Sentiment analysis systems seek to extract emotions and feelings expressed about people, organisations, nation states, goods and services, in free natural language texts.  This interdisciplinary workshop will address three related topics:
 

This workshop will deal with the recent advances in the processing of sentiment in arbitrary collections of text. Sentiment can be expressed about works of art and literature, about the state of financial markets, about liking and disliking individuals, organisations, ideologies, and consumer goods. It is necessary to examine what aspects of emotional experience sentiment analysis aims to capture, how and in what way this may be evaluated. This workshop focuses on three strands of research which will serve to enhance the development of automated sentiment analysis systems of free text for real world applications.

Firstly, in psychology and computational linguistics, the notions of emotion and metaphor interact in a number of complex ways. It has been argued that conceptual metaphors underlie human understanding and processing of emotion. In addition, it can be argued that the expression of sentiment and its interpretation can rely critically on how a speaker or writer uses metaphor. Therefore, an understanding of how emotion is expressed and perceived in language is not complete without addressing the role of figurative language and metaphor as basic scaffolding or tool for modulating affective text content.

Secondly, to date, sentiment analysis typically deals with a specific domain of ideal objects. In order to build a sentiment analysis system, one has to understand "what there is" in a given domain, i.e. the ontology of the domain. In this context, is it possible to conceive of generic sentiment analysis? Practitioners in this area need to examine the requirements and challenges of an approach that could cross boundaries of domain or time or even language where different communities of use, languages or cultures may express or even experience sentiments in different ways.

Finally, work in sentiment analysis may be regarded as work in intelligent information retrieval and success is evaluated in terms of accuracy in identifying the affective content of information segments. Yet sentiment analysis has the potential to have a powerful impact in other domains that require input about emotional context. Researchers in Human-Computer Interaction, Affective Computing, Lexicography and Terminography, may become end-users of work in sentiment analysis and sentiment analysis folks may have much to learn from how a machine artificially endowed with emotions/sentiments behaves. It may become feasible to evaluate sentiment analysis systems in terms of the performance of such applications. An examination of alternative end-user systems and evaluation mechanisms can only serve to enrich the field of sentiment analysis and present new challenges for researchers to address.

Prof Sam Glucksberg will give the keynote lecture of the workshop entitled