Beyond Aboutness
For most of its history, search has been built on aboutness. A query expresses an intent, and the search application must find results that are about that intent. That focus underlies everything from the bag-of-words model to today’s dense vector representations.
Aboutness defines relevance: a result is relevant if it is about what the query is about. This holistic framing overcomes the limitations of reductionist approaches that treat queries and documents as collections of individual words. However, aboutness alone is not sufficient. Real search requires attention to several factors that lie outside a strict interpretation of what a query is “about.”
Desirability
As I have written elsewhere, ranking and relevance are related but distinct. Relevance is a binary measure of whether a result satisfies the searcher’s need, while ranking sorts relevant results according to searcher and business objectives.
Relevance focuses on aboutness. Ranking introduces desirability. Query-independent factors such as popularity, quality, authority, and recency often matter as much or more than small differences in relevance. A product that is objectively better, safer, or more trusted should outrank one that is only a slightly closer semantic match.
Dense retrieval, a foundation of modern AI-powered search, makes this distinction even more important. Similarity scores from vector models often vary only slightly among top candidates. If desirability is not layered on top of relevance, the system risks overvaluing tiny differences in cosine similarity and undervaluing large differences in desirability.
Broad or Ambiguous Query Intent
Aboutness assumes a clear intent. Many queries do not communicate one. They are broad or ambiguous.
Broad queries are unambiguous but underspecified. A query like “shirts” spans men’s and women’s shirts, multiple styles, and many fits, colors, and materials. The intent is coherent but spans a diverse result space.
Ambiguous queries differ because they straddle domains. “Mixers” may refer to kitchen appliances, audio equipment, or industrial machines. “Jordan” may refer to a country, a brand, an athlete, or a river.
Ambiguous queries require clarification because the system cannot determine the searcher’s intent with confidence. Broad queries require guided navigation, often through category or facet refinements.
Multi-Intent Documents
Broad or ambiguous queries challenge aboutness on the query side. Multi-intent documents challenge it on the index side. A multi-intent document does not communicate a single focus either.
Long-form guides that combine several topics and articles that blend technical, historical, and practical content are common examples. Representing such documents with a single vector collapses distinct subtopics into one embedding.
Approaches that address this include segmentation, which isolates coherent sections; ColBERT, which represents documents with multiple contextualized embeddings; and SPLADE, which learns sparse token-level weights that capture multiple semantic aspects. These approaches recognize that many documents cannot be reduced to a single representation.
If a document is about more than one thing, it should match more than one cluster of queries.
Fulfillment
Keyword search conditioned users to expect results that look like their queries. A query for “shoes” returns shoes. That expectation breaks down when the searcher expresses an intent that is not literally the form of the desired result.
Results that fulfill an intent do not always mirror the wording of that intent. Answers are not the same as questions. A search for a person’s name is often an attempt to contact that person. A search for a location is often a request for directions. A factual question calls for an answer, not a paraphrase of the question.
The aboutness of a query and the fulfillment of that query often diverge. Bridging that gap is a critical part of modern search.
Summary: Search is About Satisfaction
Relevance is the primary directive of search, and relevance is rooted in aboutness. However, the ultimate goal of search is to satisfy users and support positive business outcomes. Achieving that goal requires attention to desirability, the handling of broad or ambiguous queries, the representation of multi-intent documents, and the translation required to fulfill intent.
Aboutness remains central, but satisfaction is the final measure of success.
