OASIS Mailing List ArchivesView the OASIS mailing list archive below
or browse/search using MarkMail.

 


Help: OASIS Mailing Lists Help | MarkMail Help

legalruleml message

[Date Prev] | [Thread Prev] | [Thread Next] | [Date Next] -- [Date Index] | [Thread Index] | [List Home]


Subject: Fwd: RE: Some notes for today


Hi,

As I already wrote in my earlier emails (attached), we don't need to invent another XML language for temporal representations and spend time on its design, development, and maintenance. Temporal representations have been already addressed by many communities and standards. There are more than enough existing languages and ontologies which we can simply reuse, as, at a closer look, there is nothing special about the representation of time in the legal domain.
It would be a poor organisational decision if we still started to develop
our own non-semantic data definition language in XML, e.g. for defining
granularities, different types of intervals, times etc. instead of using
semantic ontology languages. 
(Reaction) RuleML with its generic constructs for Time, Interval and
temporal operators provides the required openness and flexibility to support
this reuse. 

Looking at the examples, I do not see why we should spend time on developinga new time data definition language in XML, instead of reusing what is there:


>1. "The document entered into force immediately."
>Problem: we have to capture the time and the date and this datetime is
relevant for the criminal law and for checking if a such crime enter in to
the "temporal window" of validity of the law.

This is a situation "entered into force" which holds at an interval which
has as a starting point a relative time point "immediately". As explained in my previous emails Reaction RuleML supports continuous time models with
TimeInstants and discrete time models with time points resp. TimeIntervals.It allows supports different types of intervals from empty to all types of open and closed intervals.

The question for this example is what we actually want to represent. Is it
the initiating time point of the "inforce" situation, then we should use a
discrete time point. Since "immediate" is a relative time definition we need to either additionally define the reference time point to which it relates to or we need to transform it into an absolute  time point.

   <Time type="&ruleml;TimeInterval">< Data
xsi:type="xs:dateTime">2012-10-02T00:00:00-00:00</Data></Time>

If we want to represent the interval of "inforce" we can use an <Interval> 

Defined in terms of an interval with absolute start time point and unbounded end time point

<Interval type="&LeftClosed">
   <Time>< Data
xsi:type="xs:dateTime">2012-10-02T00:00:00-00:00</Data></Time>
</Interval>

Defined as an interval with a relative time using a duration. Note, we would need to specify an absolute reference time point (e.g. publication of the law) for a relative time which comes after this absolute reference time point (see #2 - temporal operators)  otherwise it would change according to the current time.

<Interval type="&ruleml;LeftClosed">
     < Time><Data xsi:type="xs:duration">0</Data></Time>
</Interval>

Defined as interval with user-defined temporal function called "immediately" (note, we could also refer to, e.g., a built-in function or a reference to an ontological function definition what "immediately" means etc.). Again, since "immediately" is a relative time function it would need an absolute reference time point as argument.

<Interval type="&ruleml;LeftClosed">
     < Time><Expr><Fun>immediately</Fun>... definition of an absolute
reference time point for the function immediately </Expr></ Time>
</Interval>

If we actually want to do reasoning with situations we would need a more
expressive representation which supports logical calculi for reasoning about the effects of events/actions over time, such as in KR layer of Reaction RuleML which supports fluents (changeable situations). Situations such as "efficacy", "inforce" then would be modeled as fluents in Reaction RuleML and can be initiated holding true by "law of intertia" until they are terminated.

>2. "This title takes effect 6 months after the date of the publication of
this Act."

This a temporal operation with a relative time point after some absolute
time point. Reaction RuleML therefore has a rich set of temporal operators
such as After, Every, Any, Timer (recurring timer) which compute absolute
and relative time points.

<After key="startEfficacyAct1">
      < Time key="#publicationDateAct1">< Data
xsi:type="xs:date">2012-10-02</Data></Time>
      <Time><Data xsi:type="xs:duration">6M</Data></Time>
</After>

or defined using Reaction RuleML's key reference to the definition of  a
time point:

 < Time key="#publicationDateAct1">< Data
xsi:type="xs:date">2012-10-02</Data></Time>
...
<After>
      < Time keyref="#publicationDateAct1"/>
      <Time><Data xsi:type="xs:duration">6M</Data></Time>
</After>

The temporal operator function would map the relative time "6 months" to an
absolute time point. 

>3. "On the 1st of April this year, the new Consumer Protection Act comes
into effect."

This is again a situation "comes into effect" like in #1. "this year" is a
relative time where we would need to define an absolute reference time point with the relative time "1st of April this year" after this absolute time point e.g. "2011-1-1"  or directly replace it with the absolute date
"2011-4-1". 

>4."The consumer must present the application form within 31/12/2012".
>It means in legal domain, often, that the endPoint is 30/12/201211.59.59.
>But in the legal text we have only "31-12-2012" (e.g. captured with NLP
tools).
>How to integrate this information with the nature of the interval (open on
the endPOint)?

This is an interpretation what "31-12-2012" in the legal text means in terms of a concrete data value, i.e. 30/12/201211.59.59. My interpretation would be that the time interval of the situation "consumers must present application forms" needs to be "before 30/12/201211.59.59". Reaction RuleML therefore has a rich set of interval operators (based on Allen's interval calculus) such as During, Overlaps, Preceds etc. So we can simply use these if we want to say that this obligation interval needs to be before this end time point. If we just want to say that the time point when the application form is actually presented (i.e. when the event occurs) is before 30/12/201211.59.59, we can simply use the standard arithmetic built-ins.
Moreover, the existing time ontologies and also the ISO standard for calendar datetimes in the XML datatypes (including data type facets) provide the expressiveness e.g. to define quantities such as hours, minutes, seconds, including the definition of time zones. We can simply refer to such semantic definitions in the time definition using the type attribute <Time type="...">, so that the time (data) value gets a precise semantic meaning.
 

Best,

Adrian

-----Original Message-----
From: legalruleml@lists.oasis-open.org
[mailto:legalruleml@lists.oasis-open.org] On Behalf Of monica.palmirani
Sent: Mittwoch, 24. Oktober 2012 15:03
To: legalruleml@lists.oasis-open.org
Subject: [legalruleml] Some notes for today

Hi,

I have tried to model some examples using the temporal model.
In my opinion inside to the TimeCharateristics block we have to add new tags
for capturing:
- temporal model (discrete or linear): proposal lrml:TimeMode
- granularity (second, day, etc.): proposal lrml:TimeGranularity
- if the interval is open or close or infinite, [t1, t4] or [t1, t4[
- how to distinguish a date per-se and a date that start an infinite
interval? [t1] or [t1, infinite[?
- duration

Examples:

1. "The document entered into force immediately."
Problem: we have to capture the time and the date and this datetime is
relevant for the criminal law and for checking if a such crime enter in to
the "temporal window" of validity of the law.
How to say that the time is important and not only optional?

2. "This title takes effect 6 months after the date of the publication of
this Act."
Problem: to model the duration using a well-know tag and not simply <Data
xsi:type="xs:duration">P5Y2M10D</Data>
In case we have
<Data xsi:type="xs:dateTime>2012-10-02T10:45:34-00:00</Data>
<Data xsi:type="xs:duration">P5Y2M10D</Data>
I have also xs:duration as only one way to express this information and this
is not mandatory but just leave to the end-user. Better to have a specific
lrml:duration xsi:type="xs:duration"

3. "On the 1st of April this year, the new Consumer Protection Act comes
into effect."
Infinite interval starting from 1-04-2012 "Order to suspend the traffic at 1
April 2012."
How to distinguish these two cases?

4."The consumer must present the application form within 31/12/2012".
It means in legal domain, often, that the endPoint is 30/12/201211.59.59.
But in the legal text we have only "31-12-2012" (e.g. captured with NLP
tools).
How to integrate this information with the nature of the interval (open on
the endPOint)?

I would be apprentice very much if the TC can take in consideration my
requests and provide a concrete proof-of-concept how it is possible to
manage the problem presented respecting the isomorphism principles.

yours,
mp


---------------------------------------------------------------------
To unsubscribe, e-mail: legalruleml-unsubscribe@lists.oasis-open.org
For additional commands, e-mail: legalruleml-help@lists.oasis-open.org
--- Begin Message ---
Hi Tara,

Yes, there are many existing ontologies and vocabularies/schemas available for time. Since there is nothing special about the representation of time in the legal domain which would require a domain-specific model, we should reuse them.

Reaction RuleML therefore has a generic approach which supports this reuse. There is a basic distinction in whether the temporal aspect are based on the relationships between "time instants" (type="ruleml:TimeInstant") or "time intervals" (type="ruleml:TimeInterval">). Depending on this core distinction a time becomes a linear continuous time model with points in time called "time instants"(<Time type="ruleml:TimeInstant">) or a discrete time interval model with "time points" (<Time type="ruleml:TimePoint">) as a relative time interval with an absolute reference or a relative reference to some given time interval. 

A duration is an amounts of time, .e.g. <ruleml:Time> <ruleml:Data xsi:type="xs:duration">P5Y2M10D</ruleml:Data> </ruleml:Time> or other quantities of duration values with other types of time units including nominal time units, e.g.3 years  <ruleml:Time><Ind iri="owl-time:Year">3</Ind><ruleml:Time> and discrete time functions (<ruleML:Time><Expr>).

<Interval> in Reaction RuleML is a generic concept for modeling intervals (Time/Spatial/Event/Action/Situation Intervals). Time intervals are then either based on the  segments of a linear continuum of time instants or the discrete time points (which span a smaller time interval).

Reaction RuleML also has a rich set of algebra operators for intervals (Allens relations such as During, Overlaps, ...) and general time arithmetic's and aggregators (such as Every, Timer, Any, ...-), and a general <Operator> (XML extension point) for adding further special operators from the many existing domain specific languages.

So, depending on whether the Legal RuleML collection is a set of time points it would be <lrml:TimePoints>  or <lrml:TimeInstants> in the case of time instants. 
I would propose to be more generic and say <lrml:Collection type="ruleml:TimePoint">, so that we can have different types of Collections.

-Adrian


From: legalruleml@lists.oasis-open.org [mailto:legalruleml@lists.oasis-open.org] On Behalf Of Tara Athan
Sent: Donnerstag, 18. Oktober 2012 01:02
To: legalruleml@lists.oasis-open.org
Subject: [legalruleml] Time Instants or Time Points

There is a relatively recent date-time ontology published by OMG (http://www.nist.gov/customcf/get_pdf.cfm?pub_id=911074) that addresses time scale granularity and also solves the problem of duration (is 1 month greater, equal or less than 30 days?)

While LegalRuleML is not supposed to be dependent on any particular ontology, this work does provide a useful conceptualization that allows certain intervals of time to be specified with a single datum. For example the date 2012-10-12 corresponds to a particular interval of time which can be specialized to a particular 24 hr interval when a timezone is added. This general concept of a point on a time scale with a particular granularity is called "time point". It is more general than a "point in time" or "time instant", but could include that concept if the time scale is continuous.

I proposed that we adopt this terminology, (but not the ontology) by referring to a collection of <ruleml:Time> elements as <lrml:TimePoints>

This also avoids the unfortunate plural "Times", which could be confused with multiplication or repitition, while retaining generality regarding the granularity, if any, of the time scale.

Example

<lrml:TimePoints>
  <ruleml:Time key="t1"> <ruleml:Data xsi:type="xs:dateTime">2012-07-21T00:00:00Z</ruleml:Data> </ruleml:Time>
  <ruleml:Time key="t2"> <ruleml:Data xsi:type="xs:date">2012-07-21</ruleml:Data> </ruleml:Time> </lrml:TimePoints>

Tara

---------------------------------------------------------------------
To unsubscribe, e-mail: legalruleml-unsubscribe@lists.oasis-open.org
For additional commands, e-mail: legalruleml-help@lists.oasis-open.org

--- End Message ---
--- Begin Message ---
Hi Tara,

Yes, I also think we need both. That is why I propose to use a more general <Collection> which can be typed (type="..") with different sorts.

1. Is there a default selection between continuous and discrete timescale that would resolve this ambiguity?

Yes, the default one is a time interval semantics. 
(On a side note: this distinction in interval semantics and point in time semantics is important as it avoids ambiguity in the algebra operators for complex events, as we have shown in our interval base event calculus semantic for complex event processing)

2. According to the schema, the content model appears to be essentially
unrestricted: <ruleml:Time><ruleml:content>...anything...</ruleml:content></ruleml:Time>

No, the main content type is an <arg> with e.g. Ind, Data, Var or Expr or a slotted representation with a <slot>. Therefore it reuses the content model of <arg> and <slot> from <Atom>.
The additional choice of a <content> role tag with unrestricted XML content model is an extension point of the Reaction RuleML syntax, which allows embedding one of the many existing XML serializations for time data from the various communities which deal with time.

3. Is it possible to use the element <ruleml:Time> to describe a timepoint that is 6 months after some named timepoint?

Yes, this is a temporal operation 

<After>
      <ruleml:Time key="#t1">...</ruleml:Time>
      <ruleml:Time><ruleml:Data xsi:type="xs:duration">6M</ruleml:Data></ruleml:Time>
</After>


BTW, I just encountered that I had sent my regrets for the last telecon on Wednesday not from my gmx address but from my university address  which apparently seems to be rejected by the mailing list. Sorry, for that. I attach it again with the example version which preserves the advantages and conveniently combines it with the rest of RuleML.

-Adrian

-----Original Message-----
From: legalruleml@lists.oasis-open.org [mailto:legalruleml@lists.oasis-open.org] On Behalf Of Tara Athan
Sent: Samstag, 20. Oktober 2012 17:26
To: legalruleml@lists.oasis-open.org
Subject: Re: [legalruleml] Time Instants or Time Points

Thanks Adrian.

In legal applications, it appears to me that both discrete and continuous of timescales are needed.
When describing an observation of the time at which a physical event takes place (e.g. publication), then a discrete timescale is appropriate because this captures the fact that all observations have limited precision.  Discrete timescales capture this uncertainty as intervals in time in a similar way that significant digits are used to record numerical measurements in general, (except that the time interval is assumed to lie on the future side of the time observation rather than being centered on the measurement as is done with significant digits.)

When defining a fiat event (e.g start of enforceability), then a continuous timescale may be appropriate because the definition can be made with infinite precision.

I have included the <ruleml:Time> element in the LegalRuleML schema, but so far I have only implement a restricted content model of the form

<ruleml:Time><ruleml:Data
xsi:type="xs:dateTime>2012-10-02T10:45:34-00:00</ruleml:Data></ruleml:Time>

I have a couple of questions:

1. Based on what you say below, the interpretation of this construct appears to be ambiguous because there is no @type on the Time element.

If a discrete timescale is adopted, this would represent an interval of length one second.

If a continuous timescale is adopted, this would be interpreted as a time instant, in particular the beginning of the time interval described by <ruleml:Data xsi:type="xs:dateTime>2012-10-02T10:45:34-00:00</ruleml:Data>

Is there a default selection between continuous and discrete timescale that would resolve this ambiguity?

Since the argument
<ruleml:Data xsi:type="xs:dateTime>2012-10-02T10:45:34-00:00</ruleml:Data>
already denotes a temporal entity, a time period of length 1 second (the value space of xs:dateTime is actual time periods and is based on ISO 8601), I would expect the default, if there is one, to be the discrete timescale.


2. I would like to include more of the content model for this element as defined in the Reaction RuleML schema http://www.csw.inf.fu-berlin.de/RRML1.0/xsd/modules/time_module.xsd

but the interpretation of this more general syntax is not clear to me.
According to the schema, the content model appears to be essentially
unrestricted:
<ruleml:Time><ruleml:content>...anything...</ruleml:content></ruleml:Time>

Is it possible to use the element <ruleml:Time> to describe a timepoint that is 6 months after some named timepoint?
( <ruleml:Time key="#t1">...</ruleml:Time> )


Tara

On 10/20/2012 9:24 AM, Adrian Paschke wrote:
> Hi Tara,
>
> Yes, there are many existing ontologies and vocabularies/schemas available for time. Since there is nothing special about the representation of time in the legal domain which would require a domain-specific model, we should reuse them.
>
> Reaction RuleML therefore has a generic approach which supports this reuse. There is a basic distinction in whether the temporal aspect are based on the relationships between "time instants" (type="ruleml:TimeInstant") or "time intervals" (type="ruleml:TimeInterval">). Depending on this core distinction a time becomes a linear continuous time model with points in time called "time instants"(<Time type="ruleml:TimeInstant">) or a discrete time interval model with "time points" (<Time type="ruleml:TimePoint">) as a relative time interval with an absolute reference or a relative reference to some given time interval.
>
> A duration is an amounts of time, .e.g. <ruleml:Time> <ruleml:Data xsi:type="xs:duration">P5Y2M10D</ruleml:Data> </ruleml:Time> or other quantities of duration values with other types of time units including nominal time units, e.g.3 years  <ruleml:Time><Ind iri="owl-time:Year">3</Ind><ruleml:Time> and discrete time functions (<ruleML:Time><Expr>).
>
> <Interval> in Reaction RuleML is a generic concept for modeling intervals (Time/Spatial/Event/Action/Situation Intervals). Time intervals are then either based on the  segments of a linear continuum of time instants or the discrete time points (which span a smaller time interval).
>
> Reaction RuleML also has a rich set of algebra operators for intervals (Allens relations such as During, Overlaps, ...) and general time arithmetic's and aggregators (such as Every, Timer, Any, ...-), and a general <Operator> (XML extension point) for adding further special operators from the many existing domain specific languages.
>
> So, depending on whether the Legal RuleML collection is a set of time points it would be <lrml:TimePoints>  or <lrml:TimeInstants> in the case of time instants.
> I would propose to be more generic and say <lrml:Collection type="ruleml:TimePoint">, so that we can have different types of Collections.
>
> -Adrian
>
>
> From: legalruleml@lists.oasis-open.org [mailto:legalruleml@lists.oasis-open.org] On Behalf Of Tara Athan
> Sent: Donnerstag, 18. Oktober 2012 01:02
> To: legalruleml@lists.oasis-open.org
> Subject: [legalruleml] Time Instants or Time Points
>
> There is a relatively recent date-time ontology published by OMG (http://www.nist.gov/customcf/get_pdf.cfm?pub_id=911074) that addresses time scale granularity and also solves the problem of duration (is 1 month greater, equal or less than 30 days?)
>
> While LegalRuleML is not supposed to be dependent on any particular ontology, this work does provide a useful conceptualization that allows certain intervals of time to be specified with a single datum. For example the date 2012-10-12 corresponds to a particular interval of time which can be specialized to a particular 24 hr interval when a timezone is added. This general concept of a point on a time scale with a particular granularity is called "time point". It is more general than a "point in time" or "time instant", but could include that concept if the time scale is continuous.
>
> I proposed that we adopt this terminology, (but not the ontology) by referring to a collection of <ruleml:Time> elements as <lrml:TimePoints>
>
> This also avoids the unfortunate plural "Times", which could be confused with multiplication or repitition, while retaining generality regarding the granularity, if any, of the time scale.
>
> Example
>
> <lrml:TimePoints>
>    <ruleml:Time key="t1"> <ruleml:Data xsi:type="xs:dateTime">2012-07-21T00:00:00Z</ruleml:Data> </ruleml:Time>
>    <ruleml:Time key="t2"> <ruleml:Data xsi:type="xs:date">2012-07-21</ruleml:Data> </ruleml:Time> </lrml:TimePoints>
>
> Tara
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: legalruleml-unsubscribe@lists.oasis-open.org
> For additional commands, e-mail: legalruleml-help@lists.oasis-open.org
>
>
--- Begin Message ---

Dear All,

 

unfortunately I need to send regrets for today’s telecon.

 

Following up on our discussion in the last telecon and Tara's recently uploaded example, here is a version that preserves its advantages and can be conveniently combined with the rest of RuleML.

 

Best,

 

Adrian

Attachment: ex2.1.8-defeasible_v2.lrml
Description: Binary data


--- End Message ---
---------------------------------------------------------------------
To unsubscribe, e-mail: legalruleml-unsubscribe@lists.oasis-open.org
For additional commands, e-mail: legalruleml-help@lists.oasis-open.org

--- End Message ---


[Date Prev] | [Thread Prev] | [Thread Next] | [Date Next] -- [Date Index] | [Thread Index] | [List Home]