Discover the key differences between traditional keyword search and semantic search—and learn how new technology is transforming digital collections discovery.
Keyword search finds records by matching the words a researcher enters. Semantic search looks for records that are related in meaning, even when they use different terminology.
The growing interest in semantic search is being driven partly by changing user expectations. Everyday experiences with search engines and other digital tools have made people more accustomed to asking questions in natural language and receiving results based on context, not only exact wording.
Digital collections do not always work that way. Historical records may use unfamiliar, outdated, or highly specific terminology, which means a researcher can ask a relevant question and still miss useful material.
Semantic search offers another pathway. It can help connect a researcher’s modern wording with records that express the same idea differently, while keyword search continues to support precise searches for known names, dates, phrases, and historical terms.
Keyword search relies on exact word matching. It retrieves content that includes the specific terms entered by the user—without interpreting context or intent. This makes it highly effective for targeted searches involving names, dates, or direct quotes—situations where the user knows precisely what they're looking for.
Its limitations become more apparent when relevant records use different language from the query. Historical terminology can vary between periods, regions, communities, and publications. A researcher may therefore need to try several related terms to uncover the full range of available material
Its limitations become more apparent when relevant records use different language from the query. For example, a search for “Impact of 1918 New Zealand influenza pandemic” would return only results containing those exact words. It would miss valuable content that uses different but still highly relevant terminology, such as “Spanish flu”. Additionally, it treats “impact” as a literal term rather than understanding its intended meaning. In this way, keyword search can narrow or even undermine the relevance of the results returned.
Semantic search uses natural language processing and machine learning to interpret the meaning behind a search query—not just the literal words used. Unlike keyword search, which looks for exact matches, semantic search can identify synonyms, related concepts, and contextual relevance–helping it better understand user intent, even when not explicitly stated.
Therefore, returning to the earlier example search query—“Impact of the 1918 New Zealand influenza pandemic”—a semantic search would likely return far more nuanced and relevant results. It would not only account for related phrasing such as “black flu” and "Spanish flu", but also surface content discussing economic consequences, social disruption, and public health responses related to the pandemic.
Semantic search does not replace keyword search. It provides a different route into a collection.
There’s no doubt that semantic search offers a smarter, more intuitive way to navigate historical content—unlocking new opportunities to uncover hidden narratives, unexpected connections, and deeper insights from the past.
However, keyword search remains an essential tool, particularly in areas where precision is critical. It continues to play a vital role in:
Genealogy research where family history is traced through specific names, dates, and places—especially in resources like newspaper notices, obituaries, passenger lists, and electoral rolls.
Informing how certain terms appeared and evolved in mainstream conversation. For example, the word “quarantine” shows up across multiple pandemics, but its framing and public perception likely shifted dramatically between 1918 and 2020.
Remaining connected with outdated terminology. For example, the term “lunatic”, once commonly used in medical and legal documents related to mental health, is no longer considered appropriate today. Keyword search helps surface these historical references, providing valuable insight into how different conditions were historically understood and represented.
Identifying when ideas first emerged and changed. For example, when did the concept of “climate change” first appear in public records? Keyword search can help pinpoint its earliest mentions and trace how opinions and attitudes toward it have shifted across different time periods.
Veridian’s current search experience is centered on keyword search, which remains essential for finding known names, dates, places, phrases, and historical terms.
As part of our beta testing, we are also exploring semantic search options, including RAG Search (Retrieval Augmented Generation), that could be combined with existing keyword search in a hybrid model. This approach brings together the precision of keyword matching and the broader contextual understanding of semantic search.
In practice, hybrid search will help users find exactly what they are looking for while also surfacing relevant records and connections they may not have discovered through keyword search alone.