Book Success Prediction with Pretrained Sentence Embeddings and Readability Scores
This work addresses the problem of predicting book success for publishers and readers, but it is incremental as it builds on existing embedding and readability techniques.
The paper tackles book success prediction by using pretrained sentence embeddings and readability scores, achieving a 6.4% F1-score improvement over baselines and showing that only the first 1,000 sentences are sufficient for accurate predictions.
Predicting the potential success of a book in advance is vital in many applications. This could help both publishers and readers in their decision-making process whether or not a book is worth publishing and reading, respectively. In this paper, we propose a model that leverages pretrained sentence embeddings along with various readability scores for book success prediction. Unlike previous methods, the proposed method requires no count-based, lexical, or syntactic features. Instead, we use a convolutional neural network over pretrained sentence embeddings and leverage different readability scores through a simple concatenation operation. Our proposed model outperforms strong baselines for this task by as large as 6.4\% F1-score points. Moreover, our experiments show that according to our model, only the first 1K sentences are good enough to predict the potential success of books.