Universal dependencies (UD) is a framework for morphosyntactic annotation of human language, which to date has been used to create treebanks for more than 100 languages. In this article, we outline the linguistic theory of the UD framework, which draws on a long tradition of typologically oriented grammatical theories. Grammatical relations between words are centrally used to explain how predicate–argument structures are encoded morphosyntactically in different languages while morphological features and part-of-speech classes give the properties of words. We argue that this theory is a good basis for cross-linguistically consistent annotation of typologically diverse languages in a way that supports computational natural language understanding as well as broader linguistic studies.

Manual text annotation is an essential part of Big Text analytics. Although annotators work with limited parts of data sets, their results are extrapolated by automated text classification and affect the final classification results. Reliability of annotations and adequacy of assigned labels are especially important in the case of sentiment annotations. In the current study we examine inter-annotator agreement in multiclass, multi-label sentiment annotation of messages. We used several annotation agreement measures, as well as statistical analysis and Machine Learning to assess the resulting annotations.

No abstract

Currently 19%-28% of Internet users participate in online health discussions. A 2011 survey of the US population estimated that 59% of all adults have looked online for information about health topics such as a specific disease or treatment. Although empirical evidence strongly supports the importance of emotions in health-related messages, there are few studies of the relationship between a subjective lan-guage and online discussions of personal health. In this work, we study sentiments expressed on online medical forums. As well as considering the predominant sentiments expressed in individual posts, we analyze sequences of sentiments in online discussions. Individual posts are classified into one of five categories. We identified three categories as sentimental (encouragement, gratitude, confusion) and two categories as neutral (facts, endorsement). 1438 messages from 130 threads were annotated manually by two annotators with a strong inter-annotator agreement (Fleiss kappa = 0.737 and 0.763 for posts in se-quence and separate posts respectively). The annotated posts were used to analyse sentiments in consec-utive posts. In four multi-class classification problems, we assessed HealthAffect, a domain-specific af-fective lexicon, as well general sentiment lexicons in their ability to represent messages in sentiment recognition

Currently 19%-28% of Internet users participate in online health discussions. In this work, we study sentiments expressed on online medical forums. As well as considering the predominant sentiments expressed in individual posts, we analyze sequences of sentiments in online discussions. Individual posts are classified into one of the five categories encouragement, gratitude, confusion, facts, and endorsement. 1438 messages from 130 threads were annotated manually by two annotators with a strong inter-annotator agreement (Fleiss kappa = 0.737 and 0.763 for posts in sequence and separate posts respectively). The annotated posts were used to analyse sentiments in consecutive posts. In automated sentiment classification, we applied HealthAffect, a domain-specific lexicon of affective words.