Kim Schouten

Survey on Aspect-Level Sentiment Analysis

The field of sentiment analysis, in which sentiment is gathered, analyzed, and aggregated from text, has seen a lot of attention in the last few years. The corresponding growth of the field has resulted in the emergence of various subareas, each addressing a different level of analysis or research question. This survey focuses on aspect-level sentiment analysis, where the goal is to find and aggregate sentiment on entities mentioned within documents or aspects of them. An in-depth overview of the current state-of-the-art is given, showing the tremendous progress that has already been made in finding both the target, which can be an entity as such, or some aspect of it, and the corresponding sentiment. Aspect-level sentiment analysis yields very fine-grained sentiment information which can be useful for applications in various domains. Current solutions are categorized based on whether they provide a method for aspect detection, sentiment analysis, or both. Furthermore, a breakdown based on the type of algorithm used is provided. For each discussed study, the reported performance is included. To facilitate the quantitative evaluation of the various proposed methods, a call is made for the standardization of the evaluation methodology that includes the use of shared data sets. Semanticallyrich concept-centric aspect-level sentiment analysis is discussed and identified as one of the most promising future research direction.

Semantics-based information extraction for detecting economic events

As today’s financial markets are sensitive to breaking news on economic events, accurate and timely automatic identification of events in news items is crucial. Unstructured news items originating from many heterogeneous sources have to be mined in order to extract knowledge useful for guiding decision making processes. Hence, we propose the Semantics-Based Pipeline for Economic Event Detection (SPEED), focusing on extracting financial events from news articles and annotating these with meta-data at a speed that enables real-time use. In our implementation, we use some components of an existing framework as well as new components, e.g., a high-performance Ontology Gazetteer, a Word Group Look-Up component, a Word Sense Disambiguator, and components for detecting economic events. Through their interaction with a domain-specific ontology, our novel, semantically enabled components constitute a feedback loop which fosters future reuse of acquired knowledge in the event detection process.

A semantic web-based approach for personalizing news

Hermes is an ontology-based framework for building news personalization services. This framework consists of a news classification phase, which classifies the news, a knowledge base updating phase, which keeps the knowledge base up-to-date, a news querying phase, allowing the user to search the news for concepts of interest, and a results presentation phase, showing the returned news items. The focus of this paper is on how to keep the knowledge base up-to-date. For this purpose, we elaborate on the updating phase that searches for key events in the news. Using rules based on patterns and actions, these events can be extracted and the knowledge base is updated. This is a semi-automatic process since user validation is required before updating the knowledge base.

Finding Implicit Features in Consumer Reviews for Sentiment Analysis

With the explosion of e-commerce shopping, customer reviews on the Web have become essential in the decision making process for consumers. Much of the research in this field focuses on explicit feature extraction and sentiment extraction. However, implicit feature extraction is a relatively new research field. Whereas previous works focused on finding the correct implicit feature in a sentence, given the fact that one is known to be present, this research aims at finding the right implicit feature without this pre-knowledge. Potential implicit features are assigned a score based on their co-occurrence frequencies with the words of a sentence, with the highest-scoring one being assigned to that sentence. To distinguish between sentences that have an implicit feature and the ones that do not, a threshold parameter is introduced, filtering out potential features whose score is too low. Using restaurant reviews and product reviews, the threshold-based approach improves the F1-measure by 3.6 and 8.7 percentage points, respectively.

The benefit of concept-based features for sentiment analysis

Sentiment analysis is an active field of research, moving from the traditional algorithms that operated on complete documents to fine-grained variants where aspects of the topic being discussed are extracted, as well as their associated sentiment. Recently, a move from traditional word-based approaches to concept-based approaches has started. In this work, it is shown by using a simple machine learning baseline, that concepts are useful as features within a machine learning framework. In all our experiments, the performance increases when including the concept-based features.

SPEED: a semantics-based pipeline for economic event detection

Nowadays, emerging news on economic events such as acquisitions has a substantial impact on the financial markets. Therefore, it is important to be able to automatically and accurately identify events in news items in a timely manner. For this, one has to be able to process a large amount of heterogeneous sources of unstructured data in order to extract knowledge useful for guiding decision making processes. We propose a Semantics-based Pipeline for Economic Event Detection (SPEED), aiming to extract financial events from emerging news and to annotate these with meta-data, while retaining a speed that is high enough to make real-time use possible. In our implementation of the SPEED pipeline, we reuse some of components of an existing framework and develop new ones, e.g., a high-performance Ontology Gazetteer and a Word Sense Disambiguator. Initial results drive the expectation of a good performance on emerging news.

Ontology-Enhanced Aspect-Based Sentiment Analysis

With many people freely expressing their opinions and feelings on the Web, much research has gone into modeling and monetizing opinionated, and usually unstructured and textual, Web-based content. Aspect-based sentiment analysis aims to extract the fine-grained topics, or aspects, that people are talking about, together with the sentiment expressed on those aspects. This allows for a detailed analysis of the sentiment expressed in, for instance, product and service reviews. In this work we focus on knowledge-driven solutions that aim to complement standard machine learning methods. By encoding common domain knowledge into a knowledge repository, or ontology, we are able to exploit this information to improve classification performance for both aspect detection and aspect sentiment analysis. For aspect detection, the ontology-enhanced method needs only 20% of the training data to achieve results comparable with a standard bag-of-words approach that uses all training data.

Semantics-Driven Implicit Aspect Detection in Consumer Reviews

With consumer reviews becoming a mainstream part of e-commerce, a good method of detecting the product or service aspects that are discussed is desirable. This work focuses on detecting aspects that are not literally mentioned in the text, or implicit aspects. To this end, a co-occurrence matrix of synsets from WordNet and implicit aspects is constructed. The semantic relations that exist between synsets in WordNet are exploited to enrich the co-occurrence matrix with more contextual information. Comparing this method with a similar method which is not semantics-driven clearly shows the benefit of the proposed method. Especially corpora of limited size seem to benefit from the added semantic context.

Aspect-Based Sentiment Analysis on the Web Using Rhetorical Structure Theory

Fine-grained sentiment analysis on the Web has received much attention in recent years. In this paper we suggest an approach to Aspect-Based Sentiment Analysis that incorporates structural information of reviews by employing Rhetorical Structure Theory. First, a novel way of determining the context of an aspect is presented, after which a full path analysis is performed on the found context tree to determine the aspect sentiment. Comparing the proposed method to a baseline model, which does not use the discourse structure of the text and solely relies on a sentiment lexicon to assign sentiments, we find that the proposed method consistently outperforms the baseline on three different datasets.

Review-level aspect-based sentiment analysis using an ontology

The rapid growth of the World Wide Web has led to an explosion of information that is available on this platform. This has resulted in an increased interest in sentiment analysis, where the goal is to determine the opinion regarding a topic. Aspect-based sentiment analysis aims to capture the sentiment within a segment of text for mentioned aspects, rather than for the text as a whole. The task we consider is aspect-based sentiment analysis at the review-level for restaurant reviews. We focus on ontology-enhanced methods that complement a standard machine learning algorithm. For this task we use two different algorithms, a review-based and a sentence aggregation algorithm. By using an ontology as a knowledge base, the classification performance of our models improves significantly. Furthermore, the review-based algorithm gives more accurate predictions than the sentence aggregation algorithm.