TEXT GENERATION: THE PROCESS, THE PROS, AND CONS
Text generation also called Natural language generation(NLG) is an area of research under Natural language processing (NLP) which is responsible for a number of technologies including text to speech. NLP algorithms take in some text and try to find patterns by turning these data into matrixes thus simplifying the data. By doing so, the algorithm can then pick out important phrases and words; find relationships between them and thus make predictions when it comes across them in the future. Similarly, NLG algorithms take in structured data and then generate text (paragraph(s), a sentence, etc) in natural language. Think of it this way, if 30 people were asked to write 50 word texts about elephants, for instance, text generation algorithms could take these texts as input, study patterns across the 30 different writings, and then generate its own unique write-up on elephants.
Automated text generation has quite a number of useful applications. Many organizations now use automated text to improve user experience, for reports, weather analysis, and a host of other things. An even exciting phenomenon is that texts from these algorithms are increasingly as solid as human-written ones. Thus in industries that require constant writing such as media and journalism, the time-consuming and sometimes uncomfortable task of continuous writing can be left to computers.
Unfortunately, like all other technology, text generation can be exploited negatively. Especially worrisome is the potential of these models to be used in creating fake news. In fact, one of the reasons some text generators aren’t open to the public is that researchers are worried that because texts generated from these systems are “so real”, and of high quality, it would be really difficult to make a distinction between real and fake information. Also, text generators can be abused by students who might want to cheat on essays and other school work. Prospective post-grad students for an instance could just feed a text generator with various motivation letters, statements of purpose, etc, and get a fresh write-up for use in college applications. The result would be a high-quality essay which is no true reflection of that student’s abilities.
In conclusion, text generation is an amazing technology and could potentially be a game-changer if applied correctly; think write-ups on art pieces, news summary, book reviews, and more user centric information.
