Kino : Making Semantic Annotations Easier
Download as PPT, PDF0 likes467 views
Kino is a search engine and semantic annotation tool specialized for scientific domains. It allows users to semantically annotate services, documents, and other web resources using SA-REST. Key features include faceted search integrated with NCBO ontologies, a browser plugin for annotation, and a backend server. The document describes two use cases: 1) Finding web services for gene-related tasks using a catalog more accurately by accounting for synonyms, and 2) Annotating scientific papers by looking up terms via a browser plugin to complete and submit annotations more easily. The system works by indexing annotations from the NCBO ontology repository and REST services via its backend and enabling faceted search and annotation via its browser plugin and other
Kino : Making Semantic Annotations Easier
- 1. Kino : Making Semantic Annotations Easier Ajith Ranabahu # , Priti Parikh # , Maryam Panahiazar # , Amit Sheth # and Flora Logan-Klumpler* # Ohio Center of Excellence in Knowledge-enabled Computing ( kno.e.sis ) * Sanger Institute and University of Cambridge, UK
- 2. What is Kino ? A search engine of sort currently specialized for scientific domains Semantic Annotation (using SA-REST) for services and documents (or any Web resources) Indexing system Uses SA-REST, Faceted search integrates with NCBO Tools include browser plugin & back-end server
- 3. Two use cases service lookups: eg: BioCatalogue returns about 75 Web services for the search term “gene prediction”. However, it returns only 20 Web services for the term “gene finding”, even though gene prediction and gene finding are synonyms. supporting the whole cycle of annotation and curation at Sangers
- 4. Use case 1 : Service lookups for gene data Find Web services using a service catalog for these operations Tagging is not consistent, keyword based Lookups are not accurate and do not take synonyms into account ! Genome Genome Sequencing Sequenced Genome Gene Prediction Predicted Gene Gene sequence Comparison Cellular location determination Function prediction Detailed Information about the Gene
- 5. Use case 2 : Scientific Document Annotation Read Published Paper Add notes via a browser plugin Look up terms and update the notes Are the annotations complete ? Submit annotations to the database yes no Long and cumbersome cycle of term lookup and updates. This done by scientists – is there a way to make this process easier for them ? Courtesy of Sanger Institute UK
- 6. How does it work ? NCBO Ontology Access API NCBO Ontology Repository Kino Search API SOLRJ Kino Index API SOLR Web Interface Lucene Index Kino Browser Plugin Web Pages Kino Web Front-end Other Front -ends NCBO REST Service Kino Back-end Kino browser based annotation Kino Search Interfaces
- 7. What did we learn? Back-end automatically including even the synonyms make a significant improvement in search recall. Having integrated tooling enables faster cycles in scientific literature annotations.
- 8. Check out our poster for more details Video demo at http://wiki.knoesis.org/index.php/Kino Paper at http://knoesis.org/library/resource.php?id=1553
Editor's Notes
- #2: Kino is a pearl hunter - he goes in depths and finds pearls in this novel: http://en.wikipedia.org/wiki/The_Pearl_%28novel%29
- #3: Kino is a semantic annotation and indexing system for scientific domains. It is built around SA-REST and our own APIhut technology as the backend, using the NCBO RESTful APIs. The ease of annotation comes with the use of a browser plugin and the integrated server back-end which can directly receive annotated content. Any Web document can be annotated and published to a faceted index easily. Since the NCBO repositories are tightly integrated, annotators can directly lookup the matching concepts and apply the annotation to the document.
- #4: We had two usecase – one with service lookups and one with scientific document annotations In service lookup case, we learnt that there is a significant advantage in the case of using even the synonyms associated with the concept. It’s clear that if more sophisticated information can be used from the ontologies at the time of indexing, it would definitely be helpful for better recall. For document annotation – a real use case from Sanger, the whole cycle of annotation and curation can now be reduced significantly by this integrated tooling. The current workflow is very labor intensive but using a toolset like Kino reduces the effort and time required significantly.
- #5: The lookup of services is slow and inaccurate – each tag or keyword put manually is subjected to non-uniformity and cultural biases. Often a scientist may have to sift through a number of services to find the needed one.
- #6: This document annotation process , currently in use by Sanger, has many cycles between the annotations. The cause for this is that the document is not directly annotated but notes are added instead and the concepts are looked up later in the cycle based on the notes. The lack of tool integration puts a heavy burden on the scientists and also the process is prone to much errors.
- #7: With the first animation : The annotation process starts by highlighting words, phrases or links in the web documents. When the appropriate menus are clicked, the user is presented with a the search results from NCBO and well as the options to browse for a desired concept. When the concept is selected, the in-memory document is modified to attach the annotation. With the second animation: Once the document is fully annotated, it can be submitted via the plugin, to the kino back-end instance. The index api filters out the annotations, makes lookups for synonyms for the concepts via NCBO and finally indexes the document using the Apache SOLR engine. The complete set of front-end and back-end functions are exposed via APIs so that programs (other than the plugin and the front-end) can make use of the kino system to directly submit and retrieve content. Third animation: During lookups, the users can use the AjaX front-end to lookup documents using concept names, synonyms with concept names or just plain keywords to look up documents. They can filter the results using facets and restrict the search for specific annotations.
- #8: We had two usecase – one with service lookups and one with scientific document annotations In service lookup case, we learnt that there is a significant advantage in the case of using even the synonyms associated with the concept. It’s clear that if more sophisticated information can be used from the ontologies at the time of indexing, it would definitely be helpful for better recall. For document annotation – a real use case from Sanger, the whole cycle of annotation and curation can now be reduced significantly by this integrated tooling. The current workflow is very labor intensive but using a toolset like Kino reduces the effort and time required significantly.
- #9: Take a look at our poster and also the videos in the Kino page. Paper is in the proceeding or at the link given.