【導讀】語義計算無論在廣度和深度上都對語義Web進行了擴展。它將幾個計算技術連。接并整合成一個完整統(tǒng)一的的主題。本文中通過對一種新的模式，能夠進行“問題驅(qū)。商業(yè)世界主要關于的是需求和供給。在大多數(shù)情況下的需求觸發(fā)供給，并在某些。一個更普遍的觀點是需求是問題，供應則是解決問題的方法?；ヂ?lián)網(wǎng)提供了一個全球性的基礎設施來連接問題和解決方案。取得了偉大的成績。一個基于關鍵字的搜索引擎如谷歌可能被視為一個特殊的解決問?；陉P鍵字的搜索引擎和Q＆A問答系統(tǒng)都能出色地完成他們試圖解決的問題。但是，從解決問題的角度來看，他們的效率是遠遠不夠的。已進行了一些初步的。數(shù)據(jù)庫的搜索問題，如找出在超市中超過2塊錢一斤的蘋果。綜合問題如建立一個程序，輸入一組數(shù)據(jù)，并以遞增順序返回它們。到目前為止，它已經(jīng)在商品貿(mào)易中發(fā)揮了作用。案被打包作為服務。子句中將功能作為一項條件被列入。satellite’,’star’]的字符串；ST表示，t是一個字符串，其域

　　

【正文】 SELECT g = trained(k) FROM INPUT dog k, VAR dog g WHERE * In the above, k is a dog to be trained, g is the dog after training? ?*‘ is a wild card designating ―arbitrary conditions on k and g.‖ Note that a service does not have to be rendered online。 SCDL is merely a specification language. Finally, SCDL can be extended to include metavariables to describe a service whose scope is broader. For example, given four classes account, customer, branch and depositor, the following SCDL expression specifies that it can take any SQDL query that selects objects from these classes with any bination of relational parisons between their attributes, but the number of conditions cannot exceed 4. [S4SCDL] SELECT META select FROM META from WHERE META where META IF member(t,select) THEN member(, path(account) union path(branch) union path(customer) union path(depositor)) META IF member(t,from) THEN member (, [account, branch, customer, depositor]) META IF member(t,where) AND member(s,) AND isa(s,variable) THEN member(, path(account) union path(branch) union path (customer) union path(depositor)) META IF member(t,where) THEN member(,[,=,!=,=,=]) META cardinality(where) = 4 In the above, select, from and where aremetavariables that by definition are sets. There are five constrains about the metavariables. The first constraint basically says the metavariable select can be any set of legal attributes。 the second constraint says the metavariable from can be any set of variables declared under the four 388 P. . Sheu amp。 C. V. Ramamoorthy classes of objects。 the third and fourth constraints say the metavariable where can be any set of conditions between two attributes or between an attribute and a constant, as long as the attribute(s) is legal and the parison is a relational parison。 finally the fifth constraints says the number of conditions cannot exceed four. More examples showing how SCDL can be used to describe the capability of services for each type of problems discussed in Sec. 1 may be found in [27]. Note that while an SCDL expression may be executable, in practice it is often not realistic to do so. The language is used to describe the capability of solutions only. By paring the capability of a service expressed in SCDL and a problem description, a match may be determined. 4. Describing Problems in SQDL SQDL is similar to SCDL, except that all input variables are instantiated to a constant. A query in SQDL is presented as: SELECT objects, object attributes and/or functions FROM object classes [WHERE conditions] Now, the problem ―Show all blob clusters of the images in dataset ?cmd232? that are structured like a satellite or a star‖ can be described in SQDL as: [SQDL1] SELECT i FROM image:?cmd232? im, setofblob:()sb string: [?satellite?,?star?]t WHERE (sb) and (t) As another example, the problem ―Show all blob clusters of the images in dataset ?cmd232? that are structured like a satellite or a star and do not overlap with a given blob structure‖ can be described in SQDL as: [SQDL2] SELECT i FROM image:?cmd232? im, setofblob:()sb setofblob:()sc string: [?satellite?,?star?]t WHERE (sb) and (t) and not overlap(sb,sc) In the above, SQDL1 may be solved by S1 in Sec. 4 alone. However SQDL2 has to be solved by bining S1 and S3. Problems, Solutions, and Semantic Computing 389 5. SNL — Bridging People and Descriptions Both SCDL and SQDL seem difficult for ordinary users to understand. A simpler language is needed to allow both consumers and providers to pose their sentences. Structured Natural Language (SNL) [28] is a subset of natural language whose expressive power is described by the Object Relational Algebra (ORA). It can be shown that any query expressed in SQDL can be expressed in ORA, but the reverse is not true. This may be a necessary tradeoff that we have to make to encourage more consumers (and providers) to join the new infrastructure (Factor 3 in Sec. 2). The world of ORA consists of objects that are typed. Associated with objects, we can define a set of predicates and operations. These predicates and operations (collectively called methods) can be used in a unique query language to manipulate as well as retrieve data. This allows many abstract concepts or domain dependent methods that cannot be expressed in the relational model to be included as a part of a query. Details of ORA can be found in [28]. A sentence in SNL is expressed in the following form: Verb noun(s) where adjectives SNL sentences may be posed by na168。?ve users based on simple multiple hierarchical choices without knowing any lowlevel concepts such as ―join‖ and ―selection‖. Suppose, for example, that one wishes to identify the number of patients with a clinical diagnosis of Alzheimer‘s disease who have diabetes and have been generated for a specific geic marker (. ApoE) genotype results available. The sentence would consist of a noun (patients), three adjectives/predicates (―with clinical diagnosis of AD‖, ‖having diabetes‖, ―having ApoE genotype results available‖) and a verb (―find‖). Consider SQDL1 discussed in Sec. 4. It can be described in SNL as: [SNL1] Show (VERB) blobs of images in ?cmd232? (NOUN) that have blobs structured like a satellite or a star (ADJECTIVE) And the SNL expression for SQDL2 is: [SNL2] Show (VERB) blobs of images in ?cmd232? (NOUN) that have blobs structured like a satellite or a star (ADJECTIVE) an not overlap with other blob structure (ADJECTIVE) 6. Matching Problems with Solutions is a problemdriven search service that takes a user problem described in SNL/SQDL and tries to identify one or more services whose capability 390 P. . Sheu amp。 C. V. Ramamoorthy can match the problem description. Service discovery can be done in two phases: service registration fo