HEAL Semantic Search Resources
This page contains resources on HEAL Semantic Search including FAQs, tutorials, and other instructional materials. HEAL investigators interested in submitting their study information should visit the "Submit Your Study Information" section.
Click here to navigate to the HEAL Semantic Search interface.
FAQs
What is semantic search?
Semantic search is “search with meaning.” Search results are enhanced by incorporating knowledge about the search concepts, including synonyms and relationships to other concepts. For example, a semantic search for “high heart rate” would find results containing those three words or their variants (such as “frequency” instead of “rate”) as well as synonyms (such as “tachycardia” or “heart racing”) and related concepts (such as “arrhythmia”).
Why should I use HEAL Semantic Search (HSS)?
HSS provides a smart way to search disparate HEAL datasets to uncover related studies, variables, Common Data Elements (CDEs), and research concepts. Use HSS if you are looking for studies, variables, or Common Data Elements (CDEs) related to biomedical concepts, rather than a particular researcher, project or dataset.
When should I use HEAL Semantic Search (HSS), and when should I use the HEAL Platform’s search function?
HSS and HEAL Platform search are complementary search functions providing quick and effective HEAL study data search. The table below highlights HSS and HEAL Platform functions and how each can be used to dive deeper with the other. Our goal is to eventually integrate these two tools so users can query the HEAL Data Ecosystem seamlessly using either.
These tools complement each, enhancing result exploration. The HEAL Platform delves deeper into the studies and variables uncovered through HSS. Meanwhile, HSS aids in discovering related information pertaining to studies, variables, or concepts of interest identified within the HEAL Platform.
HEAL Platform Search
- Discover HEAL-funded projects
- Find researchers (PIs) working on these projects
- Conduct keyword search
- View study metadata
- Discover datasets that might be of interest
HEAL Semantic Search
- Find connections between biomedical concepts
- Expand the search by looking at synonyms and related terms
- Identify HEAL studies, variables, and Common Data Elements (CDEs) related to search terms and search results
- Explore knowledge graphs
What is a “concept” in HEAL Semantic Search (HSS)?
In HSS, a concept is a named entity, defined in an ontology or other formal knowledge source. Concepts have a unique identifier, preferred label, and a precise definition. For example, chronic pain is identified as HP:0012532, which is labeled “chronic pain” and is defined as “Persistent pain, usually defined as pain that has lasted longer than 3 to 6 months.” In HSS, concepts include categories such as disease, phenotype, biological process, or anatomical entity.
Can I search multiple terms at the same time on HEAL Semantic Search (HSS), like “knee pain AND obesity?”
Yes, you can search for multiple terms by using the "+" operator. For example, to search for chronic pain and back, enter "chronic pain + back" into the search bar.
How does HEAL Semantic Search (HSS) find related concepts or studies for a search term?
HSS uses named-entity recognition tools to identify biomedical concepts in the variable names, descriptions, and measurement instruments (question text, permissible values). When you search for a term, it will start by looking for concepts, studies, CDEs, and variables that contain your search term in its name, description or list of synonyms. It will also use relationships between concepts from the ROBOKOP Knowledge Graph to suggest additional items that might be related to your search term.
How does HEAL Semantic Search (HSS) rank or score search results?
Search results are scored based on their search term relevance/closeness. Higher scoring, more relevant results are first. Further down the search results screen, the returned results become more obscure, are further from the search term, and score lower. HSS understands related biomedical concepts and search synonyms, and it uses these to prioritize search results.
Where does HEAL Semantic Search (HSS) ingest data/information from?
HSS ingests HEAL study variable-level metadata (VLMD) and integrates them in existing knowledge graphs (ROBOKOP) to combine distinct concepts and disparate data types. You can learn more about metadata in the HEAL Data Ecosystem on the Metadata page.
In addition, HSS includes common data element (CDE) information collected by the HEAL CDE program, so we can point you to relevant CDEs.
Why is it critical that investigators share their data dictionaries with the HEAL Stewards for HEAL Semantic Search (HSS) development?
HSS quality and robustness depends on ingested data dictionary volume. More ingested data dictionaries will result in richer, more relevant user search results, since HSS indexes data dictionary information to well-established biological ontologies (including biomedical concepts and their synonyms) to connect biomedical concepts.
Please note, data dictionaries contain information about study variables, not the data itself. Sharing your data dictionary allows others to discover your study and may help identify potential collaborators and novel research questions. It also contributes to NIH’s overarching goal of making HEAL data FAIR (Findable, Accessible, Interoperable, and Reusable).
A concept I searched for doesn’t have many results, but I know the data is out there. Why doesn’t HEAL Semantic Search (HSS) know about it?
HSS returns search results related to the data dictionaries it ingests, indexes, and builds knowledge graphs for. If HSS hasn’t ingested related information, it cannot connect between your search term and what is available. We are always looking for new data dictionaries to add, improving search results. We encourage you to submit your own data dictionaries for HSS through the HEAL Platform’s variable-level metadata (VLMD) submission system or by emailing the HEAL Stewards. If you’re sure that the concept you searched for must exist in HSS, please report this issue on the HSS Support Portal.
Why does HEAL Semantic Search (HSS) return results in the order it does?
HSS returns results based on their relevance to your search term. Usually, the result order is because:
- The concept name or description includes the search term.
- It knows the search term is a synonym for the concept name, e.g., “arthralgia” instead of “joint pain.”
- A concept related to your search term is also displayed (e.g., bleomycin is displayed when searching for “joint” because the knowledge graph shows it can cause arthritic joint disease as a side-effect).
How can I contribute my study’s information to HEAL Semantic Search (HSS)? Are there any legal restrictions on what I can submit?
Please submit your data dictionaries through the HEAL Platform’s variable-level metadata (VLMD) submission system or by emailing the HEAL Stewards. Your study does not have to be complete; we are happy to accept any data dictionary, even if it is not finalized. Data dictionaries inherently should not contain sensitive information such as personal health information (PHI) or personally identifiable information (PII). For more information about sensitive data, click here.
How can I share feedback or suggestions for future enhancements for HEAL Semantic Search (HSS)?
We welcome your suggestions to build a better HSS! Please visit our Help Portal and submit an improvement ticket.