The NeuroSent System at ESWC-2018 Challenge on Semantic Sentiment Analysis

摘要: The approach described in this paper explores the use of semantic structured representation sentences extracted from texts for multi-domain sentiment analysis purposes. presented algorithm is built upon a domain-based supervised using index-like representing information text. extracts dependency parse relationships containing training set. Then, such are aggregated together with either polarity and domain information. Such exploited order to have more fine-grained learned When new text has be computed, converted same that used (i) detecting which belongs to, then (ii), once assigned text, structure. First experiments performed by Blitzer dataset system demonstrated feasibility proposed approach.