Constrained Natural Language Generation (NLG) Task
Jump to navigation
Jump to search
A Constrained Natural Language Generation (NLG) Task is an NLG task that accepts an linguistic constraint.
- Context:
- It can support NLG Applications, such as: Machine Translation, Dialogue Systems, and Text Summarization.
- …
- Example(s):
- General Constraints NLG Task, such as:
- Generate Text(includeWords='cat, dog', subject='animals') => “The cat and the dog are common pets."
- Generate Text(accuracy='high', subject='history') => “The Treaty of Versailles was signed in 1919."
- Generate Text(audience='children', purpose='entertain') => “Once upon a time in a magical kingdom..."
- Generate Text(deadline='2023-12-31') => “Submitted before the end of the year."
- Structural Constraints NLG Task, such as:
- Generate Article(length={200}, paragraphCount={3}) => “Introduction. Body. Conclusion."
- Generate Report(structure='intro, body, conclusion', format='pdf') => “PDF file with clearly defined sections."
- Stylistic Constraints NLG Task, such as:
- Generate Text(vocabulary='simple') => “The cat sits on the mat."
- Generate Text(grammar='advanced') => “Although she was busy, she made time for her family."
- Generate Text(stylisticDevices='metaphor') => “Time is a thief."
- Generate Text(tone='humorous') => “Why did the chicken cross the playground? To get to the other slide!"
- Special Constraints NLG Task, such as:
- Generate a Poem(lineCount={2}, syllableCount={7, 5}, wordOrder='alphabetical') => “Children in the laughing park, / Kites fly high."
- Generate Sentence(sentiments='positive, polite, non-toxic') => “Thank you for your kind consideration; you're doing great."
- Generate a Sentence(syntax='imperative', adverbPlacement='before verb', structure='[verb] [Name2] the [object]') => “Eagerly give Mary the book."
- Practical Examples, such as:
- Generate Product Description(length={100}, features='include', style='clear, concise') => “This product has A, B, C features and offers numerous benefits."
- Generate News Article(length={200}, mainPoints='summarize', tone='neutral') => “Event X occurred, involving Y and Z."
- Generate Legal Contract(clauses='include', style='clear, unambiguous') => “All necessary clauses are present."
- Generate Technical Manual(language='simple', illustrations='clear') => “Step 1: Do this. Step 2: Do that."
- Generate Language Exercise(grammarStructure='particular') => “Fill in the blanks with the correct form of the verb."
- …
- General Constraints NLG Task, such as:
- Counter-Example(s):
- See: ....