
(ATTEMPTED)
	
MURDER,
	
(ATTEMPTED )
KIDNAPPING, DEATH THREAT incidents .
Number
	
of
	
fillers :
	
Null,
	
or one
	
or more
	
fillers.
Note :
	
Use
	
this
	
slot
	
for
	
the
	
entities
	
that
	
are
	
the
	
targets of incidents
	
(facilities ,
buildings,
	
vehicles,
	
etc .),
	
not
	
for
	
whatever
	
parts
	
happened to have been
	
damaged
as
	
a
	
result
	
of
	
the
	
incident
	
(e.g .,
	
"windows")
	
unless
	
the text doesn't
	
mention
	
the
entity
	
(target)
	
that
	
they
	
are
	
part
	
of.
7.10 SLOT 9 -- PHYSICAL TARGET : TOTAL NUM
Applicability : Not applicable to (ATTEMPTED) MURDER, (ATTEMPTED )
KIDNAPPING, DEATH THREAT incidents .
Number of fillers : Null if ID null, else one filler .
Set list:
PLURAL
Note : Use PLURAL if the only information available is that there was more tha n
one physical target . Otherwise, fill this slot with a digit generated by the syste m
from the info available in the text . If the text indicates a range of values lik e
"between 6 and 8" or "at least 6," use a hyphen as an operator, making the fills "6 -
8" and "6-", respectively .
7.11 SLOT 10 -- PHYSICAL TARGET : TYPE(S )
Applicability : Not applicable to (ATTEMPTED) MURDER, (ATTEMPTED )
KIDNAPPING, DEATH THREAT incidents .
Number of fillers : Null if ID slot null, or one filler per filler in ID slot .
Cross-referencing : Include cross-reference to PHYSICAL TARGET : ID(S).
A— 9
Set list :
CIVILIAN RESIDENCE
COMMERCIAL
COMMUNICATIONS
DIPLOMAT OFFICE OR RESIDENCE
ENERGY
FINANCIAL
GOVERNMENT OFFICE OR RESIDENCE
LAW ENFORCEMENT FACILITY
POLITICAL FIGURE OFFICE OR RESIDENCE
ORGANIZATION OFFICE
TRANSPORT VEHICLE
TRANSPORTATION FACILIT Y
TRANSPORTATION ROUTE
WATER
OTHER
Note : Use OTHER only if the target doesn't fit any of the specific categories.
7.12 SLOT 11 -- HUMAN TARGET: II)(S)
Applicability : Applicable to all types of incidents .
Number of fillers : Null, or one or more fillers
Note : For incidents that have a physical target, the human target will tend to be
some person or group that was at or near that target (either intentionally o r
unintentionally) at the time of the incident . Anyone who suffers injury or death a s
a result of the incident should definitely be mentioned in this slot .
7.13 SLOT 12 -- HUMAN TARGET: TOTAL NUM
Applicability : Applicable to all types of incidents .
Number of fillers: Null if ID null, else one filler
Set list:
PLURAL
Note : Use PLURAL if the only information available is that there was more tha n
one human target . Otherwise, fill this slot with a digit generated by the system
from the info available in the text . If the text indicates a range of values lik e
"between 6 and 8" or "at least 6," use a hyphen as an operator, making the fills "6 -
8" and "6-", respectively .
7.14 SLOT 13 -- HUMAN TARGET : TYPE(S)
Applicability : Applicable to all types of incidents .
Number of fillers : Null if ID slot null, or one filler per filler in ID slot .
A–10
Cross-referencing : Include cross-reference to HUMAN TARGET : ID(S) .
Set list :
CIVILIAN
DIPLOMAT
GOVERNMENT OFFICIAL
FORMER GOVERNMENT OFFICIAL
FORMER ACTIVE MILITAR Y
LEGAL OR JUDICIA L
ACTIVE MILITAR Y
POLITICAL FIGURE
LAW ENFORCEMENT
SECURITY GUARD
Notes :
1)
	
Use CIVILIAN as the default category .
2) The categories ACTIVE MILITARY and LAW ENFORCEMENT would be use d
when military and/or police personnel are targets in an incident that ha d
nonmilitary targets as well .
3)
	
Use GOVERNMENT OFFICIAL in preference to POLITICAL FIGURE in case s
where both apply .
4)
	
Use LAW ENFORCEMENT for all kinds of police forces and paramilitary force s
that are working on the side of the government .
5)
	
Use SECURITY GUARD for personal/facility security personnel who are no t
described as being members of a military or police force .
7.15 SLOT 14 -- TARGET : FOREIGN NATION(S )
Applicability : Applicable to all types of incidents .
Number of fillers : Null if country of location of incident is same as nationalit y
of target, else one filler if nationality is explicitly mentioned in the text .
Cross-referencing : Include cross-reference to PHYSICAL TARGET : ID(S) and/o r
to HUMAN TARGET : ID(S).
Example: The fill (disregarding cross-reference) should be HONDURAS if th e
incident occurred in a country other than Honduras and if the physical target i s
the Honduran embassy and/or the human target is Honduran embassy employees .
Set list: See separate file called set-list-foreign-nation .
Note : The list, in principle, could be extended to contain all the countries of th e
world. However, it should be the case that it already includes at least all th e
possibilities covered by the MUC-3 corpus .
A—11
7.16 SLOT 15 -- INSTRUMENT : TYPE(S )
Applicability : Applicable only to (ATTEMPTED) MURDER, (ATTEMPTED )
HIJACKING, (ATTEMPTED) ROBBERY, and ATTACK events .
Number of fillers : Null, or one or more fillers
Set list (choices are hierarchical) :
GUN
MACHINE GUN
MORTAR
HANDGUN
RIFLE
EXPLOSIVE
BOMB
VEHICLE BOMB
DYNAMITE
MINE
AERIAL BOMB
GRENADE
MOLOTOV COCKTAIL
CUTTING DEVICE
FIRE
TORTURE
Notes :
1) The FIRE option is included because it appeared as the cause of death in a
murder. It is not clear from the article whether the fire was set as an act o f
arson or whether it resulted from an explosion.
2)
	
Various kinds of rifles, including assault rifles, fit in the category of RIFLE .
3)
	
The supercategory of EXPLOSIVE includes a mixture of explosive devices an d
explosive materials .
4)
	
DYNAMITE includes dynamite in any of its various forms, including sticks ,
gel, etc.
5) AERIAL BOMB is for bombs dropped from airplanes . Note that reportabl e
incidents with AERIAL BOMB as instrument are to be labeled ATTACK rathe r
than BOMBING .
6)
	
Use hyphen (null) if the reported instrument does not fit any of th e
categories on the set list .
7.17 SLOT 16 -- LOCATION OF INCIDENT
Applicability : Applicable to all types of incident s
Number of fillers : One filler
A—12
General method for filling : The country where the incident occurred may b e
explicit in the text or may have to be inferred . When further information i s
available about the location, give the names in order of increasing specificity ,
separated by a colon, and put the location type in parentheses after each name .
Examples : COLOMBIA : MEDELLIN (CITY)
PERU: LIMA (CITY): MIRAFLORES (NEIGHBORHOOD)
Set list: See separate files called set-list-foreign-nation and set-list-location .
Notes :
1)
	
The incident may have occurred in the vicinity of the named place rathe r
than specifically in that place .
2)
	
The incident may have occurred between two named places, in which cas e
use a dash to separate them .
Example : COLOMBIA : CUCUTA (CITY) - CAMPANARIO (REGION)
3)
	
If the article says the specific location is unknown, just put the mor e
general location(s) .
4) Do not provide specific place name information if the message is unclear a s
to whether that place is the location of the incident or the location of th e
reporting of the incident (or some other location) .
5) This slot should never be left blank . Always put at least the country name ,
even if it must be inferred . The country where the article originated is no t
always a good default .
7.18 SLOT 17 -- EFFECT ON PHYSICAL TARGET(S )
Applicability : Not applicable to (ATTEMPTED) MURDER, (ATTEMPTED )
KIDNAPPING, DEATH THREAT incidents .
Number of fillers : Null, or one or more fillers . The EFFECT slot may be filled i n
even if the ID/NUMBER/TYPE slots are all null (in the case where the article make s
no mention at all of the target but says something in summary about damage) .
Conversely, the EFFECT slot may be null even if the ID/NUMBER/TYPE slots ar e
filled .
Cross-referencing : Include cross-reference to PHYSICAL TARGET : ID(S) ; if that
slot is null, generate a dummy cross-reference, e .g ., NO DAMAGE : "-" .
Set list :
DESTROYED
SOME DAMAGE
NO DAMAGE
A—13
Additional set list items for ROBBERY incidents :
MONEY TAKEN FROM TARGET
PROPERTY TAKEN FROM TARGE T
TARGET TAKEN
Notes :
1)
	
Use DESTROYED when the article indicates total destruction, as when it say s
that the target was "blown up" by a bomb .
2)
	
ARSON of a target implies at least SOME DAMAGE . Other incident types, suc h
as BOMBING, do not imply damage .
7.19 SLOT 18 -- EFFECT ON HUMAN TARGET(S )
Applicability : Not applicable to MURDER incidents .
Number of fillers : Null, or one or more fillers . The EFFECT slot may be filled i n
even if the ID/NUMBER/TYPE slots are all null (in the case where the article make s
no mention at all of any human targets or unwitting victims but says something i n
summary about (lack of) victims) . Conversely, the EFFECT slot may be null even i f
the ID/NUMBER/TYPE slots are filled .
Cross-referencing : Include cross-reference to HUMAN TARGET : ID(S) ; if that
slot is null, generate a dummy cross-reference, e .g., NO INJURY : "-".
Set list for all incidents except DEATH THREAT, ATTEMPTED MURDER :
INJUR Y
DEATH
NO INJUR Y
NO DEATH
NO INJURY OR DEATH
Set list for KIDNAPPING incidents, in addition to above list :
REGAINED FREEDOM
Set list for ATTEMPTED MURDER:
ESCAPED
INJURY
NO INJUR Y
Set list for DEATH THREAT incidents :
INJUR Y
NO INJUR Y
RESIGNATION (i .e., from job )
NO RESIGNATION
A—14
