Silicon Updates: The AI Poet – “Deep-speare” – mastered the dynamics of Shakespeare’s work

Artificial Intelligence (AI) has been the most evolving technology from few years in the past. The tech world has recently seen a very astonishing application of AI – Literature. After the advent of AI into the creative branches like music, paintings, sculpture and choreography. 

Image source : Google

 

Following are the stanzas from the sonnet written by William Shakespeare and Deep-speare.

Sonnet written by William Shakespeare

Sonnet written by Deep-speare

The piece of work created by Deep-speare is completely non-sense if you closely observe. But at a first glance, you cannot differentiate between the stanza created by Shakespeare and Deep-speare.

The research team consisted of three machine learning researchers and one scholar of literature. The team trained Deep-spear with 2700 sonnets taken from Project Gutenberg. Apart from these sonnets, no other input was given to Deep-speare. Out of these inputs, the “poet” cooked up three rules on itself – rhyming, rhyming scheme and the fundamentals of human language.

The models generated by AI systems are exceptional at discovering the patterns. Some of the patterns recognized by AI systems are not even intended. Such unexpectedly recognized patterns are referred to as “Accidental creativity.”

A sonnet is a kind of poetry that comprises of fourteen lines. Sonnets are used to as a tool to address a problem and a respective solution. The problem the poet wants to express is presented in three stanzas called Quatrains and the solution/resolution is presented in a couplet, following the quatrain. English poets developed a sonnet style with 10-syllable lines in an unstressed-stressed rhythmic pattern called “Iambic pentameter.”  Such frequent sonnet usage by Shakespeare was named as Shakespearean sonnet.

Shakespearean sonnet
Image source : Wikihow

With the Deep-speare project, the scholars produced the individual quatrains from various Shakespearean sonnets and the verses in iambic pentameter with regular schemes.

The Poetic process:

Following are the various steps involved in the poetic process followed by Deep-speare.

Step 1:

Deep-speare first chooses the last word of a line and assesses each word’s frequency to appear at the end of a line, across the inputs provided. A list of top five words is generated, out of which, samples are made and positioned as the word in the new line.

Step 2:

The process is repeated for every word of the line. By the end of the process, the frequency of the words appearing next to each other is generated.

Steps 1 and 2

Step 3:

This procedure is repeated to generate as many candidate lines as possible. In comparison with the rhythm model, rhythm score is calculated and assigned to the lines generated. Out of these lines, the Deep-speare samples a line that fits good into the iambic pentameter sonnet model.

Step 3

Step 4:

Deep-speare repeats these steps in a bottom-to-top approach to all the lines under consideration.

Step 5:

In order to conclude a poem, a score called “rhymability” score is decided. Sampling out of this rhymability score, the Deep-speare finishes the poem.

Steps 4 and 5