This is the readme for the Illinois Intentional Tort Qualitative Dataset.
The dataset was developed by Joseph Blass (joeblass@u.northwestern.edu) as part 
of his thesis work. He would love your comments and feedback on it!

The dataset currently consists of 88 cases representing 112 claims in Tort, in 
Assault, Battery, Trespass, and Self-Defense. Each case is contained in its 
own file. The files are Knowledge Representation Files with extension .krf, 
designed for use in a system like NextKB. That said, you can open the files 
as text files or change the file extension to .txt without any trouble.

The dataset is encoded in CycL-style predicate logic representation, of the 
form (<pred> <arg1> <arg2> . . . <argn>). Cases are also scoped within microtheories, 
keeping metadata about a case separate from the predicate logic representations 
of that case. Cases are defined within the NextKB ontology, publicly available 
at https://www.qrg.northwestern.edu/nextkb/index.html .

The rest of the readme is concerned with explaining the different microtheories 
you can expect to encounter and the predicates used to describe the cases.

CASE SYMBOLS
Every case is given a unique symbol based on the case name. For example, the case 
Amos v. State has the unique symbol Amos_v_State. This symbol is used to attach 
information to the case and create microtheories.

MICROTHEORIES
Microtheories are a way of scoping knowledge contextually. In this dataset, 
when you see (in-microtheory <microtheory-name>), that means that all subsequent 
facts are scoped in the microtheory <microtheory-name>, until the next 
in-microtheory statement appears.
Microtheories can inherit from one another, such that within a child microtheory, 
a reasoner can see all knowledge from the parent microtheories from which the  
child microtheory inherits. Inheritance is expressed using (genlMt <childMt> <parentMt).

There is a global microtheory, CaseLawCorpusMt, in which all the cases are identified 
as being members of the dataset. This is defined by the statement 
(isa <case-name> CaseLawCorpusCase). All other information is kept in a microtheory specific 
to the case itself, (CaseLawCorpusMtFn <case-name>), or in child microtheories of that 
case microtheory. For example, metadat for our case Amos v. State comes after the statement
(in-microtheory (CaseLawCorpusMtFn Amos_v_State))

The submicrotheories of each case are:
 - (rawLanguageOutputMtFn <case-name>), which contains the raw language output of CNLU for that case;
 - (cleanLanguageOutputMtFn <case-name>), which contains the language output with erroneous facts 
removed, without the duplicate unnesting of facts that is a raw output of CNLU, and with temporal 
ordering statements rerepresented to indicate temporal ordering;
 - (LegalCaseMtFn <case-name>), which contains the cleanest output of the case facts, including 
having extra facts that are reintroduced at each statement (e.g., (isa <plaintiff> Plaintiff)) removed;
 - (LegalCaseConclusionMtFn <case-name>), which contains the predicate logic representations of the 
case conclusions; and
 - variations on the LegalCaseConclusionMtFn (negated; reversed; both) for use in multiple choice tests.
 
PREDICATES
(caseType <case-name> <claim-type>) indicates that <case-name> is a case illustrating the tort claim 
<claim-type>. Cases can have multiple such statements, e.g., many cases illustrate both assault and battery.

(caseValence <case-name> <claim-type> <Positive or Negative>) indicates whether case-name is a Positive or 
Negative example of the claim <claim-type>.

(caseName <case-name> <name-string>) encodes the original case name of <case-name>

(caseCourt <case-name> <court>) shows which court <case-name> was decided in

(caseYear <case-name> <year>) keeps the case year

(caseReporter <case-name> <reporter index>) contains the index for the particular case reporter where 
<case-name> was originally reported. Some of these are in Westlaw or LexisNexis specific reporters, 
but most are in Illinois or Federal reporters.

(caseOriginalText <case-name> <casefact-string>) contains the judicial opinion's original statement of 
the case facts. These are quotes taken directly from the judicial opinions and only minimally edited for 
coherence.

(caseConclusion <case-name> <conclusion>) contains a plain english description of the case's legal conclusion. 
Cases can have more than one conclusion.

(caseSimpleText <case-name> <casefact-string-simplified>) contains the simplified english statement of 
the original case facts. Please refer to the 2022 JURIX paper for a description of how facts were simplified.