Markov Chains Text Generation for Ideation Platforms

Humankind has never experienced the geometric growth and interest in computing particularly in Artificial Intelligence (AI)
like in recent times. Efficient processes, automated systems and improved decision making are some of the milestones of
this trend. The increased demand for text related solutions from generation, learning, classification and several other tasks has  motivated the use of different techniques and tools of AI. Creative text ideas have been sought after for innovation, problem solving and improvements. These ideas are needed in all endeavours of life as coming up with them can be a daunting task for individuals, organizations and ideation platforms. In this work, an idea generation system based on  improvements of Markov chains approach using a corpus of text is presented. First, a web system was created to collect solutions from people on a case study problem. They were required to make submissions based on purpose and mechanism with examples to guide them. Next, the collected text was clustered based on similarity measure into groups, then abstractive summaries of the respective groups were computed. Markov chains model was then used to generate new text from the submitted text corpus and the most similar Markov chains generated text was compared with each clustered group’s abstractive summary using a similarity measure and returned as an idea result. Finally, a pipeline to execute all the components of the system at once was developed. The result was sent for human evaluation based on
the metrics of quality, novelty and variety and compared with output from a generative transformer system using the same
text corpus and this work’s system performed better. The work majorly addresses a challenge faced by ideation  platforms.