Chart accessible name. Tokens: {name} {series} {points} {help}.
Event-marker key instructions, appended to the accessible name only when the chart has annotations. No tokens.
Accessible label for the opt-in CSV download control. No tokens.
Keyboard operating instructions woven into the chart's accessible name. No tokens.
Accessible name for the legend button group. No tokens.
OHLC value words for candle series (announcements, table headers). No tokens.
Accessible labels for the single-pointer pan buttons. No tokens.
Word appended to an event-marker announcement when it becomes pinned/selected. No tokens.
Per-series visibility state words (visible-only; aria-pressed carries state for AT).
One-line summary, all series hidden. Tokens: {label} {points}.
Event-annotations clause appended to the summary. Tokens: {count} {labels}.
One-line summary sentence. Tokens: {label} {points} {span} {parts}.
One-line summary, empty dataset. Tokens: {label}.
Per-series clause within the summary. Tokens: {name} {min} {max} {last} {dir}.
x-range phrase within the summary. Tokens: {start} {end}.
Data-table caption. Tokens: {caption} {series} {rows}.
Direction phrases. {pct} for up/down; flat has no token.
Localizable UI strings — every fixed phrase the accessibility layer emits.
The library's own prose (keyboard help, legend label, data-summary sentence, table caption) is English by default. In a non-English host document those would be untagged different-language passages (WCAG 3.1.2 Language of Parts). Because each phrase interleaves author content (series names, axis labels, formatted numbers) with connective prose, the correct fix is to let the integrator supply translations that match their page — not to wrap mixed subtrees in a
langattribute. Strings are{token}templates filled byformat.