Completely homographic in all cases with different stress

Completely identical in all cases, except one form has different stress

One form is the nominative plural of two different nouns

Homographic inflections of a single noun

Inflections of a single noun that are homographic when ё is not used (let’s call them “ё-mographic” inflections)

Ё-mographic nouns that are otherwise identical

Polysemous nouns with different plural forms for different meanings

Homographic nouns distinguished by only one case

Homographic nouns of different genders

Nouns whose stress is affected by the presence of a preposition

Homographic adjectives, distinguished by stress

Homographic adjectives, distinguished by comparative

Homographic adjectives, distinguished by short form spelling

One form is the nominative plural (and possibly genitive feminine singular) of two different adjectives

Homographs across uninflected parts of speech

Homographs across inflected parts of speech

Ё-mographs across different parts of speech

Homographic verbs in the infinitive, distinguished by stress

Homographic verbs in the infinitive, distinguished by nonpast forms

Homophonous verbs in the infinitive, distinguished by nonpast stress class

Verbs whose aspect partners are homographs of themselves in the infinitive

Verbs that are homophones but of different aspects, with different aspect partners (or none)

Homographic verbs in the infinitive (possibly homophonous, and also possibly homographic in nonpast tense), distinguished by aspect partner

Given verb pairs A1 A2 and B1 B2, where A1 and A2 are identical in the infinitive and A2 and B2 are identical in the infintive, but B1 and B2 have different stress either in nonpast or past tense

Nonpast forms that could be an inflection of one verb or of its aspect partner if stress is not marked or ё is not used

Homographic inflected forms of unrelated verbs

A verb has a past passive participle P that is homographic (or ё-mographic) with adjective A (whose meaning is related but not identical), and the adjective A follows the pattern -ен(ный), -енна while participle P follows the pattern -ён(ный), -ена́, or vice-versa (let’s call this the yenny-yonny problem)

Similar to the above, but the stress changes instead of the form

Large language models (LLMs) like ChatGPT excel at generating coherent text, but they’re not actually translation tools. Unlike DeepL or Yandex Translate, they aren’t trained on parallel texts, so they don’t learn true word-to-word correspondences between languages. Instead, they rewrite your input in a new style or language, which can make text sound more fluent, but may drop or add details.

Slovarish is a dictionary and is designed to do what dictionaries do best, but better. Handling full sentences is beyond the scope of any dictionary, and for extracting the meaning of a text, a dedicated machine translation (MT) tool is better suited to giving at least a rough idea of it.

However, especially with a highly inflected language like Russian, merely knowing the translation of a sentence doesn’t necessarily provide you with the information to make that sentence your own so that you can use those words in new contexts.

For example, MT tools don’t show things like inflections, multiple meanings, aspect partners, verb families, and same-root words (однокоренные слова). Slovarish provides all of this in a thoughtfully-designed layout that will help you remember tricky patterns and expand your vocabulary.

LLMs might be able to give you some of the same information, but it would require multiple rounds of wordy prompts at best. More fundamentally, LLMs are statistical models, not knowledge bases. They might give different answers to the same question on different days (even for simple questions about stress) and they struggle with analyzing isolated words—often hallucinating when uncertain.

Slovarish offers efficiency and reliability. As mentioned in the data sources section, the data in Slovarish originally comes from authoritative, human sources and has been thoroughly cross-checked. While it does not translate sentences, it serves as a one-stop shop for comprehensive information about Russian words that is relevant to learners of the language.

Slovarish combines data from two sources: English Wiktionary and the Smirnitsky dictionary. Unlike many dictionary apps that put each source in its own tab, here definitions are merged into a single unified view so that you don’t have to click through just to get more definitions.

However, these sources often overlap—sometimes exactly. Even within one source, aspectually paired verbs often have identical example sentences, just with a different verb. To reduce clutter, Slovarish merges definitions and examples that are very similar.

This process isn’t always perfect: it’s done dynamically, not by hand, so you may sometimes see definitions that appear duplicated or redundant. If something looks especially noisy, please report it to the Main Directorate.

One of the primary sources for the Slovarish dictionary data was Russian Wiktionary, whose coverage is far more extensive than bilingual Russian–English dictionaries.

As a fallback for words that do not exist in the Smirnitsky dictionary or did not exist in English Wiktionary at the time the Slovarish dictionary data was compiled, the Russian-language definitions are shown because that is the only available data. This was preferable to simply excluding those words and also to using machine translation.

That said, since Russian Wiktionary is a monolingual dictionary (a dictionary for native speakers), these definitions are more like conventional dictionary definitions and would not be the most efficient for learners even if they were translated. In other words, вода́ would be described as “a transparent, colorless liquid” (in Russian) rather than simply “water.”

If you would like to contribute any definitions in the form of translations, you can click the “Improve this entry” link within the relevant entry.

Wiktionary, which was one of the primary data sources for Slovarish, does contain audio files for Russian words. However, including audio files for dozens of thousands of headwords would incur significant bandwidth and storage costs, and even then, the result would not be satisfactory.

If a word is going to have an audio file, then that audio file should contain the pronunication not only of the word’s dictionary form, but also of all its inflections. This is especially true for verbs: verbs are traditionally listed by their infinitive, but hearing the infinitive of a verb is usually not helpful for learning the pronunication of its conjugated forms.

Given that most nouns have nine or ten forms, most adjectives have a few more, and verbs have a couple dozen, this is not practically feasible. There is an app called Russian Verb Trainer that contains audio for the conjugated forms of verbs, but the coverage is only in the hundreds (and there is no way to hear a single form of a verb without listening to the whole audio file from the beginning). And even on Forvo, whose purpose is to be a pronounciation dictionary, coverage of non-dictionary forms of words is far from guaranteed.

Thankfully, Russian pronunciation is highly predictable, with few exceptions, as long as you know where the stress falls in a word. So all inflections and examples are marked with stress (and sometimes even secondary stress, in the case of compound words), and exceptional pronunciations are highlighted in a callout box at the top of relevant entries.

That said, while Russian pronunication may be predictable, it is not easy for a non-native speaker to acquire a clear and natural accent. Teaching pronunciation is beyond the scope of this dictionary, and is best left to tutors who can give you live feedback.

There’s no denying that Russian words are often very long and can be a nuisance to type. However, Russian has so many inflected forms and ambiguities across those forms (e.g. замок castle, lock, or froze up) that guessing what you meant isn’t always helpful, and neither is forcing you to choose among identical-looking words.

Instead, Slovarish shows you all the relevant entries for exactly what you typed and highlights those inflections in the entries themselves. If you want to see a substantial list of example test cases that illustrate why this is a better solution than autocomplete, see the demo page.

The most feasible offline dictionary would be a simple word-to-definition mapping, which is perhaps what most people think of when they think of a dictionary. For bilingual dictionaries, this works well when the morphology of both languages is very simple, such as English, Chinese, and Swedish. However, in the case of Slavic languages such as Russian and languages with highly complex morphology in general (such as Baltic languages and Latin), it is worth reimagining what a dictionary should be.

Slovarish is a response to real stumbling blocks experienced by a learner of Russian (the developer). Among many other things, a learner-oriented Russian dictionary should:

• make it possible look up words by inflected forms
• display inflections in a way that is not overwhelming and highlight irregularities
• highlight information related to case government
• show paired verbs together and connect “families” of verbs
• allow reverse lookups based on English definitions and examples
• make it possible to look up words by shared roots, accounting for alternations

Most of these features require a large database and a sophisticated database architecture. Slovarish is designed as a Progressive Web App, so it can be used on any device and “installed” in a way that feels like a native app. However, since PWAs work through the browser, an offline version would be limited by browser database capabilities, which aren’t sophisticated enough to deliver the above features. (And downloading a database of this size on every first visit would create a poor user experience.)

An alternative solution might be to build a mobile app with a bundled database, but this would mean developing separate versions for iOS and Android, and dealing with app store restrictions and fees. And similarly to in-browser databases, there’s no guarantee that the database solutions available would be sophisticated enough. Small corrections and adjustments to the dictionary data would also require updating the app, which would be annoying for users.

Given that Slovarish is a solo project focused on creating a high-quality linguistic resource rather than the product of a mobile development company, it is more important that the dictionary be as widely accessible as possible across platforms. The web-based approach ensures Slovarish works seamlessly across all devices and platforms through any modern browser.

There are similar precedents in at least a couple of languages: Lexin indexes Swedish verbs by the present-tense form, and in Latin, the equivalent of the я-form is conventionally treated as the dictionary form.

It’s actually quite surprising that more dictionaries (in any language) don’t do this. It would make sense if the so-called “dictionary form” of a word was a form that you could (1) immediately use in a sentence and (2) extrapolate other forms from.

For example, both маши́на and бы́стрый can be immediately used in a sentence, and once you know a little bit of grammar, it’s not hard to change them into the correct form.

Not so with Russian verbs. Dictionaries conventionally treat the infinitive as the “dictionary form.” There are two problems with this.

First, you cannot immediately use the infinitive in a sentence by itself to describe an event or action. But more critically, you cannot extrapolate from the infinitive to other forms.

For example, a verb like делать is often taught to beginners as a model for Conjugation I verbs. Meanwhile, сказать, спать, and танцевать have the same ending but conjugate completely differently. This makes Russian verbs seem completely irregular and unpredictable, when only a handful of verbs are truly irregular.

There are several benefits to referring to verbs by the они-form:

• Combined with stress, this significantly reduces ambiguity. For example, the verb болеть could mean hurt, ache (боля́т) or be sick (боле́ют).
• It is easy to work backwards from the они-form to any other nonpast ending, and if you consider the nonpast stem and past stem of a verb to be generally unrelated, then virtually no verb’s conjugation will come as a surprise.

As a side note, there are two exceptions to this in the dictionary. Verbs that can only be used impersonally are indexed on the он-form (e.g., рвёт, distinct from рвут), and verbs that do not have a nonpast tense are indexed on the masculine past tense (e.g., слыха́л).

Slovarish contains most common Russian given (i.e., first) names and patronymics, but explicitly excludes surnames (i.e., last names). This is because, in terms of inflection, surnames in Russian occupy a position somewhere in between nouns and adjectives. While given names and patronymics are already gendered and decline like regular nouns, surnames must agree with the gender of the first name and can take masculine, feminine, or plural forms.

The declension of surnames or full names is complicated enough that there are even tools for it for native speakers.

Adding surname declension would require significant additional complexity in both the database structure and user interface. Since surnames don’t carry semantic meaning like other dictionary entries, this complexity isn’t justified for a learner-focused tool.

While surname declension can be useful for formal writing, learners are better served by specialized tools such as the one linked above for this specific need.