《A Wiki for Business Rules in Open Vocabulary, Executable English.pdf》由会员分享，可在线阅读，更多相关《A Wiki for Business Rules in Open Vocabulary, Executable English.pdf（9页珍藏版）》请在三一文库上搜索。
1、1 A Wiki for Business Rules in Open Vocabulary, Executable English Adrian Walker Reengineering Bristol, CT 06011-1412, USA Abstract The problem of business-IT alignment is of widespread economic concern. As one way of addressing the problem, this paper describes an online system that functions as a
2、kind of Wiki - one that supports the collaborative writing and running of business and scientific applications, as rules in open vocabulary, executable English, using a browser. Since the rules are in English, they are indexed by Google and other search engines. This is useful when looking for rules
3、 for a task that one has in mind. The design of the system integrates the semantics of data, with a semantics of an inference method, and also with the meanings of English sentences. As such, the system has functionality that may be useful for the Rules, Logic, Proof and Trust requirements of the Se
4、mantic Web. The system accepts rules, and small numbers of facts, typed or copy-pasted directly into a browser. One can then run the rules, again using a browser. For larger amounts of data, the system uses information in the rules to automatically generate and run SQL over networked databases. From
5、 a few highly declarative rules, the system typically generates SQL that would be too complicated to write reliably by hand. However, the system can explain its results in step-by-step hypertexted English, at the business or scientific level As befits a Wiki, shared use of the system is free. Introd
6、uction The well known layer cake diagram (Berners-Lee 2004) outlines a high level agenda for work on the Semantic Web. Figure 1. The Semantic Web Layer Cake 2 There is much current work in progress under the heading of Semantics, as in the interleaving of metadata with data in RDF or OWL, stemming f
7、rom (Berners-Lee et al, 2001). Such work fits into the layers from XML to Ontology in Figure 1. It may be useful to think of this as data semantics, or Semantics1. In the diagram, there are boxes labeled Rule, Unifying Logic, Proof and Trust, User Interface and Applications. This paper describes one
8、 way of meeting some of the requirements indicated by those boxes, with an online system that combines Semantics1 with two further kinds of meaning. Semantics2 specifies what conclusions a reasoning engine should be able to infer from any set of rules and facts (Walker 1993), using a logical model t
9、heory (Apt et al 1988) that takes into account the semantics under which databases are used in practice. Semantics3 concerns the meaning of English concepts at the author- and user-interface. The design of the system described here rests on making Semantics1, 2, and 3 work together. Adding and integ
10、rating these semantic dimensions has the potential not only to support aspects of the Semantic Web, but also to ease some significant problems in commercial IT. According to (Forrester 2005) : Aligning IT strategy with business strategy has been one of the top three issues confronting IT and busines
11、s executives for more than 20 years. Polling of CIOs and business executives conducted in 2004 revealed that aligning IT and business goals remains their No. 1 or 2 priority. As one way of addressing the problem, this paper shows how to support the writing and running of applications as rules in ope
12、n vocabulary, executable English. The system that does this takes a lightweight approach to English, backed by an inference engine that assigns a highly declarative meaning to a set of rules. The system accepts rules in English, and small numbers of RDF or relational facts, typed or copy-pasted dire
13、ctly into a browser. For larger amounts of data, the system uses information in the rules to automatically generate and run SQL over networked databases. From a few rules, the system typically generates SQL that would be too complicated to write reliably by hand. However, the system can explain its
14、results in step-by-step hypertexted English, at the business or scientific level. From the point of view of a person writing and running rules, the system is simply a browser, used actively rather than for passive browsing. Thus, the system functions as a kind of Wiki for executable English content.
15、 Here are some examples. A Semantic Resolution Example The paper (Peng et al 2002) describes an example of name resolution for e-commerce, using three namespaces: retailer, manufacturer, and shared. In the example, a retailer orders computers from a manufacturer. However, in the retailers terminolog
16、y, a computer is called a PC for Gamers, while in the manufacturers terminology, it is called a Prof Desktop. 3 Fortunately, the retailer and the manufacturer can agree that both belong to the class of Workstations/Desktops. We also find out to what extent a Prof Desktop has the required memory, CPU
17、 and so forth for a PC for Gamers. One of the rules for this is shown in Figure 2. for the retailer the term some-item1 has super-class some-class in the some-ns namespace for the manufacturer the term some-item2 has super-class that-class in the that-ns namespace - the retailer term that-item1 and
18、the manufacturer term that-item2 agree - they are of type that-class Figure 2. A Rule for Semantic Resolution The rule says that, if the two premises above the line are true, then so is the conclusion. In the rule, some-item1, that-class and so on are place holders, or variables, that will be filled
19、 in with actual values, such as PC for Gamers and Computers when the rule is run. Apart from the place holders, the rest of the words in the rule are in open vocabulary, open syntax, English. So, the rule defines the meaning of the last sentence in terms of the meanings of the first two. To avoid in
20、finite regress, this process stops at the headings of data tables. The headings are similar sentences, and the number of place holders in a heading is the number of columns in the table. To view and run the example, one can point a browser to (Reeng 2006) and select SemanticResolution1. An RDF Query
21、 Example The paper (Haase et al 2004) describes a use case in which 14 test questions are put to some RDF data about published papers and their authors. The paper describes several Semantic Web query languages, and shows that none of those languages can answer all 14 questions. It is straightforward
22、 to write rules that allow the system to answer all 14 questions. One such rule is shown in Figure 3. some-paper is related by fact#:title to some-title that-paper is related by fact#:author to some-description that-description is related by some-rdf-node to some-home-page that-home-page is related
23、by fact#:name to some-name that-home-page is related by fact#:email to some-email - that-name is an author , with email that-email , of that-title Figure 3. A Rule for the RDF Query Example 4 Reasoning over RDF, even in this relatively simple example, is quite hard for a person to follow. A nontechn
24、ical user, who was unsure whether to trust an answer, would have a hard time convincing himself of its validity simply by looking at the rules and the RDF data, and would probably find it impossible to do so with a SQL-like query language. To help with this matter of trust, the system can supply a s
25、tep-by-step English explanation of any answer that it produces. It can also explain in English why it failed to give an expected answer. The answer to a request such as show me the authors of papers and their email addresses is a table saying that, amongst others, “Jeen Broekstra is an author , with
26、 email jbroekscs.vu.nl , of RDF Query Languages”. A step in an explanation is shown in Figure 4. Paper is related by fact#:title to An Overview of RDF Query Languages Paper is related by fact#:author to _Description1 _Description1 is related by rdf:_1 to http:/www.cs.vu.nl/jbroeks/ http:/www.cs.vu.n
27、l/jbroeks/ is related by fact#:name to Jeen Broekstra http:/www.cs.vu.nl/jbroeks/ is related by fact#:email to jbroekscs.vu.nl - Jeen Broekstra is an author , with email jbroekscs.vu.nl , of RDF Query Languages Figure 4. An Explanation Step in the RDF Query Example On the other hand, if we ask wheth
28、er Adrian Walker is the author of the paper, we get a “No” answer, and a step in the explanation is similar to Figure 4, except that the last two premises are marked as “missing”, and the conclusion is marked “not shown”. An explanation always starts out with the general justification of an answer,
29、and provides hyperlinks so that one can drill down into more detail as needed. In particular, we could write additional rules so that a rather technical, RDF-based explanation step is preceded by something more suitable for an end user to read. To view and run the example, one can point a browser to
30、 (Reeng 2006) and select RDFQueryLangComparison1. An OWL Inferencing Test Example The W3C provides a number of test cases for OWL (W3C 2004). One of these requires the inference that the items in a list are different if they are of rdf:type owl:AllDifferent. One of the rules for this task is shown i
31、n Figure 5. 5 some-tag is related by rdf:type to owl:AllDifferent that-tag is related by owl:distinctMembers to some-start-tag from that-start-tag we can follow a list to some-item that-item is related by rdf:type to some-type - that-tag names a collection of distinct items of type that-type that in
32、cludes that-item Figure 5. A Rule for the OWL Inferencing Example To view and run the example, one can point a browser to (Reeng 2006) and select OwlTest1. An Oil Industry Supply Chain Example When a geographic region has a demand for a quantity of an oil product, it is in general possible to meet t
33、he demand using a number of equivalent products. Many factors influence the proportions of component products that are combined to make an optimal supply chain decision. The factors include the season of the year, the locations of available equivalent products, and the availability of suitable and t
34、imely transportation. For our example (Kowalski and Walker 2005), we project that the target region NJ will need 1000 gallons of product y in October. We then ask what alternative routes and modes-of-transportation (truck, train, boat, pipe) do we have to get that product to the region. Next we ask
35、whether theres a refinery nearby that can produce the base product for finished product y. With all of that, we finally say that we need a delivery plan that is optimized to deliver on time, make a profit, and beat the competition. However, if there is not enough of product y, then, depending on the
36、 region and the customers, product x or z will do as well; theyre just variations of y using different additives. But theyll only do just as well in region NJ for the season including October. This makes sales projections and marketing more complicated, but also gives us more competitive flexibility
37、. One of the rules for the supply chain task is shown in Figure 6. Estimated demand some-id in some-region is for some-quantity gallons of some-finished-product in some-month of some-year for demand that-id for that-finished-product refinery some-refinery can supply some-amount gallons of some-produ
38、ct for demand that-id the refineries have altogether some-total gallons of acceptable base products that-amount / that-total = some-long-fraction that-long-fraction rounded to 2 places after the decimal point is some-fraction - for estimated demand that-id that-fraction of the order will be that-pro
39、duct from that-refinery Figure 6. A Rule for the Oil Industry Supply Chain Example As of the writing of this paper, typing “estimated demand some-id in some-region” into Google finds the rules on the world wide web. 6 Here is a fragment of the SQL that is automatically generated from the rules: sele
40、ct distinct x6,T2.PRODUCT,T1.NAME,T2.AMOUNT,x5 from T6 tt1,T6 tt2,T5,T4,T3,T2,T1,T6, (select x3 x6,T6.FINISHED_PRODUCT x7,T6.ID x8,tt1.ID x9,tt2.ID x10,sum(x4) x5 from T6,T6 tt1,T6 tt2, (select T6.ID x3,T3.PRODUCT1,T1.NAME,T2.AMOUNT x4,T2.PRODUCT from T1,T2,T3,T4,T5,T6,T6 tt1,T6 tt2 where T1.NAME=T2
41、.NAME and T1.REGION=T6.REGION and T2.MONTH1=T4.MONTH1 and T2.MONTH1=T6.MONTH1 and T2.PRODUCT=T3.PRODUCT2 and T4.MONTH1=T6.MONTH1 and T3.PRODUCT1=T6.FINISHED_PRODUCT and T3.SEASON=T4.SEASON and T3.SEASON=T5.SEASON and T4.SEASON=T5.SEASON and T6.ID=tt1.ID and T6.ID=tt2.ID and tt1.ID=tt2.ID) union . It
42、 would be difficult to write such SQL reliably by hand, or to manually validate a result that the system has found. As a way of establishing trust, the system can explain each result in step-by-step, hypertexted English, at the business or scientific level. To view and run the example, one can point
43、 a browser to (Reeng 2006) and select Oil-IndustrySupplyChain1. A Bioinformatics Ontology Example The paper (Smith et al 2005) describes a way of assigning a formal meaning to relations in bioinformatics ontologies. For example, the paper defines what it means for a continuant class to be a part of
44、another class, in terms of some easily understood primitive relations and some reasoning over time. If C and C1 are classes, then the paper defines C part_of C1 = definition for all c, t, if Cct then there is some c1 such that C1c1t and c part_of c1 at t where Cct is shorthand for c is an instance o
45、f C at time t Figure 7 shows some of the rules that make this definition executable over ground data. 7 for all c, t, if some-C c t then there is some c1 such that some-C1 c1 t and c part_of c1 at t - that-C is a part_of the continuant class that-C1 (A c,t) some-C c t = (E c1) some-C1 c1 t and c par
46、t_of c1 at t - for all c, t, if that-C c t then there is some c1 such that that-C1 c1 t and c part_of c1 at t some-C and some-C1 are two different Non-process classes with instances not : (E c,t) that-C c t and not (E c1) that-C1 c1 t and c part_of c1 at t - (A c,t) that-C c t = (E c1) that-C1 c1 t and c part_of c1 at t Figure 7. Some Rules for the Bioinformatics Ontology Example To view and run the example, one can point a browser to (Reeng 2006) and select RelBioOntDefn3. System Design In any syste