Adam Bermingham

Classifying sentiment in microblogs: is brevity an advantage?

Microblogs as a new textual domain offer a unique proposition for sentiment analysis. Their short document length suggests any sentiment they contain is compact and explicit. However, this short length coupled with their noisy nature can pose difficulties for standard machine learning document representations. In this work we examine the hypothesis that it is easier to classify the sentiment in these short form documents than in longer form documents. Surprisingly, we find classifying sentiment in microblogs easier than in blogs and make a number of observations pertaining to the challenge of supervised learning for sentiment analysis in microblogs.

Topic-dependent sentiment analysis of financial blogs

While most work in sentiment analysis in the financial domain has focused on the use of content from traditional finance news, in this work we concentrate on more subjective sources of information, blogs. We aim to automatically determine the sentiment of financial bloggers towards companies and their stocks. To do this we develop a corpus of financial blogs, annotated with polarity of sentiment with respect to a number of companies. We conduct an analysis of the annotated corpus, from which we show there is a significant level of topic shift within this collection, and also illustrate the difficulty that human annotators have when annotating certain sentiment categories. To deal with the problem of topic shift within blog articles, we propose text extraction techniques to create topic-specific sub-documents, which we use to train a sentiment classifier. We show that such approaches provide a substantial improvement over full documentclassification and that word-based approaches perform better than sentence-based or paragraph-based approaches.

A study of inter-annotator agreement for opinion retrieval

Evaluation of sentiment analysis, like large-scale IR evaluation, relies on the accuracy of human assessors to create judgments. Subjectivity in judgments is a problem for relevance assessment and even more so in the case of sentiment annotations. In this study we examine the degree to which assessors agree upon sentence-level sentiment annotation. We show that inter-assessor agreement is not contingent on document length or frequency of sentiment but correlates positively with automated opinion retrieval performance. We also examine the individual annotation categories to determine which categories pose most difficulty for annotators.

Sentiment in Oil Markets

Sentiment is shown to influence both West Texas Intermediate (WTI) and Brent futures prices during the period 2002–2013. This is demonstrated while controlling for stock indices, exchange rates, financial costs, inventory and supply levels as well as OPEC activity. Sentiment indices are developed for WTI and Brent crude oils using a suite of financial proxies similar to those used in equity research where the influence of sentiment has already been established. Given the novel nature of this study, multiple hypothesis testing techniques are used to ensure that these conclusions are statistically robust.

Automatically recommending multimedia content for use in group reminiscence therap

This paper presents and evaluates a novel approach for automatically recommending multimedia content for use in group reminiscence therapy for people with Alzheimers and other dementias. In recent years recommender systems have seen popularity in providing a personalised experience in information discovery tasks. This personalisation approach is naturally suited to tasks in healthcare, such as reminiscence therapy, where there has been a trend towards an increased emphasis on person-centred care. Building on recent work which has shown benefits to reminiscence therapy in a group setting, we develop and evaluate a system, REMPAD, which profiles people with Alzheimers and other dementias, and provides multimedia content tailored to a given group context. In this paper we present our system and approach, and report on a user trial in residential care settings. In our evaluation we examine the potential to use early-aggregation and late-aggregation of group member preferences using case-based reasoning combined with a content-based method. We evaluate with respect to accuracy, utility and perceived usefulness. The results overall are positive and we find that our best-performing approach uses early aggregation CBR combined with a content-based method. Also, under different evaluation criteria, we note different performances, with certain configurations of our approach providing better accuracy and others providing better utility.

Design and Field Evaluation of REMPAD: A Recommender System Supporting Group Reminiscence Therapy

This paper describes a semi-automated web-based system to facilitate digital reminiscence therapy for patients with mild-to-moderate dementia, enacted in a group setting. The system, REMPAD, uses proactive recommendation technology to profile participants and groups, and offers interactive multimedia content from the Internet to match these profiles. In this paper, we focus on the design of the system to deliver an innovative personalized group reminiscence experience. We take a user-centered design approach to discover and address the design challenges and considerations. A combination of methodologies is used throughout this research study, including exploratory interviews, prototype use case walkthroughs, and field evaluations. The results of the field evaluation indicate high user satisfaction when using the system, and strong tendency towards repeated use in future. These studies provide an insight into the current practices and challenges of group reminiscence therapy, and inform the design of a multimedia recommender system to support facilitators and group therapy participants.

Recommending Video Content for Use in Group-Based Reminiscence Therapy

REMPAD is a semi-automated cloud-based system used to facilitate digital reminiscence therapy for patients with mild-to-moderate dementia, enacted in a group setting. REMPAD uses profiles for participants and groups to proactively recommend interactive video content from the Internet to match these profiles. In this chapter, we focus on the design of the system and then the system architecture, the system build, data curation, and usage scenarios. We also report a series of steps carried out as part of our user-centered design approach to system development, and a series of analyses on interaction logs which indicate various levels of effectiveness for different configurations of the recommendation algorithm we use. The results indicate high user satisfaction when using the system, and strong tendency towards repeated use in future.