PATHOS 2013 - 2nd Workshop on Practice and Theory of Opinion Mining and Sentiment Analysis

The abundance of opinions available on the World Wide Web represents an information repository of enormous intellectual and economic value. Automated methods to exploit this rich knowledge mine have become more and more relevant within the last decade and the availability of large amounts of data is an ideal premise for the application of empirical methods.
Although many researchers from different nations and institutes intensively work on the development of these techniques, many challenges have been left uncovered: Among the highly relevant social media are new text types, that are fairly different from the types on which research in natural language processing has been conducted for the last 20 years. The most prominent example may be Twitter; with its condensed tweets it presents a sublanguage of its own. Traditional approaches often leave questions about the true nature of opinions unanswered. The actual emotion hidden in opinionated text is still hard to uncover; current lexical resources for sentiment analysis mostly only contain information about the polarity or subjectivity of terms, lacking relevant information of their emotional category. Fear or anger are treated equally, as are hope and joy. Emotional categories are projected onto two polarities, i.e. positive and negative, which oversimplifies reality. A further important aspect requiring methodological coverage is the exact analysis of the entities participating in the event evoked by an opinion. Robust linguistic techniques embedded in data-driven methods will possibly guide the way to answer questions on the actual target of an opinion or on the peer group holding this opinion. Another challenge is the automated assignment of polarity or subjectivity labels to the plethora of sentiment-related textual data readily available on the Web. State-of-the-art learning approaches, such as weakly-supervised or semi-supervised methods or distant supervision still need to be thoroughly examined for this purpose.
Another important aspect of this workshop is its focus on multi-linguality. Therefore, submissions for approaches devised and applied on languages other than English, in particular German, are strongly encouraged.
The workshop aims for providing a platform for researchers interested in the upcoming challenges of sentiment analysis and opinion mining. It intends to attract researchers settled in computational linguistics, natural language processing, and artificial intelligence alike. The workshop will be organized by the recently founded GSCL-Interest Group on German Sentiment Analysis (IGGSA).
Topics of Interest
We invite contributions on, but not necessarily limited to:
Genre- and domain-specific sentiment analysis
Cross-genre and cross-domain sentiment analysis
Semi-supervised/weakly supervised learning for sentiment analysis
Contrast of machine learning vs. linguistic approaches vs. hybrid methods
Sentiment analysis on twitter and social media in general
Fine-to-coarse sentiment analysis
Emotion detection and classification
Representation of and calculus on emotions
Multi-lingual sentiment analysis
Lexical resources for opinion mining and sentiment analysis
Gold standards and methods for evaluation
Real-world applications and large-scale sentiment analysis
Trends and perspectives in the field