Does a morphological dictionary explain word meanings?

It may include some semantic information, but its defining focus is the relationship between forms and morphological analyses.

Can one form have several tags?

Yes. Different grammatical uses can share a spelling, and a form may also correspond to different lemmas.

Can the same resource analyze and generate?

Some implementations support both directions. The available operations and treatment of unknown forms depend on the resource and software.

AI Glossary / Natural language and speech / Beginner

Morphological dictionary

Definition

A morphological dictionary is a lexical resource that connects word forms with morphological analyses, such as a lemma and grammatical features. It can help identify which forms belong to the same lexical item and what properties a form may express. Some implementations also support generating forms from specified lemmas and features.

What it means

A dictionary headword is not always the form encountered in a sentence. Words may vary with number, tense, case, person, or other categories, depending on the language. A morphological dictionary records or encodes these relationships.
Its focus differs from a general dictionary's explanation of meaning. A morphological entry may identify a form as a plural noun or past-tense verb without providing a full semantic definition.
One surface form can have several analyses. The resource can supply those possibilities, but selecting the analysis intended in a particular sentence usually requires contextual processing. Morphological lookup and contextual tagging are related operations, not automatically the same one.

How it works

Define the language's relevant feature categories and a documented tag scheme. Connect forms to lemmas and feature combinations, either through explicit records or a representation that encodes morphological patterns.
For analysis, look up an observed form and return compatible lemma-and-feature candidates. Some systems supplement dictionary lookup with a guesser for forms absent from the resource; guessed results should be distinguishable from recorded ones where that matters.
For generation, start with a lemma and requested features, then retrieve or construct supported forms. Ambiguity or missing coverage can produce several results or none.
Use context when a downstream task needs one analysis. Verify that the dictionary's tags and the application's interpretation agree, especially when converting between annotation schemes.

Three practical examples

A search system groups inflected forms under a common lemma when this improves retrieval for its language and domain.
A grammatical analysis tool obtains candidate features for a word before a contextual tagger selects among them.
A language-learning application displays supported inflected forms for a lemma, with the relevant grammatical categories attached.

Think of it this way

Think of an index connecting different versions of a product to a common model and a list of attributes. Looking up a version reveals how it relates to the model.
The analogy has limits: a written word form can correspond to several lexical items or feature combinations. The spelling alone may not uniquely identify the intended entry, and morphology is not simply a catalog of suffixes.

Worked example

Use an original simplified English example with the form “saw.” A miniature morphological dictionary contains at least these three analyses: lemma see with past-tense verb features; lemma saw with singular-noun features; and lemma saw with base-form verb features.
In “Yesterday Mina saw the ferry,” context favors the past-tense analysis of see. In “Mina stored the saw,” the noun analysis fits. In “Please saw the plank,” the verb meaning cut with a saw fits.
The lookup therefore returns three candidates in this deliberately limited resource, not three meanings asserted simultaneously in each sentence. A larger dictionary and more detailed feature scheme could distinguish additional analyses.
In the reverse direction, generating the past-tense form of see returns saw under the toy mapping. Merely stripping letters from saw would not recover see, which illustrates why irregular relationships need suitable lexical information or modeling.

Illustrative case study

In a fictional project, Linden Search normalizes maintenance documents by lemma. An early implementation always selects the first dictionary analysis and consequently mixes a tool named “saw” with some occurrences of the verb “see.”
The team preserves candidate analyses until sentence context is available. It evaluates ambiguous forms separately from straightforward regular inflections and records cases where the resource has no suitable entry.
The lesson is that a morphological dictionary supplies structured evidence, while application behavior depends on how that evidence is used. This invented case claims no measured retrieval improvement.

Common misconceptions and mistakes

Do not assume that each surface form has exactly one lemma or tag.
Do not confuse lemmatization with simple suffix removal. Irregular forms and ambiguous spellings require more than mechanical trimming.
Do not assume all morphological dictionaries explicitly list every possible form; some encode patterns or use compact representations.
Do not treat a missing entry as proof that a form is invalid. New words, names, domain terms, and coverage limits matter.
Do not interpret a candidate analysis as a contextually confirmed one, and do not silently merge dictionary-backed results with guesses when provenance affects review.

Business use

For a business working with languages that have substantial inflection, morphological resources can support search, text analysis, and language-learning functions. The practical question is whether the resource covers the forms and grammatical distinctions the application encounters.
Test irregular forms, ambiguous spellings, specialist vocabulary, and names. Check compatibility between dictionary tags and the rest of the processing pipeline before relying on feature-based rules.
Retain resource versions and document any local extensions. Measure downstream task behavior rather than assuming that a large number of stored forms guarantees useful analysis in the intended domain.

Check your understanding

Why does finding the past-tense analysis see for “saw” not justify using it in “Mina stored the saw”?

Answer: Lookup identifies a possible analysis of the spelling. The sentence uses saw as a noun, so contextual selection is needed to avoid assigning an available but inappropriate candidate.

Related terms

Lemma
Paradigm (experimental)
Part-of-speech tagging
Inflection
Morphological analysis
Morphological generation
Paradigm

Learn next

Create a tiny table with columns for surface form, lemma, and grammatical features. Include one regular inflection, one irregular form, and one spelling with multiple analyses.
Then test both lookup directions and identify which questions cannot be answered without a sentence. Inspect MorfFlex and MorphoDiTa documentation as a concrete example, while remembering that their Czech tag conventions are not universal.

Lemma

Sources

Charles University, MorfFlex CZ official resource description: https://ufal.mff.cuni.cz/morfflex Charles University, MorphoDiTa official API tutorial, morphological dictionary analysis and generation: https://ufal.mff.cuni.cz/cs/morphodita/api-tutorial The English saw examples use a deliberately simplified original resource, not outputs from the Czech system. Independent review: official analysis, generation, candidate-return and guesser documentation consulted.

Previous: Morphogenetic robotics Next: Morphological gradient