Combining Semantic Wikis and Controlled Natural Language

Combining Semantic Wikis and Contr olled Natural Language T obias K uhn Depar tment of Inf or matics University of Zurich, Switzerland tkuhn@ifi.uzh.ch ABSTRA CT W e demonstrate AceWiki that is a semantic wiki using the con trolled natural language Attempto Con trolled English (A CE). The goal is to enable easy creation and mo dification of ont ologies through the web. T exts in ACE can automati- cally be translated in to first-order logic and other languages , for example O WL. Previous ev aluation sh ow ed that ordinary people are able to use AceWiki without b eing instructed. K eywords Con trolled Natural Language, Attempto Controlled English (A CE), Semant ic W eb, Semantic Wiki, AceWiki, Ontology 1. INTR ODUCTION Most of the tasks the Semantic W eb is even tually supp osed to fulfill rely on the av ailabilit y of ontologies. Ho wev er, the creation and maintenance of ontologies is difficult b ecause a n umber of domain experts — most of which are not familiar with formal languages — hav e to agree on a conceptualiza- tion of the respective domain. F or that reason, it is crucial for the future of the Semantic W eb to provide tools that mak e the creation of ontologies easy for everybo dy . AceWiki 1 tac kles this problem by combining seman tic wikis with controlled natural language. The goal of AceWiki is to enable ordinary people with no background in formal lan- guages to create expressive on tologies in a collab orative and in tuitive wa y without the need of installing an application. 2. B A CKGR OUND There are severa l existing semantic wiki systems, see e.g. [3] for a brief survey . Unfortunately , most of those wikis do not support expressiv e ontology languages in a general wa y . F urthermore, they are often hard to understand for p eople who are not familiar with the technical terms. A ttempto Controlled English (ACE) 2 is the controlled nat- ural language that is used for AceWiki. Being a subset of English, ACE lo oks completely natural. Restrictions of the syn tax and the definition of a small set of interpretation rules mak e it a formal language that is automatically translatable in to first-order logic. A CE cov ers a large part of natural En- glish: singular and plural noun phrases, activ e and passive v oice, relative phrases, anaphoric references, existential and univ ersal quan tifiers, negation, and muc h more. 1 See [3], [4], and http://attempto.ifi.uzh.ch/acewik i 2 See [1] and http://attempto.ifi.uzh.ch Figure 1: A screenshot of the predictive editor of AceWiki. The fragment “Every area is” has already b een ente red and no w the editor shows all p ossibil- ities to contin ue the sen tence. A CE has been used as a natural language fron t-end to OWL with a bidirectional mapping of ACE to OWL [2]. AceWiki uses this for translating ACE sentences into OWL. The same w ork also introduces a Prot´ eg ´ e plugin called “ACE View” whic h enables to manage onto logies in ACE within the Pro- t ´ eg´ e environmen t. 3. SYSTEM In AceWiki, the ontolo gical ent ities are represented by nat- ural language words and phrases. Proper names (e.g. “Zu- ric h” , “Switze rland” , “Europe” ) are interpreted as individu- als, nouns (e.g. “cit y” , “coun try” ) are in terpreted as classes, and transitive v erbs (e.g. “borders” ), of -constructs (e.g. “part of ” ), and transitive adjectives (e.g. “located-in” ) are in terpreted as binary relations. Using those words together with the predefined function w ords of ACE (e.g . “a” , “ev ery” , “if ” , “then” , “and” , “n ot” , “is” , “that” ), on tological statemen ts are expressed as ACE sentences: As those examples sho w, the formal statemen ts are easily readable and understandable b y any English sp eaking p er- son. In order to enable easy creation and modification of A CE sentences, AceWiki in tegrates a predictiv e editor that sho ws step-b y-step the words that are syntactically p ossi- Figure 2: A screenshot of the w eb interface of AceWiki showing the wiki article for the class “con- tinen t” . ble at a giv en p osition in the sen tence. Figure 1 sho ws a screenshot of this editor. Eac h of the ontological entities gets its o wn wiki article. Figure 2 sho ws an example. Every article consist of A CE sen tences most of which can b e translated into OWL, e.g: A CE is more expressive than OWL, and thus we can write statemen ts that go beyond the semantic expressivit y of OWL (e.g. rule-lik e statemen ts). Such statements are mark ed with a red triangle (and are curren tly ignored by the reasoner): F urthermore, questions can b e used to query the knowledge base, e.g: Th us, ACE is an ontology language, a rule language, and a query language at the same time. AceWiki uses the O WL reasoner Pellet 3 to p erform rea- soning tasks ov er the sentences of the wiki that are OWL- complian t. In order to ensure the consistency of the ontol- ogy , ev ery new sentence is chec ked — immediately after its creation — whether it wo uld in troduce a con tradiction. If this is the case then the sente nce is not included in the on- tology and display ed in red font: 3 http://pellet.owldl.com/ The reasoner is also used to infer the class memberships of individuals. The results are presen ted in A CE again: The same is done for class hierarchies: This shows that not only asserted but also inferred knowl- edge is represented in AC E. Finally , the reasoner is also used to answer questions: In general, we can sa y that AceWiki communicates with the users on a very natural lev el. No knowledge ab out formal languages is required to deal with AceWiki. 4. EV ALU A TION In our previous w ork [3], w e conducted a user experiment that show ed that ordinary people with no background in logic and on tologies are able to deal with AceWiki. The par- ticipan ts — without b eing instructed ho w to interact with the interface — w ere ask ed to add knowledge to AceWiki. About 80% of the created sen tences w ere correct and sen- sible. Remark ably , more than 60% of those sen tences were complex in the sense that they con tained an implication or a negation. 5. CONCLUSIONS AceWiki shows how ontologies can be created and modified in a natural wa y within a wiki. It demonstrates how seman- tic wikis using controlle d natural language can b e expressiv e and easy to use at the same time. Our previous ev aluation sho wed that AceWiki is indeed easy to learn. 6. REFERENCES [1] Norbert E. F uchs, Kaarel Kaljurand, T obias Kuhn. A ttempto Con trolled English for Knowledge Represen tation. R e asoning Web, F ourth International Summer Scho ol , Springer, 2008 [2] Kaarel Kaljurand. Attempto Contr ol le d English as a Semantic Web L anguage . PhD thesis, F aculty of Mathematics and Computer Science, Universit y of T artu, 2007 [3] T obias Kuhn. AceWiki: A Natural and Expressive Seman tic Wiki. Pro c. of Semantic Web User Inter action at CHI 2008: Exploring HCI Chal lenges , CEUR W orkshop Pro ceedings, 2008 [4] T obias Kuhn. AceWiki: Collaborative Ontology Managemen t in Controlled Natural Language. Pr o c. of the 3r d Semantic Wiki Workshop , CEUR W orkshop Proceedings, 2008