While microblogging has gained popularity on the Internet, analyzing and processing short messages has become a challenging task in natural language processing. This paper analyzes the differences between Internet short messages (or “microtext”) and general articles by comparing the Plurk Corpus and the Sinica Balanced Corpus. Likelihood ratio and the tóngyìcícílín thesaurus are adopted to analyze the lexical semantics of frequent terms in each corpus. Furthermore, the NTUSD sentiment dictionary is used to compare the sentiment distribution of the two corpora. The result is also applied to sentiment transition analysis.

We present in this paper, the application of a novel approach to computational modeling, understanding and detection of social phenomena in online multi-party discourse. A two-tiered approach was developed to detect a collection of social phenomena deployed by participants, such as topic control, task control, disagreement and involvement. We discuss how the mid-level social phenomena can be reliably detected in discourse and these measures can be used to differentiate participants of online discourse. Our approach works across different types of online chat and we show results on two specific data sets.

A content analysis of Wikipedia barnstars personalized tokens of appreciation given to participants reveals a wide range of valued work extending beyond simple editing to include social support, administrative actions, and types of articulation work. Barnstars are examples of short semi-structured text characterized by informal grammar and language. We propose a method to classify these barnstars which contain items of interest into various work type categories.We evaluate several multilabel text categorization classifiers and show that significant performance can be achieved by simple classifiers using features which carry context extracted from barnstars. Although this study focused specifically on work categorization via barnstar content for Wikipedia, we believe that the findings are applicable to other similar collaborative systems

Analyzing chat traffic has important applications for both the military and the civilian world. This paper presents a case study of a real-world application of chat analysis in support of team training exercise in the military. It compares the results of an unsupervised learning approach with those of a supervised classification approach. The paper also discusses some of the specific challenges presented by this domain.

This paper focuses on performing Sentiment Analysis of Twitter by adapting data from other domains, commonly referred to as Domain Adaptation. While we show that Domain Adaptation is useful in predicting sentiments, we propose different techniques to select an out-of-domain data source that would aid in Sentiment Analysis. Additionally, we suggest two iterative algorithms based on Expectation-Maximization (EM) and Rocchio SVM that filter noisy data during adaptation and train only on valid data. Finally, we explore a couple of metrics, Mutual Information and Cosine distance to measure similarity between different domains of data. We use Twitter and Blippr as data sources and perform binary sentiment (positive and negative sentiments) classification.

How widespread is the phenomenon of commenting or editing a tweet in the practice of retweeting by members of political communities in Twitter? What is the nature of comments(agree/disagree), or of edits (change audience, change meaning, curate content). Being able to answer these questions will provide knowledge that will help answering other questions such as: what are the topics, events, people that attract more discussion (in forms of commenting) or controversy (agree/disagree)? Who are the users who engage in the processing of curating content by inserting hashtags or adding links? Which political community shows more enthusiasm for an issue and how broad is the base of engaged users? How can detection of agreement/disagreement in conversations inform sentiment analysis - the technique used to make predictions (who will win an election) or support insightful analytics (which policy issue resonates more with constituents). We argue that is necessary to go beyond the much-adopted aggregate text analysis of the volume of tweets, in order to discover and understand phenomena at the level of single tweets. This becomes important in the light of the increase in the number of human-mimicking bots in Twitter. Genuine interaction and engagement can be better measured by analyzing tweets that display signs of human intervention. Editing the text of an original tweet before it is retweeted, could reveal mindful user engagement with the content, and therefore, would allow us to perform sampling among real human users. This paper presents work in progress that deals with the challenges of discovering retweets that contain comments or edits, and outlines a machine-learning based strategy for classifying the nature of such comments.

This paper reports on a work in progress whose goal is to use Latent Dirichlet Allocation (LDA) to discover topic clusters within a small set of Twitter posts. Preliminary results indicate that micro-documents are amenable to topic clustering via LDA provided that a) only nouns and verbs are used; b) posts are “padded” with words of similar meaning to those used in the posts. These preliminary findings are consistent with the fact that probabilistic topic models look for word co-occurrences in documents and hence require that topic-indicative words appear together many times throughout the data sample. The results of this pilot study are to be extended to a larger data set of Twitter posts to see whether padding” can counteract the growing size of the data and the presence of numerous information-sparse posts.

We build a mechanism to form an ontology of entities which improves a relevance of matching and searching microtext. Ontology construction starts from the seed entities and mines the web for new entities associated with them. To form these new entities, machine learning of syntactic parse trees (syntactic generalization) is applied to form commonalities between various search results for existing entities on the web. Ontology and syntactic generalization are applied to relevance improvement in search and text similarity assessment in commercial setting; evaluation results show substantial contribution of both sources to microtext processing.

Rather, the goal is to Microtext media (Ellen, 2011), such as SMS, IM, Twitter, gather relevant messages, organize them, and extract some and text chat, have in common that they use short strings other kind of useful information from them, such as how for immediate communication or broadcast. Microtext can well a team is performing or what people are talking about be construed as one form of micro-messaging (e.g., and when. However, micro-messages do not exist in a Milstein, et al., 2008) which we extend here to include any vacuum; they are contextually oriented and may be part of of a number of other modalities (e.g., telephone calls, a larger network of communications which includes email, face-to-face interaction) used for short, immediate and telephone and other media, including "macro-text." Given (potentially) persistent message passing among this, we have found that natural language processing of the coordinating agents. In this paper, we describe several microtext must be paired with temporal or network recent attempts to study micro-messaging military and analysis of the context. To demonstrate this process, we related organizational contexts.

In a separate study, we were interested in understanding people's Q&A habits on Twitter. Finding questions within Twitter turned out to be a difficult challenge, so we considered applying some traditional NLP approaches to the problem. On the one hand, Twitter is full of idiosyncrasies, which make processing it difficult. On the other it is very restricted in length and tends to employ simple syntactic constructions, which could help the performance of NLP processing. In order to find out the viability of NLP and Twitter, we built a pipeline of tools to work specifically with Twitter input for the task of finding questions in tweets. This work is still preliminary, but in this paper we discuss the techniques we used and the lessons we learned.