Paradigms

← Guide index · 中文 · seventh item in the navigation bar

A paradigm combines grammatical dimensions into slots and gives each slot a generator, so every inflected form can be derived from the stem automatically. Inflection, agglutination, templatic alternation and reduplication can all be expressed; a language without inflection can ignore this page entirely.

1. Structure {#1-structure}

• Dimensions are the categories a word changes for — number, case, tense — and are defined on the lexicon’s Parts of speech & dimensions sub-page.
• A paradigm picks some dimensions and combines their values into slots: sg.nom, sg.acc, pl.nom… A slot label is its value names joined with ., and is also the key in an entry’s forms. The order of the dimension chips decides which comes first in slot names (and which forms the rows or columns of the table); move a chip with its arrows, or press and drag it onto another chip. When you reorder them, each slot’s generator, variants and disabled state follow along, and inflected forms stored under slot names in the lexicon move to the new names.
• The number of slots is the product of the number of values in each dimension, so it grows quickly: when it would exceed 500, a dialog warns you before the dimension is added (you can still add it), and the slot summary turns to a warning colour. With too many slots, the Paradigms page, derivation and entry editing all slow down.
• Slots you don’t need can be disabled.
• A paradigm can inherit from another: the child only defines the slots that differ and takes the rest from the parent.
• Once a part of speech is bound to a paradigm, its entries get a slot panel in the lexicon’s Edit mode; a single entry can also name a different paradigm.
• A part of speech can be bound to several paradigms (say, a verb’s first and second conjugations): tick them under Bound parts of speech in the inspector. The first one bound is the default, used by entries that haven’t chosen a paradigm; the other paradigms’ rows offer Make default, and unticked parts of speech show in small print which paradigms they are already bound to. An entry chooses its paradigm in the Edit mode drop-down (where “paradigms of this part of speech” form their own group); the test bench first lists the words currently using this paradigm, then those whose part of speech is bound to it but haven’t switched to it.

2. Main area: the slot table

To the right of the Slots title you can switch between three views:

• Visual (default): one row per slot — enabled checkbox, slot name, gloss abbreviation, generator and pipeline — edited right in the row.
• Table: the first dimension forms the rows and the second the columns; from a third dimension on, each of its values (each combination, with more dimensions) gets its own table, labelled above it. With a single dimension it is one column.
• Tree: branches level by level in dimension order, with slots at the last level; click a branch point to collapse or expand it (with more than 200 slots, branches start collapsed and only the ones you open are drawn).

Each cell of the table and tree shows what the test bench word becomes (change the word in the inspector’s test bench; with no word there, the cell only shows how it is built), with small print explaining how the cell is built — the affix of each step, which rule set runs, the first line of an adjustment, and “⟨stem name⟩” when the stem isn’t the headword. A form that doesn’t match the one typed in the entry is shown in red, disabled cells say Off, and cells filtered out by the top-bar search are faded. Click any cell to return to the visual view, which scrolls to that cell and flashes it.

3. Variants

When a slot has two parallel forms (colloquial / literary, form A / form B), you don’t need a whole new paradigm:

1. Create a variant in the Variants bar above the slot table and name it.
2. With that variant selected, edit slots — only the slots you change belong to the variant; the rest follow the base set.
3. An entry can choose which variant to use; switching recomputes the derived forms and leaves hand-typed ones alone.
4. Hovering a variant button shows a pencil on its right: click it to rename in place. With no variant selected you rename the base set (called Base by default — you might rename it “written”); entries’ variant drop-downs show the new name, and clearing the name brings back Base.

4. The pipeline: start from the stem and add step by step

Each slot’s generator is a pipeline: it starts with just the stem, and you add whatever steps you need; each step acts on the result of the previous one. Change the order with the left / right arrows on a step, or press a step (not on an input) and drag it onto another step. The Add step menu closes a moment after the mouse leaves it, or when you press Esc.

stem[strong] → prefix @DEF → suffix -s → sound change (Proto-Shikrin → Theusrin) → tweak -at

Available steps:

The same kind of step can appear several times: add a prefix, run sound changes, add a suffix, then tweak — any order you like. Only three generator kinds remain: None (not derived), Table (manual) (typed per word) and pipeline (this pipeline).

The four older generator kinds — affixation, affixation + sound changes, pattern and reduplication — are converted to equivalent pipelines automatically when an old project is opened; derived results don’t change.

5. Notation quick reference

5.1 Prefixes and suffixes

Letters that change by condition: in prefixes, suffixes, infixes, circumfixes and tweaks, write {condition:form|condition:form|default}; during derivation one branch is chosen for this entry and this cell, and everything else stays the same. A condition is a dimension value’s name or abbreviation; both the entry’s own grammatical features (gender, noun class and other things that aren’t paradigm dimensions) and this cell’s dimension values count, with the cell taking precedence for the same dimension:

Steps that use such affixes list every possible choice in small print beside them, and the derivation trace says which branch this word took; when the corpus segments words, every choice is tried as an affix to strip.

5.2 Infix positions

The second box of an infix step is the insertion position; empty means V1:

Segments are split using the language’s phoneme inventory and digraphs, so th, ng and the like count as one segment.

5.3 Tweaks

A tweak step holds one operation per line, applied in order:

A tweak takes effect wherever it sits in the pipeline — to tidy up a word ending before sound changes run, put the tweak before the sound-change step.

6. Stem slots

The stem at the start of a pipeline comes from the entry’s stem slots. Stem slots are defined per part of speech in Lexicon → Parts of speech & dimensions (a name + a description, such as a noun’s strong / middle / weak forms) and filled in per entry in Edit mode; the stem drop-down lists the stem slots of the bound parts of speech, and the hint next to it explains where stems come from. lemma, an empty value, or a slot the word hasn’t filled all mean the headword.

Paradigms that apply to all words

A paradigm with Applies to all words ticked in the inspector (initial mutations, sandhi and the like) doesn’t need a part of speech and doesn’t write inflected forms to entries; its reconciliation only runs the project consistency check. It is used by the corpus: while segmenting, the software feeds a batch of lexicon words through the paradigm, compares before and after, and builds tables such as “initial X becomes Y” and “final X becomes Y”; when a word in a sentence doesn’t match anything, its start or end is changed back and it is looked up again, with the slot’s abbreviation added after the gloss. For example, if the “soft mutation (LEN)” slot of an “initial mutation” dimension has one tweak m > w / #_, then na-wener in the corpus is recognised as mener “sky.LEN”. The test bench lists every word of the current language. Once ticked, you can also choose Language it applies to, so one language’s mutations aren’t used to analyse another language.

7. Inspector: the test bench

• The search box matches the whole lexicon loosely (headword or definition); words bound to this paradigm come first. Click one to try it; the test bench runs whichever variant you are editing.
• See the derived result and the step-by-step trace for every slot.
• Derive this word and store: writes the results into this word’s inflected forms (cells overridden by hand are left alone); the table flashes green afterwards.
• Derive all: writes derived forms for every entry bound to this paradigm.

8. Reconciliation report

The Reconciliation report button next to the title compares derived results with the hand-entered inflected forms in entries (imported from CSV or typed in Edit mode), giving a match rate and a list of differences per slot. With many entries a progress bar is shown, as it is for bulk derivation. Use it to check whether your rules cover your data: a slot with a low match rate usually means a sound change or tweak is missing.

Project consistency

Below it, the same report includes a check-up of the whole project (it runs even if no entry is bound to the paradigm), grouped by severity; click an item to jump to it. It checks entries, morphemes and sentences of the current language, or of all languages when All languages is selected; parts of speech, dimensions and paradigms are project-wide anyway:

9. Connections to other modules

• Morphemes: @morpheme references, with allomorphs picked by environment.
• Sound changes: the sound-change step refers to a rule set and its stages.
• Lexicon: bound parts of speech, per-entry paradigm and variant, the slot panel, inflected forms.
• Corpus: automatic glossing looks words up through their inflected forms; affixes written in paradigms are also used to work out heavily inflected words, so hover cards can find the matching entry.

10. Examples

The noun paradigm of Aelith (an example project): dimensions number × case, and every slot is two steps, “suffix → sound change”. The suffixes are written with archiphonemes, ¢lAr and ¢dA; in the rule set, A is realised as a / e depending on whether the preceding vowel is back or front, Ŭ is a linking vowel dropped after stems ending in a vowel, and ¢ is deleted at the end.

An agglutinative verb head can be handed over to paradigms entirely too: pick the dimensions mood-particle × mood × aspect × tense, and give each slot a pipeline — tweak -· removes the middle dot at the end of the head, then suffix @some-particle, suffix @some-mood… in turn (@ refers to indivisible morphemes in the morpheme list, with allomorphs picked by environment); where another slot follows, a tweak removes the previous slot’s marker vowel, and finally tweak +· puts the middle dot back. The morpheme list then only holds atomic morphemes, every compound head is derived by the paradigm, and hovering in the corpus breaks it down piece by piece.