Rude waiter but mouthwatering pastries! An exploratory study into Dutch Aspect-Based Sentiment Analysis

Rude Waiter but Mouthwatering Pastries! An Exploratory Study into Dutch Aspect-Based Sentiment Analysis Orph´ee De Clercq and V´eronique Hoste LT3, Language and Translation Technology Team Ghent University Groot-Brittanni¨elaan 45, 9000 Ghent, Belgium firstname.lastname@ugent.be Abstract The fine-grained task of automatically detecting all sentiment expressions within a given document and the aspects to which they refer is known as aspect-based sentiment analysis. In this paper we present the first full aspect-based sentiment analysis pipeline for Dutch and apply it to customer reviews. To this purpose, we collected reviews from two different domains, i.e. restaurant and smartphone reviews. Both corpora have been manually annotated using newly developed guidelines that comply to standard practices in the field. For our experimental pipeline we perceive aspect-based sentiment analysis as a task consisting of three main subtasks which have to be tackled incrementally: aspect term extraction, aspect category classification and polarity classification. First experiments on our Dutch restaurant corpus reveal that this is indeed a feasible approach that yields promising results. Keywords: user-generated content, aspect-based sentiment analysis, semantic processing 1. Introduction In today’s information society, it cannot be ignored that large parts of our lives are spent and shared online. The arrival of Web 2.0 sites allowed site visitors to add content, called user-generated content (Moens et al., 2014). Typical for user-generated content is that it contains a lot of sub- jective material. As the amount of online information has grown exponentially, so has the interest in new text mining techniques to handle and analyze this growing amount of subjective text. One of the main research topics is sentiment analysis, also known as opinion mining. The objective of sentiment anal- ysis is the extraction of subjective information from text, rather than factual information. Originally, it focused on the task of automatically classifying an entire document as positive, negative or neutral (Liu, 2012). More recently, the focus has shifted from coarse-grained to fine grained sen- timent analysis, where sentiment has to be assigned at the clause level (Wilson et al., 2009). Often, users are not only interested in people’s general sen- timents about a certain product, but also in their opinions about specific features, i.e. parts or attributes of that prod- uct. The task of automatically detecting all sentiment ex- pressions within a given document and the concepts and aspects (or features) to which they refer is known as aspect- based or feature-based sentiment analysis, i.e. ABSA (Pon- tiki et al., 2014). Such systems do not only try to distinguish the positive from the negative utterances, but also strive to detect the target of the opinion, which comes down to a very fine-grained sentiment analysis task. In this paper we present two Dutch domain-specific cor- pora annotated for ABSA and the first pipeline for auto- matically performing this task on Dutch customer reviews. In Section 2 we describe which two corpora were collected, one comprising restaurant reviews and another comprising smartphone reviews, and how these have been manually an- notated using newly developed guidelines that comply to standard …

Content truncated to 500 words