| Hi John, All good things come to an end and this will be the last Q&A from the PoolParty Summit 2023. Whether you attended or couldn't make it, this is a great opportunity to review the sessions and get an overview of the questions and answers that were triggered during the live event. This time we share with you "Taxonomy Linking" hosted by Heather Hedden,
Product Communication Representative at Semantic Web Company and Donna Popky, Senior Taxonomy and Information Architecture Specialist at Harvard Business School. 1. How do you ensure you can move forward with your project rather than spending time constantly refining where you think something is in the taxonomy? Heather gave the example of red wine being similar rather than narrower - what are good tips to ""accept"" or ""be happy"" with your taxonomy and work with it rather than constantly refine it? Donna answers: Test it with card sorting or tree testing. There are several good tools for this. Also – remember that a taxonomy is a living thing, it’s never final. You can launch it and use analytics to see how users are clicking through it and you should always think about changes based on user behavior and feedback, new content, new concepts, new language, and dead concepts. Heather answers: You can move forward before a taxonomy is "finalized." You should do some evaluation of a taxonomy, to make sure it is thorough enough and is of good quality and some testing for tagging and retrieval, but such testing is only a sample. A taxonomy can be more improved after it has gone into production, and should be periodically checked over time.
2. How many people are involved in creating and maintaining the taxonomies and managing the hub, search, etc.? Donna anwers: At HBS we have an Information Management Team of 5 members that work with content, metadata, our KG, and our ontology. We work closely with the tech team of four members for implementation. Heather answers: While there could be only one active taxonomist, even that one person needs to get input from others, so at the smallest there could be a team of three or so. In larger, dynamic taxonomies there could be a team of several people editing a taxonomy. It really depends.
3. Can you link ontologies the same way as taxonomies in PoolParty? Donna answers: You can have multiple custom ontologies and pick and choose classes and attributes from all of them in custom schemes – so basically, yes.
Heather answers: You can link ontologies in PoolParty, but it is usually done differently. The linking feature in PoolParty does support project linking by means other than SKOS matching relationships to include OWL sameAs and RDFS seeAlso, but I think these are not often used, as ontologies are usually not managed as "projects" in PoolParty Thesaurus Management, but rather in PoolParty Ontology Management. In PoolParty Ontology Management you can "link" by selecting classes, relations, or attributes for reuse from an existing ontology into a new ontology, which could be a kind of mashup of different ontologies plus having new classes, relations and attributes of its own. The URIs of each class, relation, and attribute specify the original ontology source. So, this can be considered a kind of linking. It's like linked open data.
4. Heather, how do you utilize Batch Linking when one of your Taxonomy projects is ingested through Unified Views? Heather answers: The method of importing a taxonomy, whether through UnifiedViews, RDF import, or tabular import, does not impact how batch linking is done. Batch linking requires that two taxonomy projects reside in the same PoolParty server, although one or both could be imported rather than created in PoolParty.
5. Is poolparty smart enough to advise the user that this concept should be exposed as an ontology and not a taxonomy or vice versa via the discovery reasoning process, seems like a lot of things are manually driven where the user needs to know how to use the platform to make best use of it, how does it scale to large amount of conceptual linkage and tagging… Donna answers: I think the user needs to know their metadata more that the platform. Ontology elements are used to classify and relate concepts. Example – you can have a skos vocabulary of topics where each topic is a concept. The class in the ontology is topics. You need to think about what best suits your content and metadata and model accordingly. Unfortunately it is not really an automatic process – you need to make and implement the decisions yourself. The system can manage vocabularies and ontologies at large scale but you need to consider what is manageable for you. Heather answers: No, PoolParty does not try to guess whether an OWL vocabulary is a schema or a vocabulary. For such decisions, domain knowledge is at least as important as data formats.
6. How do you handle close matches or related terms in PoolParty? Donna answers: The skos schema, the basis for vocabularies in PoolParty, has several match types to handle these.
Heather answers: I would recommend using closeMatch for matches that are sufficiently equivalent for the use case, content, and context, but not exact matches and thus may not be appropriate matches in a different context, if the taxonomy has a different future use. The relatedMatch is used in similar situations as related Concept, but between separate taxonomy projects. Related concept is used traditionally within a thesaurus, so would normally be used within a topical concept scheme. As for what to use between concept schemes within the same project, both relatedMatch and related are permitted, but you must be consistent, and either use only matching relationships across concept schemes or only the basic thesaurus relationships between concept schemes and not mix them.
7. @Heather Hedden sometimes you want to update the vocab in a different graph that has data. You can do this using UV, yes? Heather answers: Yes, you can.
8. How do you activate/enhance internal involvement of domain experts? Donna answers: I cultivate relationships with people in my organization that I know can help me with this. I have also been hired as a consultant in this capacity so if you don’t have local SMEs you could hire one for a particular project. Heather answers: This is more of a knowledge management question. You could say you plan to revise the taxonomy and you want their input and schedule brainstorming meetings, etc. As for PoolParty use, domain experts who are not PoolParty users for editing the taxonomy can contribute comments through the wiki front end of PoolParty or as a Read only PoolParty user.
9. How do you take a 3rd party Taxonomy, being consumed via unified views, as a Taxonomy Project so you can leverage batch linking? Meaning, Does batch linking work with taxonomies or projects imported via unified views? Heather answers: Yes there are multiple ways to import taxonomies in PoolParty, one way is via UnifiedViews. As soon as the taxonomy is imported you can use batch linking. By the way, UV could be used to do anything involving the PoolParty API. So it could be used to process RDF data that contains concept definitions, and push to PP via SPARQL or use PoolParty API calls to create concepts, etc.
10. Hi Donna, did HBS start with creating a taxonomy for one specific department and then having that as a POC to persuade other departments to create their own taxonomies? If so, what was the product for the POC that convinced the other departments to follow suit? Donna answers: I think the taxonomy that got us exposure at first was one we did for faculty research publications. These publications are considered very high value to the school and discoverability was important to a lot of people. It took a while but once it was clear that this method of content discovery could be successful, we got buy in from other groups on campus and were asked to help them out with metadata for their content. Want the full picture? Click the button below to access this session, and all of the PoolParty Summit 2023 sessions, on-demand! |
