The Hidden Vocabulary of Hotels: Why Taxonomy Matters as Much as Design

A guest booking system and a guest's own words rarely speak the same language — and the tourism sector faced that problem, and largely solved it, before most other industries did.

A guest searches for "a quiet room away from the elevator." The hotel's booking system classifies it as "Deluxe Room, Floor 3 to 5, Standard View." Nobody made a mistake. It is simply that the person searching and the person organising the information speak two different languages. This is a problem as old as libraries, and the tourism sector faced it before many others did, with results that are worth rediscovering.

Controlled vocabularies as the infrastructure of information architecture

Before getting into the tourism case, it is worth framing the problem within the discipline it belongs to: Information Architecture. Information Architecture designs and organises information environments, physical and digital, structuring their content in a way that is semantically coherent and logically understandable, so that access is simple and meaningful for the people who use it. It is concerned not only with consistency within a single system, but also with maintaining it across different platforms, contexts, and modes of use, preserving structures of meaning over time.

Within this discipline, a controlled vocabulary functions as an umbrella concept: it defines a predefined and authorised list of terms, words, or phrases, whose meaning is clear and unambiguous within a specific domain, and it is meant to be used for indexing, cataloguing, archiving, and retrieving information. Under this umbrella sit tools of very different complexity, from simple lexicons and glossaries all the way to taxonomies, thesauri, and ontologies. Moving from one tool to the next is not just a jump in technical sophistication. It is a progressive enrichment of the system's ability to capture and make explicit the meaning of concepts, turning a plain list of words into an actual map of knowledge for a domain.

Taxonomies, for instance, resolve ambiguity through the position of a term within a hierarchical tree that runs from general to specific (the relevant standard is ANSI/NISO Z39.19), but they remain weak on the synonym front: a taxonomy will rarely tell you that "double room," "matrimonial room," and "queen room" are the same thing. This is where the thesaurus comes in, the tool at the centre of this article: a controlled, structured vocabulary in which concepts are represented by terms organised so that the relationships between concepts are made explicit, and in which preferred terms are accompanied by entry points for synonyms and near synonyms. More complex than a taxonomy, capable of handling synonymy while also making explicit relationships that a hierarchy alone cannot express.

A precedent more solid than it sounds

In 2001 the World Tourism Organization published, together with the French Secretariat of State for Tourism, the Thesaurus on Tourism and Leisure Activities: over eight hundred pages, three languages, twenty years of work behind it. It is not a dictionary. It is a controlled vocabulary, a structure of preferred terms, synonyms, hierarchical and associative relationships, built to ensure that whoever indexes tourism content and whoever searches for it ultimately use the same language.

The method behind building such tools is codified in the ISO 25964 standard, which precisely distinguishes two concepts that everyday language tends to blur: the concept (the unit of thought) and the term (the label we give it). A "double room," a "matrimonial room," and a "queen room" can all be the same concept expressed through three different terms, and a well-built domain vocabulary knows which one to prefer, which ones to treat as synonyms, and how to organise them into a coherent hierarchy.

More recently, the UNWTO thesaurus itself was converted into SKOS format for the semantic web, with readable identifiers for each concept linked to DBpedia — a sign that this is not library archaeology but infrastructure still very much alive for search and data interoperability in tourism.

From documentary thesaurus to hotel ontology

The hospitality sector has developed its own, more targeted variant: not a generic tourism thesaurus, but full domain ontologies designed specifically for hotel search systems. One instructive example is an ontology for a semantic hotel search system built around three levels: hotels and their relationships to context (neighbourhood, attractions, transport), the internal components of a hotel with their respective properties (room types, services, amenities), and — the most interesting part — the subjective and ambiguous terms travellers actually use when searching for or evaluating a property: "nice view," "convenient location," "quiet."

This third level is exactly what most hotel websites ignore. Search filters are built on the vocabulary of the management system (the PMS), not on the vocabulary of the guest. The result is a search experience that works well for people already familiar with hotel nomenclature, and frustrates everyone else — which is, by definition, the majority of a site's visitors.

The relational structure of a thesaurus: equivalence, hierarchy, association

The ISO 25964 definition of a thesaurus actually describes three types of semantic relationship, and it is precisely the combination of these three that makes the difference compared to a taxonomy.

The equivalence relationship handles synonyms and near synonyms through the pair of codes USE and UF, Used For: if "matrimonial room" points to "double room" through a USE instruction, the reverse path, from "double room" to "matrimonial room," is expressed with UF. This is the mechanism that lets a guest search for "queen room" and still find the matrimonial rooms in the catalogue, without anyone having to duplicate content under three different labels.

The hierarchical relationship, expressed through BT and NT, organises concepts according to relationships of subordination: genus and species, whole and part, class and instance. This is the same logic taxonomies use, but here it coexists with the other two relationships instead of standing alone.

The associative relationship, marked RT, links concepts that are semantically close but not connected hierarchically. It is the least rigid relationship and, somewhat paradoxically, the most useful one for anyone browsing a site without knowing exactly what they are looking for: it supports search models such as berrypicking, first described by Marcia Bates in 1989, in which the user does not formulate a precise query from the start, but gathers information a little at a time, moving laterally through the conceptual network and following connections that are not obvious but are relevant. A good set of associative relationships in a hotel thesaurus is what lets a guest who started from "breakfast" arrive, without having to search for it explicitly, at the restaurant hours page or at options for food intolerances.

What this means in practice for a digital project

Translated into operational terms, for anyone working on a hotel website, a booking engine, or a destination portal, this line of research points to a few very concrete things.

• A content audit should always include a vocabulary audit: which terms the PMS uses, which terms marketing uses, and which terms guests actually use in reviews and customer service requests. The three lists rarely match, and the gaps between them are often the hidden cause of low conversion rates on internal filters and search engines.
• Search filters and facets deserve the same methodological rigour as a thesaurus: clear preferred terms, synonyms mapped under the same concept, an explicit hierarchy between broad categories and subcategories. This is not an academic exercise — it decides whether a guest finds the room they are looking for in three clicks or abandons the search.
• Multilingual consistency is critical for international hospitality: a well-built controlled vocabulary does not translate word for word, it maps equivalent concepts across languages — exactly the problem the UNWTO thesaurus solved for documentary indexing, and that a multilingual hotel website must solve for navigation.

Why this is a design problem, not just a data problem

It is tempting to dismiss all of this as a technical matter for librarians or data engineers. But vocabulary structure is, quite literally, the structure of the information a guest navigates: it precedes visual design, shapes it, and often makes or breaks it. An elegant interface built on a confused taxonomy still produces a confusing experience — only a harder one to diagnose, because the problem is invisible and only shows up in abandonment numbers.

Anyone designing digital experiences for hospitality would do well to treat content taxonomy as a project phase in its own right, not as a byproduct of the CMS or PMS structure. It is a task that resembles rigorous card sorting and content auditing more than a naming exercise, but it is, at bottom, the same discipline that produced an eight-hundred-page thesaurus twenty years ago: understanding what people actually call things, before deciding how to organise them.

The case of "invisible" filters

It is worth pausing on one specific point, because it is where controlled vocabulary theory most often collides with the practice of online booking: filters the system cannot offer because it has no concept to represent them. A guest searching for a "quiet" room or a "good view" is using attributes that, in the typical data model of a PMS, simply do not exist as structured fields. At best they exist as scattered phrases buried in reviews.

The more mature hotel ontologies address exactly this: alongside easily structured concepts (room type, floor, size, amenities) they explicitly model a category of subjective and ambiguous terms — the ones guests actually use to evaluate a property. Treating them as noise to ignore, rather than as a legitimate layer of the vocabulary to map (for instance linking "quiet" to a set of objective properties: far from the elevator, high floor, courtyard side), is one of the most common causes of internal searches that return zero results even when the right inventory exists in the catalogue.

The same principle applies, on a larger scale, to destination portals: a site that lists "things to do" by administrative category (museums, restaurants, events) instead of by visitor intent (romantic, with children, in a few hours, free) is imposing its own internal organisational scheme on people who simply do not know it and should not have to learn it.

A practical exercise before any redesign

One concrete way to bring this work into a project, even before opening a design tool, is to build, even in a lightweight form, three parallel lists for the same content domain: the terms used by the management system, the terms used in marketing materials, and the terms that actually appear in guest requests and reviews. The simple exercise of laying them side by side, concept by concept, almost always reveals where unmapped synonyms and missing categories are hiding — an input no visual walkthrough of the site can provide on its own.

There is no need to build a full ISO 25964-style thesaurus for every project: the discipline that standard formalises — distinguishing concepts from terms, choosing a preferred label, explicitly mapping synonyms, building coherent hierarchies — is applicable at a smaller scale even to a simple content audit, and it is often the difference between a filter system a guest understands on the first try and one that takes several attempts.

How Esario applies these tools

At Esario, controlled vocabulary work does not stay theoretical. It is one of the first things we tackle when working on the information architecture of a hotel website or a tourism portal, before we even start sketching wireframes or interfaces. The starting point is almost always a terminology audit, carried out alongside the content audit: we map the vocabulary used by the client's PMS or CMS, the vocabulary used in marketing materials, and the vocabulary that actually shows up in reviews, customer service requests, and internal search logs when available. It is the fastest way to spot where the system's language and the guests' language diverge, and to work out which synonyms need to be explicitly mapped before touching the interface at all.

This work almost always produces a lightweight domain vocabulary: not a full ISO 25964-style thesaurus, which would require resources hard to justify for a single hotel project, but a structure that applies its core principles — one preferred term per concept, synonyms explicitly mapped following the USE and UF logic, a coherent hierarchy between categories and subcategories, and the associative relationships most useful for guiding search. This is the tool that, translated into search filters and booking engine facets, decides whether a guest finds the room they are looking for in three clicks or abandons the search.

For multilingual projects — which in international hospitality and tourism are the norm rather than the exception — this terminology work extends to mapping equivalent concepts across languages: not word-for-word translations, but verified conceptual correspondences, the same principle the UNWTO thesaurus applies on an international scale. It is an investment that pays off quickly in navigation consistency and relevant search results. The same approach, adapted to each domain, guides our Information Architecture work in the other sectors we operate in as well, from e-commerce to finance.

The methodology step by step: what ISO 25964 actually says

It is worth going deeper into the method, because this is where the standard stops being theory and becomes a replicable procedure. A public application guide that closely follows ISO 25964 (produced by the MoTIF project of the Digital Repository of Ireland, which built a pilot thesaurus on Irish folklore as a testbed) breaks the work down into four main phases, preceded by a planning phase.

Planning

Planning comes first: the purpose of the vocabulary, who will use it, which content it covers, and who is responsible for maintaining it over time all need to be written down. Translated into a content strategy project for a hotel website, this is the step where you decide whether the vocabulary serves internal search, site navigation, or both, and who is in charge of updating it when services or room types change.

Term collection

The first operational phase is term collection: it starts with a content audit of existing vocabularies (in the case of a hotel: PMS categories, CMS tags, booking engine labels), supplemented by secondary sources such as reviews, customer service requests, and internal search logs. The guide suggests starting with a manageable number of terms, between one hundred and five hundred, rather than attempting an exhaustive list from the outset.

Selection

The second phase is selection: the collected terms are grouped by meaning, a preferred term is chosen for each group, and the others are treated as synonyms. This is where a decision comes in that ISO 25964 treats with particular care: what to do with compound concepts, such as "double room with sea view." The standard lists five options, from breaking the concept down into simpler, linked terms (for example "double room" plus "sea view"), to keeping it intact when splitting it would lose meaning or create ambiguity. The practical rule that follows from this is easy to apply even outside a formal thesaurus: split a compound term only if combining the simple terms creates no ambiguity, and never split it into more than two components, because consistency quickly becomes unmanageable.

Facet analysis

The third phase is facet analysis, the most common method for building a thesaurus's hierarchical structure, and it is rooted in the theory of faceted classification developed by S. R. Ranganathan in the 1930s. Instead of starting from a classification by subject, which tends to mix heterogeneous categories under the same heading, Ranganathan proposed breaking a domain down into five fundamental categories, known by the acronym PMEST.

A simple but effective test the standard suggests for checking whether a term sits in the correct hierarchical position is the so-called "all and some test": if all dogs are animals and only some animals are dogs, then "dog" correctly sits under "animal." Applied to a hypothetical "kennel" under "dog," the test fails, because not all kennels are dogs.

Associative relationships

The fourth phase adds the associative relationships described earlier, the ones marked with the RT code. The standard recommends building these last, after the hierarchies are complete, because it is precisely the relationships between different hierarchies that are most useful to someone searching for information that is related but not obviously nearby, following the berrypicking logic already mentioned.

None of this requires building a full thesaurus for every hotel project. What it requires is treating vocabulary as a design decision rather than an accident of whichever system happens to structure the content first — the same discipline, at a smaller scale, behind an eight-hundred-page reference built to make sure indexers and searchers speak the same language.