Tekstaro de Esperanto is a searchable, curated corpus containing millions of words from foundational texts, Zamenhof, translations, literature, periodicals, and more recent writing. It can show that a form is attested, reveal its neighbors and contexts, and compare distribution across periods or sources. It cannot by itself declare a form grammatical, representative of all speakers, or appropriate in every register.
Corpus results become especially useful when combined with the analytical categories in PMEG and descriptive grammar: the corpus supplies examples and distributions, while grammatical analysis explains what contrast the distribution may encode.
Ask a corpus-sized question
A productive corpus question is specific enough to be tested. “Is this good Esperanto?” is too broad. Better questions include:
- Does the form occur at all?
- Which complement or preposition follows this verb?
- Is one variant concentrated in older texts?
- Does an expression occur across independent authors and genres?
- Is the apparent frequency caused by one repeatedly quoted work?
Mi volas ekscii ne nur ĉu la formo aperas, sed ankaŭ kiam, kie kaj kun kiu signifo ĝi aperas.
I want to learn not only whether the form occurs, but also when, where, and with what meaning it occurs.
That sentence defines a distributional inquiry. It separates attestation from interpretation.
— Ĉu vi trovis la esprimon? — Jes, sed ĉiuj kvin trafoj venas de la sama romano.
— Did you find the expression? — Yes, but all five hits come from the same novel.
Five hits are not five independent speakers. Source concentration changes the evidential value.
Know what the collection represents
Tekstaro began as an ESF-supported linguistic corpus project and has expanded in phases. Its holdings are selected and prepared texts, not a random sample of every Esperanto conversation. Written sources dominate; spontaneous speech, private chat, and some modern communities are underrepresented.
La formo estas ofta en romanoj de la frua dudeka jarcento, sed tio ne montras, kiel junuloj hodiaŭ parolas rete.
The form is frequent in novels from the early twentieth century, but that does not show how young people speak online today.
Translations also require attention. A pattern repeated in several translations from one source language may reflect translation practice. Original writing and translated writing can both provide valid Esperanto evidence, but they answer different questions.
La konstruo aperas ĉefe en tradukoj; mi do komparos ĝin kun originalaj Esperantaj tekstoj.
The construction occurs mainly in translations, so I will compare it with original Esperanto texts.
Likewise, Zamenhofian evidence is crucial for historical and foundational interpretation, yet it cannot alone establish twenty-first-century conversational register.
Search the form, then search the system
Esperanto endings make forms visible, but an exact surface search can miss related inflections. Searching only respondi will not automatically answer how respondas, respondis, or respondu behave unless the query is designed to include them. Compare forms deliberately and document what was included.
Mi serĉis aparte la formojn «respondas», «respondis» kaj «respondu», ĉar iliaj kuntekstoj povas diferenci.
I searched separately for the forms ‘respondas,’ ‘respondis,’ and ‘respondu’ because their contexts may differ.
For a preposition or collocation, inspect a wide enough window. Immediate adjacency may fail when an object, adverb, or parenthesis intervenes.
Inter la verbo kaj la prepozicia komplemento ofte staras kelkaj aliaj vortoj.
Several other words often stand between the verb and the prepositional complement.
Orthographic variants matter too. Older sources may contain apostrophized forms; imported names vary; legacy surrogate spellings belong to particular digital contexts. Do not merge distinct strings silently. State whether the search was case-sensitive and which variants were combined.
Mi ripetis la serĉon kun majusklo kaj minusklo, sed la rezulto apenaŭ ŝanĝiĝis.
I repeated the search with uppercase and lowercase, but the result hardly changed.
Read concordance lines in context
A hit can be part of a quotation, a rejected example, a foreign-language passage, a title, or an author's discussion of an error. The surrounding sentence is the minimum; often the paragraph and bibliographic metadata are necessary.
La aŭtoro ne uzis la formon; li citis ĝin por klarigi, kial li malrekomendas ĝin.
The author did not use the form; he quoted it to explain why he advises against it.
En la plena alineo evidentiĝis, ke la vorto estis ironia, ne laŭvorta.
In the full paragraph it became clear that the word was ironic, not literal.
Automatic counts cannot reliably classify such pragmatic functions. Manual reading of a sample is part of corpus analysis, not an optional decorative step.
Compare variants fairly
Suppose two expressions compete. Search equivalent morphological ranges, comparable spelling variants, and the same subcorpus. Comparing all forms of A with one exact inflection of B builds the answer into the method.
Por justa komparo mi serĉis ĉiujn tempajn formojn de ambaŭ verboj en la sama periodo.
For a fair comparison I searched all tense forms of both verbs in the same period.
Raw counts should be related to corpus size when subcorpora differ. One hundred hits in a huge collection may represent a lower rate than twenty hits in a small one. If exact normalization is not possible, report the limitation.
La moderna tekstaro estas multe pli granda, tial la nudaj nombroj ne estas rekte kompareblaj.
The modern text collection is much larger, so the raw numbers are not directly comparable.
Also separate senses. A word may be frequent overall but rare in the meaning under dispute.
Plej multaj trafoj havas la konkretan signifon; nur du montras la novan metaforan uzon.
Most hits have the concrete meaning; only two show the new metaphorical use.
What positive evidence supports
One genuine hit establishes attestation in one source. Multiple independent hits support broader use. Distribution across authors, decades, and genres strengthens a claim of conventionality. Repeated occurrence in dialogue supports conversational representation more than repeated occurrence in legal documents, though written dialogue is still constructed.
| Observation | Safe conclusion | Unsafe leap |
|---|---|---|
| one clear hit | attested in that source | standard everywhere |
| many authors, one period | distributed in that period | equally common today |
| many genres and periods | broad written attestation | required by grammar |
| no hits | absent from this search and selection | impossible or incorrect |
La esprimo aperas ĉe multaj sendependaj aŭtoroj, do ĝi ne estas nur persona strangaĵo de unu verkisto.
The expression occurs in many independent authors, so it is not merely one writer's personal oddity.
This still leaves register and norm to investigate.
Negative evidence is especially fragile
No result may reflect a rare form, the wrong inflection, spelling variation, gaps in the holdings, or a medium absent from the corpus. Phrase the conclusion narrowly.
Mi trovis neniun trafon en la elektitaj tekstoj inter 1950 kaj 2000.
I found no hit in the selected texts between 1950 and 2000.
That is reproducible. La formo ne ekzistas en Esperanto is not justified by the same search.
La manko de trafoj igas la formon suspektinde malofta, sed ne pruvas ĝin kontraŭgramatika.
The lack of hits makes the form suspiciously rare, but does not prove it ungrammatical.
For newly coined technical words, corpus absence may simply mean the corpus predates the domain.
Corpus evidence and normative evidence
Tekstaro can reveal usage that informs grammatical analysis and institutional decisions. It does not amend the Fundamento. Conversely, official status does not predict how often a form appears. State both dimensions when relevant.
La radiko estas oficiala, sed en ĉi tiu korpuso ĝi aperas preskaŭ nur en fakaj tekstoj.
The root is official, but in this corpus it occurs almost only in specialist texts.
La varianto estas neoficiala, tamen ĝi havas vastan modernan atestadon.
The variant is unofficial, yet it has broad modern attestation.
Neither statement alone tells a learner what to use. Add the communicative setting and a pedagogical recommendation.
Make the search reproducible
For a serious claim (academic), record:
- the query and included spelling or inflectional variants;
- the search date;
- selected texts, period, and filters;
- raw hit count and number of independent sources;
- exclusions such as quotations or foreign passages;
- how senses and constructions were classified.
Mi konservis la serĉesprimon kaj liston de fontoj, por ke alia esploranto povu ripeti la analizon.
I saved the search expression and source list so that another researcher could repeat the analysis.
The corpus continues to grow, so a later search may legitimately return a different total.
A strong corpus note ends with an interpretation proportional to the method. Separate what you directly observed from what you infer: “twelve relevant examples in eight post-1990 sources” is an observation; “the construction is established in modern edited prose” is a reasoned inference. Then state what remains untested, such as spontaneous speech or informal messaging. This discipline makes a modest result useful and prevents a large result from sounding more universal than the corpus permits.
Common Mistakes
Treating one hit as a rule
❌ Mi trovis unu ekzemplon, do la formo estas norma en ĉiuj registroj.
Incorrect — one hit proves only limited attestation.
✅ Mi trovis unu ekzemplon en romano de 1910; mi bezonas pli da fontoj por pli vasta konkludo.
I found one example in a novel from 1910; I need more sources for a broader conclusion.
Counting hits without counting sources
❌ La frazo aperas kvindekfoje, do kvindek aŭtoroj uzis ĝin.
Incorrect — one text or repeated quotation may supply many hits.
✅ La kvindek trafoj venas de tri verkoj de du aŭtoroj.
The fifty hits come from three works by two authors.
Equating no hit with impossibility
❌ La serĉo donis nul rezultojn, do la vorto ne povas ekzisti.
Incorrect — the query or corpus coverage may miss it.
✅ Mi ne trovis la vorton en la elektita korpuso; mi kontrolos variaĵojn kaj aliajn fontojn.
I did not find the word in the selected corpus; I will check variants and other sources.
Ignoring period and genre
❌ La formo estas ofta en malnovaj poemoj, do ĝi estas neŭtrala en moderna teknika raporto.
Incorrect — period and genre strongly affect register.
✅ En malnovaj poemoj la formo estas ofta; en moderna prozo ĝi ŝajnas literatura.
In old poems the form is frequent; in modern prose it appears literary.
Key Takeaways
Tekstaro provides attestation, contexts, and distribution within a documented collection. Design comparable queries, inspect sources, separate senses, count independent texts, and report period and genre. Combine corpus results with the Fundamento, formal Akademio guidance, and grammatical analysis; no single hit count can replace those distinct kinds of evidence.
Related Topics
- The Fundamento and the Living LanguageC1 — Understand how Esperanto combines an immutable foundational standard with productive vocabulary growth, changing usage, and accepted modern variation.
- Akademio RecommendationsC1 — Interpret Akademio decisions, Official Additions, recommendations, dictionary status labels, and consultation answers without overstating their force.
- PMEG and Descriptive GrammarC1 — Use PMEG as a detailed evidence-based grammar while distinguishing its analysis and recommendations from the Fundamento and formal Akademio decisions.
- Accepted Variation and Personal StyleC2 — Distinguish stable grammar, accepted alternatives, register choices, individual style, and genuinely nonstandard forms in Esperanto.
- Historical Esperanto and Modern UsageC1 — Read older Esperanto accurately while distinguishing unchanged grammar, dated style, accepted modern alternatives, and later innovation.
- Age, Medium, and Community VariationC2 — Analyze how generation, communication medium, and social networks shape Esperanto without turning tendencies into rules.