A History of Poem Generation
In the recent years, several attempts have been made to generate creative compositions automatically. Composing musical pieces, creating visual arts, generating poetic texts are only some of the ingenious objects that computers can generate through machine learning. With the help of programs and computers, you can now create poems even if you don’t know all the rules for poem composition.
The Basis of Automatic Poem Generation
Natural language generation is a promising subfield of computational linguistics and artificial intelligence. Its primary objective is to develop software and computer programs that can generate human-readable texts. There are different types of compositions that can be automatically generated like jokes, story narratives, weather forecasts, songs, and poems.
Attempts to Poem Generation
Automatic poem generation was a hot topic in the 90s. During these years, this concept was first theorized. It was avidly discussed and debated in scientific meetings and then published in magazines and literature.
Since the late 1990s, many attempts have been made in generating poems automatically, and they include the following:
The ALAMO group
On the ALAMO group, various French poem generation programs can be utilized. Their Gervais and Rimbaudelaires Rubaud J and P are two of the French poem generators that exist on their website. They mainly use existing Rimbaud sonnets as a template and then their program replaces the adjectives, verbs, and nouns by words that appear in Baudelaire’s poems.
The Manurung’s chart system
In 1999, Hisar Maruli Manurung, a doctor of philosophy, introduced a chart system that generates syntactically correct texts, which follow a rhythmic pattern. By reversing a chart parser, utilized for parsing, Marunung came up with a chart generator that translates abstract forms to strings. The strings are then used to produce natural language strings that match certain stress patterns.
The first-order logic, which represents sentences, illustrated the inputted semantics. While generating the final output, the result’s stress pattern, which is based on a new rule, is checked, and analyzed against the pre-defined stress pattern before adding the rule to the Manurung’s chart.
The stress patterns are described as a list of strong and weak syllables that are obtained from a certain dictionary. The Manurung’s chart system maintains a rhythmic, syntactic, and semantic quality at every step while sacrificing flexibility and frugality. As a result, the system sometimes produces a zero output.
WASP
The WASP is a rule rule-based system, which means that it is based on a set of if-then statements. It is considered as the first attempt at automatic poem generation. WASP is a kind of system that provides intelligent answers to given problems. It is comprised of multiple programs. These programs implement various construction heuristics from formal metric constraints that aim to create a poem based on existing poetic text in the system or on the internet.
Tra-La-Lyrics
Tra-la-lyrics is a system that produces Portuguese lyrics. Its output is based on the rhythm of the song inputted by the user. Using the pattern of weak and strong melody beats, this system’s way of generating song lyrics is almost the same to the process of poem generation.
By following the generate and test technique, the Tra-la-la system can produce grammatical sentences and scores them based on the restrictions derived from the metric pattern. To achieve specified goals, tra-la-lyrics follows three strategies:
• Random words and rhyme
• Words following system templates and rhyme
• Grammar and rhyme
In tra-la-la lyrics, the best sentence, clause, or phrase that fits the required rhythm is chosen. While the first strategy doesn’t consider grammar rules, the second and third strategy doesn’t ensure the presence of rhyme.
Final Words
The quality, length, properties, and elements comprising the final output of poem generators are often varied. But, with generateapoem.com, you can always get the creative poem that you wish to have.
This website provides users an innovative poetry generator that utilizes machine learning to retrieve, translate, and generate a rhythmic poem that is based on the specified format of the user. It applies unsupervised learning to describe word stress syllable patterns and to analyze discourse models.
The website also allows you to choose the type of poem that will be generated. You can either choose from couplet, haiku, sonnet, and complex. By specifying some rules, you can easily have a poem that conforms to the standards that you’ve set.
