NTUA-ISLab at SemEval-2019 Task 9: Mining Suggestions in the wild

Proceedings of the 13th International Workshop on Semantic Evaluation (SemEval-2019), pages 1224–1230 Minneapolis, Minnesota, USA, June 6–7, 2019. ©2019 Association for Computational Linguistics 1224 NTUA-ISLab at SemEval-2019 Task 9: Mining Suggestions in the wild Rolandos Alexandros Potamias, Alexandros Neofytou, Georgios Siolas Intelligent Systems Laboratory, National Technical University of Athens, Zografou, Athens , Greece rolpotamias@gmail.com, alex.neofytou@gmail.com gsiolas@islab.ntua.gr Abstract As online customer forums and product com- parison sites increase their societal influence, users are actively expressing their opinions and posting their recommendations on their fellow customers online. However, systems capable of recognizing suggestions still lack in stability. Suggestion Mining, a novel and chal- lenging field of Natural Language Process- ing, is increasingly gaining attention, aiming to track user advice on online forums. In this paper, a carefully designed methodology to identify customer-to-company and customer- to-customer suggestions is presented. The methodology implements a rule-based classi- fier using heuristic, lexical and syntactic pat- terns. The approach ranked at 5th and 1st position, achieving an f1-score of 0.749 and 0.858 for SemEval-2019/Suggestion Mining sub-tasks A and B, respectively. In addition, we were able to improve performance results by combining the rule-based classifier with a recurrent convolutional neural network, that exhibits an f1-score of 0.79 for subtask A. 1 Introduction Nowadays, online platforms and e-commerce sites have become increasingly popular. In such online environments people are willingly sharing views, feelings and opinions about specific products and services by conveying written recommendations on several topics. The pool of advising items that float in contemporary e-commerce environ- ments posts a critical task: how to identify, capture and extract useful information from them. Dig- ging into these information pools of unstructured commenting dialogues about ideas and thoughts expressed by reviewers on the product, its fea- tures, components or functions can be seen as an Opinion Mining (OM) problem (Banitaan et al., 2010), with the relevant research gaining increas- ingly significant attention. However, OM strategies attempt to extract the sentiment of users written passages, failing to cor- rectly capture putative advice, especially in the presence of mixed emotions. For example, a sug- gestion in an online trip advising platform could contain the sentence “If you sleep lightly, this would be a problem because of all the trams that go by - so ask for an interior room.”. The pas- sage conveys a negative sentiment which is im- possible to grasp correctly. Confronted with this problem, computational methodologies that auto- matically identify and highlight clear-cut advice in written recommendations raises as a major need. Towards this end, Suggestion Mining (SM) strives to track suggestions and tips in passages. Sug- gestive sentences could be positive, negative or they may not contain any sentiment at all, a fact that makes their tracking quite puzzling (Negi and Buitelaar, 2015). In general, advice may be ex- pressed by using either explicit suggestions, which propose direct and conventionalized forms of rec- ommendations (Mart´ınez Flor, 2005), or implicit suggestions where the precise recommendation is in a way veiled and should be inferred (Negi and Buitelaar, 2015). The …

Content truncated to 500 words