The module will use the PowerAqua semantic question-answering system to allow a smart product to continue to search beyond locally held knowledge using the semantic web. It operates in an online mode, in the smart products environment but looks beyond it onto the (semantic) web.
PowerAqua? allows the user to express his information needs in Natural Language, as a full (interrogative) sentence. Because of the need to use resource-constrained devices on the Smart Products scenario, proactive knowledge bases (PaK) contain only limited knowledge expressing the main information needs. Therefore, sometimes it will be also desirable to explore, apart from the proactive internal information, additional knowledge and information available in other larger sources from external online repositories on the Web, like for example to find nutritional information about a new ingredient, if available. Scalability cannot be addressed on current resource constrained devices. Therefore, our proposed approach to extending ontology-based search capabilities to several external sources of information at the same time, is to use the PowerAqua? as an online remote web service. In this scenario any internet-aware smart product, independently of its resources, can interact with the PowerAqua? semantic component to extend the scope of proactive knowledge bases by locating and exploiting, at run time, existing relevant external large-scale knowledge sources in response to a query posed by the user.
In short, PowerAqua? is the only ontology-based QA system out there able to query multiple ontologies (instead of just one, selected a priori) and furthermore has a user friendly, natural language query interface which fits the usability requirements of "Products that talk to you".
Developer: Vanessa Lopez
License
The Licensor, Knowledge Media Institute at the Open University, is the owner of, and/or has acquired rights to PowerAqua?, consisting of machine readable object code, and any subsequent error corrections or updates supplied ( http://technologies.kmi.open.ac.uk/poweraqua). PowerAqua? uses GATE under the GNU Library General Public License (www.gate.ac.uk). PowerAqua? uses jwnl library under jwnl License ( http://sourceforge.net/projects/jwordnet/). PowerAqua? uses SecondString? library of string matching techniques under the University of Illinois/NCSA Open Source License ( http://secondstring.sourceforge.net/).
Download
will be made available ... http://technologies.kmi.open.ac.uk/poweraqua/
How to get started
Examples -- The types of queries that can be answered in PowerAqua? are factual queries, representing explicit or implicit relationships between one or various terms, that can be translated to one or many ontological facts or triples. This includes all the big set of queries constituted by a wh-question, such as the ones starting with: what, who, when, where, are there any, does anybody/anyone or how many, etc. Also imperative commands like list, give, tell, name, etc . Queries may vary in length and complexity: from simple queries, with adjunct structures or modifiers, to complex queries with relative sentences and conjunctions / disjunctions. However, the current version of PowerAqua? is not able to answer queries that require temporal reasoning (e.g.: last year, recently, in the 80’s, etc), casual reasoning (i.e., why- type of queries), that contain negations (e.g.: do not, never), or require any kind of ranking, percentages, comparatives or superlatives (e.g.: the most, same as, the best, the les, the fastest)
Which queries can be answered ultimately depends on the knowledge and coverage represented in the sources. Considering the philips cooking scenario, here we present an example list of competency questions that it will be desirable to answer:
• What kind of oils contain oleic acid? • Which recipes take less than 10 minutes? • What is the daily maximum recommendation of muesli? • Which recipes can be made using the microwave? • Show me recipes made with gorgonzola cheese and mashed potatoes. • Which foods contain calcium? • What are the ingredients in a minestrone? • What are the typical ingredients in Jewish cuisine? • Show me low-fat recipes.
http://poweraqua.open.ac.uk:8080/poweraqualinked
API Documentation
PowerAqua? is a java-based web interface. It also provides an API, which allows future integration in other platforms and the independent use of its components. The user types a query in Natural Language and she can browse and navigate through the set of ranked ontological answers and the relevant ontologies and ontological mappings that translate her query. In order to do that, the different components of PowerAqua? interpret the user’s query expressed in NL using the relevant available semantic information, automatically discovery from anywhere on the SW, and translates the user terminology into ontology terminology, retrieving accurate semantic entity values as response to the user’s request. PowerAqua? components are:
- Linguistic Component: it analyzes a NL query, and translates it into a set of linguistic triples, by identifying associations that relate terms together. This component is based on the GATE linguistic infrastructure (www.gate.ac.uk).
- PowerMap? mapping component: which is responsible for identifying the suitable semantic sources that are likely to provide the information requested by the user and answer the given query, producing initial element mappings between the terms in the linguistic triples and entities in these sources. PowerMap? uses both WordNet? and the SW itself as sources of background knowledge to perform query expansion and to find lexically dissimilar (but semantically similar) matches
- The Triple Similarity Service (TSS): returns a set of ontology compliance triple mappings for each linguistic triple. The TSS chooses if possible the ontologies that better cover the user query and domain. Then, it extracts, by analyzing the ontology relationships, a small set of ontologies that jointly covers the user query and contains enough information to generate an answer to the question.
- The Merging and Ranking Component generates the final answers. This component fuses similar information together and ranks the answers in terms of their relevance to the query at hand.
PowerAqua? provides a plug-in specification mechanism that supports a common API to query ontologies residing on different repositories. Currently plug-ins are available for Sesame versions 1 and 2 (www.openrdf.org), the Watson semantic search engine ( http://watson.kmi.open.ac.uk), and Virtuoso ( http://virtuoso.openlinksw.com). PowerAqua? can be configured to query its own list of repositories (e.g. in the food and diet domain) and it offers the capability to index and add new online ontologies at any time.
Additional Material
