Natural language processor extracts time-related phrases from sentences, offering customizable configurations for interpreting ambiguous values within English text. Supports parsing repeating time patterns.
KWhen is a natural language processor for Kotlin Multiplatform that extracts time-related expressions from English sentences.
Add the dependency using Gradle. Replace <version> with the latest version:
repositories {
mavenCentral()
}
dependencies {
implementation("io.github.pahinaa.kwhen:kwhen:<version>")
}
You can also download the .jar directly from Maven Central.
To use the processor, create an instance of TimeParser. An optional configuration (Config) can be provided to customize interpretation. Currently, only English (ENConfig) is supported.
val tp = TimeParser()
tp.parse("I will go to the gym tmrw at 6")
You can customize the meaning of ambiguous phrases using ENConfig:
val tp2 = TimeParser(
ENConfig(
evening = 18,
morning = 9,
afternoon = 15,
night = 20,
use24 = false
)
)
tp2.parse("He goes swimming in the morning")
A call to parse() returns a list of Parsed objects. Each object contains:
val tp = TimeParser()
val parsed = tp.parse("Johnny has school from 8am to 4pm")
val firstParsed = parsed[0]
firstParsed.text // "from 8am to 4pm"
firstParsed.range // IntRange: (17..32)
firstParsed.startTime // List<LocalDateTime>: [2024-06-25T08:00]
firstParsed.endTime // LocalDateTime: 2024-06-25T16:00]
firstParsed.tagsTimeStart // Set<TimeTag>: [HOUR, MINUTE]
firstParsed.tagsTimeEnd // Set<TimeTag>: [HOUR, MINUTE]
val parsed = tp.parse("It's a holiday every Jul 4th")
val firstParsed = parsed[0]
firstParsed.repeatTag // TimeUnit.YEAR
firstParsed.repeatOften // 1
If multiple unrelated or detached time expressions are found, each is returned as a separate Parsed object in the result list.
Surfaced from shared tags and platforms — no rankings paid for.