Soylent is a word processing tool used for proofreading. It goes through many contributors sourced through Amazon Mechanical Turk, and produces crowdsourced edited documents for writers. Soylent consists of three main tasks: shorten, proofread, and format. This paper believes that crowdsourcing the task of proofreading allows writers to gather a holistic and advanced final draft that has been approved by a mass of people. In the past, works have used crowdsourcing mainly for past artifacts, while Soylent focuses on present tasks. In addition, human power may be more suited for such literary problems than artificial intelligence.
Shortening is a much needed tool, not only for the authors, but also the future readers. As humanity evolve and grow, the mass of literature only grows larger. Parsimoniousness is a key factor to consider in research. However, I wonder if it would be necessary for the author to provide a word limit range of sorts, and stopping the Shortn program when the limit has been met. I see Soylent being potentially useful for not only those in higher education, but all education levels. Thus, this feature may prove useful, as students in high school and below typically struggle to meet word limits. Though a minor and silly criticism, I feel that the system should allow author set constraints, for many benefits even beyond this specific one.
The feature Crowdproof is very interesting to me. Without researched rationale, I already feel that distributing the task of proofreading will improve accuracy. Certainly, it is easier to point out grammatical or spelling errors from a sentence than a huge paragraph. I feel that a study could be done to see what length of text is best for this distribution. How would the system know where to begin or end an excerpt? If the segment is too short, there may not be enough context to determine whether a sentence is correct. For example, what if a sentence lacks structure, but it was because it was a direct quote? Or poetic metaphor? However, if an excerpt it too long, then the tool loses its purpose.
Regarding how the participants are recruited to do exactly what task, I am still somewhat confused. If the crowd-sourced authors are randomly assigned sections, I feel this system could work. However, certain features still seem like they are simply hiring multiple proofreaders to achieve greater results than using fewer editors. For example, if a certain quota of proofreaders were assigned to each section or segment, the issue of lazy proofreaders should be eased.