Gizmodo has an interesting story about writing. Researchers from Stony Brook University in New York have created an algorithm that attempts to predict how well a novel will sell.
http://aclweb.org/anthology/D/D13/D13-1181.pdf
They have tested their algorithm against the text of many books over several genre, and compared their results against the books' sales.
The basic results show that high numbers of prepositions, nouns, pronouns, determiners and adjectives are all indicative of highly successful novels, whilst higher percentages of verbs, adverbs, and foreign words characterise less successful novels (ref. Table 6).
I would be intrigued to put some of my writing through this, and I can imagine that the researchers would be able to make a little money from desperate literary novelists (or publishers and agents wanting to reduce the size of their slush pile).
I generally tend to be sceptical about this sort of research, but these results seem to make sense, at least to a certain extent. However one of the most successful recent novels - Dan Brown's (yes, he) "The Lost Symbol", scored poorly. Which shows that plotting and oodles of publicity can overcome the predictive power of algorithms.
8K display on laptop
1 month ago
No comments:
Post a Comment