Exploring Machine Creativity through Text-Generative Algorithms
Author : Harsh Singh
Greenwood High International School
In recent times, artificial and machine intelligence has found applications through a wide variety of fields. The rising number of authors throughout the world who want to express a story or a viewpoint may not necessarily be equipped with the proper writing skills to develop an article and may need guidance in initial steps. Technology has found application here as well, and by using text-generative algorithms, they can produce rough drafts, short stories, plotlines and initial outlines customized to the writer’s needs by using specific given inputs. The aim of this research is to analyze two text-generative algorithms, different in their fundamental and logical approach to the problem, on the basis of criteria: speed, accuracy and grammar in order to understand which algorithm is more proficient at the given task. A 1000 trials were performed for both algorithms. The number of successful trials were recorded and that percentage gave the accuracy for both algorithms. Each trial’s results and the amount of time it took to generate that output was recorded. The algorithm that used the phrase method was both fast and also accounted for lesser grammatical mistakes comparatively. However, the algorithm that used the Markov’s Chain Method was more accurate in terms of output and results. The study’s results establish an elaborate manner through which the proficiency of text-generative algorithms can be analyzed. It mainly served to evaluate two widely used logical techniques used in algorithms for text-generation.