A straightforward baseline for this subtask involves learning a predictor function for helpfulness scores. Features used can range from product rating to reviewer reputation, count of POS tags and even the length of the review. A sample solution can be found in the RevOpiD-2017 shared task description paper and a participating system (CYUT).
The following two baselines can be referred when developing a participating system for Subtask B of RevOpiD-2018:
The above mentioned baselines for Subtask B can be extended easily to rank reviews of any given product.