A Personalized Sentiment Model with Textual and Contextual Information

Proceedings of the 23rd Conference on Computational Natural Language Learning, pages 992–1001 Hong Kong, China, November 3-4, 2019. c⃝2019 Association for Computational Linguistics 992 A Personalized Sentiment Model with Textual and Contextual Information Siwen Guo Sviatlana H¨ohn ILIAS Research Lab, CSC, University of Luxembourg {siwen.guo,sviatlana.hoehn,christoph.schommer}@uni.lu Christoph Schommer Abstract In this paper, we look beyond the traditional population-level sentiment modeling and con- sider the individuality in a person’s expres- sions by discovering both textual and contex- tual information. In particular, we construct a hierarchical neural network that leverages valuable information from a person’s past ex- pressions, and offer a better understanding of the sentiment from the expresser’s perspective. Additionally, we investigate how a person’s sentiment changes over time so that recent in- cidents or opinions may have more effect on the person’s current sentiment than the old ones. Psychological studies have also shown that individual variation exists in how easily people change their sentiments. In order to model such traits, we develop a modified at- tention mechanism with Hawkes process ap- plied on top of a recurrent network for a user- specific design. Implemented with automati- cally labeled Twitter data, the proposed model has shown positive results employing different input formulations for representing the con- cerned information. 1 Introduction Sentiment is one of the key factors affecting hu- man behavior. Studying the way in which sen- timent is perceived, evolved and expressed is an essential part in artificial intelligence. To an- alyze sentiment in text, researchers have made different assumptions on linguistic behaviors that are leveraged with approaches developed based on the nature of the text, the representation of the related information and the objectives. How- ever, majority of the studies are conducted at the population-level which assumes that people fol- low a common understanding with regard to the use of language. Such approaches can be inaccu- rate in the cases where people use the same lexical choices to convey different messages or vice versa. Harris (2006) stated that ‘no two alike’ showing the inherent difference in human that motivates the research of personalized sentiment analysis. Grounded in the psychological works, we argue that it is significant to study the effect of individu- ality in the expressions and to investigate the pos- sibility of providing a deeper understanding of the expressions from the writers’ own perspectives. In this work, we concern the use of preferred lexi- cal choices when expressing sentiment (Reiter and Sripada, 2002) and the level of consistency in re- taining a sentiment (Janis and Field, 1956). Besides the targeted text message itself, we ex- ploit two types of contextual information for the purpose of realizing the psychological aspects: a person’s expressions in the past and the time when the expressions were made. With the goal of dis- covering the effect of the contextual information, distinct formulation methods are proposed to in- tegrate the information in the personalized senti- ment model. The backboned model is a hierarchi- cal neural network which follows a conventional embedding – recurrent – attention …

Content truncated to 500 words