In probabilistic approaches to classification and information extraction, one typically builds a statistical model of words under the assumption that future data will exhibit the same regularities as the training data. In many data sets, however, there are scope-limited features whose predictive power is only applicable to a certain subset of the data. For example, in information extraction from web pages, word formatting may be indicative of extraction category in different ways on different web pages. The difficulty with using such features is capturing and exploiting the new regularities encountered in previously unseen data. In this paper, we propose a hierarchical probabilistic model that uses both local/scope-limited features, such as word formatting, and global features, such as word content. The local regularities are modeled as an unobserved random parameter which is drawn once for each local data set. This random parameter is estimated during the inference process and then used to perform classification with both the local and global features--- a procedure which is akin to automatically retuning the classifier to the local regularities on each newly encountered web page. Exact inference is intractable and we present approximations via point estimates and variational methods. Empirical results on large collections of web data demonstrate that this method significantly improves performance from traditional models of global features alone.

In this paper we describe the architecture of the OntoGene Relation mining pipeline and some of its recent applications. With this research overview paper we intend to provide a contribution towards the recently started discussion towards standards for information extraction architectures in the biomedical domain. Our approach delivers domain entities mentioned in each input document, as well as candidate relationships, both ranked according to a confidency score computed by the system. This information is presented to the user through an advanced interface aimed at supporting the process of interactive curation.

Social networking websites such as Twitter have invigorated a wide range of studies in recent years ranging from consumer opinions on products to tracking the spread of diseases. While sentiment analysis and opinion mining from tweets have been studied extensively, surveillance of beliefs, especially those related to public health, have received considerably less attention. In our previous work, we proposed a model for surveillance of health beliefs on Twitter relying on the use of hand-picked probe statements expressing various health-related propositions. In this work we extend our model to automatically discover various probes related to public health beliefs. We present a data driven approach based on two distinct datasets and study the prevalence of public belief, disbelief or doubt for newly discovered probe statements.

Financial statements contain quantitative information and manager's subjective evaluation of firm's financial status. Using information released in U.S. 10-K filings. Both qualitative and quantitative appraisals are crucial for quality financial decisions. To extract such opinioned statements from the reports, we built tagging models based on the conditional random field (CRF) techniques, considering a variety of combinations of linguistic factors including morphology, orthography, predicate-argument structure, syntax, and simple semantics. Our results show that the CRF models are reasonably effective to find opinion holders in experiments when we adopted the popular MPQA corpus for training and testing. The contribution of our paper is to identify opinion patterns in multiword expressions (MWEs) forms rather than in single word forms. We find that the managers of corporations attempt to use more optimistic words to obfuscate negative financial performance and to accentuate the positive financial performance. Our results also show that decreasing earnings were often accompanied by ambiguous and mild statements in the reporting year and that increasing earnings were stated in assertive and positive way.

Sentiment Classification (SC) is about assigning a positive, negative or neutral label to a piece of text based on its overall opinion. This paper describes our in-progress work on extracting the meaning of words for SC. In particular, we investigate the utility of sense-level polarity information for SC. We first show that methods based on common classification features are not robust and their performance varies widely across different domains. We then show that sense-level polarity information features can significantly improve the performance of SC. We use datasets in different domains to study the robustness of the designated features. Our preliminary results show that the most common sense of the words result in the most robust results across different domains. In addition our observation shows that the sense-level polarity information is useful for producing a set of high-quality seed words which can be used for further improvement of SC task.

Twitter sentiment analysis (TSA) has become a hot research topic in recent years. The goal of this task is to discover the attitude or opinion of the tweets, which is typically formulated as a machine learning based text classification problem. Some methods use manually labeled data to train fully supervised models, while others use some noisy labels, such as emoticons and hashtags, for model training. In general, we can only get a limited number of training data for the fully supervised models because it is very labor-intensive and time-consuming to manually label the tweets. As for the models with noisy labels, it is hard for them to achieve satisfactory performance due to the noise in the labels although it is easy to get a large amount of data for training. Hence, the best strategy is to utilize both manually labeled data and noisy labeled data for training. However, how to seamlessly integrate these two different kinds of data into the same learning framework is still a challenge. In this paper, we present a novel model, called emoticon smoothed language model (ESLAM), to handle this challenge. The basic idea is to train a language model based on the manually labeled data, and then use the noisy emoticon data for smoothing. Experiments on real data sets demonstrate that ESLAM can effectively integrate both kinds of data to outperform those methods using only one of them.

Event extraction systems typically locate the role fillers for an event by analyzing sentences in isolation and identifying each role filler independently of the others. We argue that more accurate event extraction requires a view of the larger context to decide whether an entity is related to a relevant event. We propose a bottom-up approach to event extraction that initially identifies candidate role fillers independently and then uses that information as well as discourse properties to model textual cohesion. The novel component of the architecture is a sequentially structured sentence classifier that identifies event-related story contexts. The sentence classifier uses lexical associations and discourse relations across sentences, as well as domain-specific distributions of candidate role fillers within and across sentences. This approach yields state-of-the-art performance on the MUC-4 data set, achieving substantially higher precision than previous systems.

In digital libraries, a table, as a specific document component as well as a condensed way to present structured and relational data, contains rich information and often the only source of .that information. In order to explore, retrieve, and reuse that data, tables should be identified and the data extracted. Table recognition is an old field of research. However, due to the diversity of table styles, the results are still far from satisfactory, and not a single algorithm performs well on all different types of tables. In this paper, we randomly take samples from the CiteSeerX to investigate diverse table styles for automatic table extraction. We find that table headers are one of the main characteristics of complex table styles. We identify a set of features that can be used to segregate headers from tabular data and build a classifier to detect table headers. Our empirical evaluation on PDF documents shows that using a Random Forest classifier achieves an accuracy of 92%.

Web 2.0 has changed the ways people communicate, collaborate, and express their opinions and sentiments. But despite social data on the Web being perfectly suitable for human consumption, they remain hardly accessible to machines. To bridge the cognitive and affective gap between word-level natural language data and the concept-level sentiments conveyed by them, we developed SenticNet 2, a publicly available semantic and affective resource for opinion mining and sentiment analysis. SenticNet 2 is built by means of sentic computing, a new paradigm that exploits both AI and Semantic Web techniques to better recognize, interpret, and process natural language opinions. By providing the semantics and sentics (that is, the cognitive and affective information) associated with over 14,000 concepts, SenticNet 2 represents one of the most comprehensive semantic resources for the development of affect-sensitive applications in fields such as social data mining, multimodal affective HCI, and social media marketing.

Although web-based information extraction systems draw much attention, most of such systems assume that the web directly reflects the real world. For instance, Google flu trend, which is one of the-state-of-the-art influenza surveillance systems, relies on the basic idea that the amount of the influenza related search queries directly correlates with the number of the influenza patients. However, the real patients suffering from influenza symptoms are invisible in the web, because they do not use Internet. Considering this gap, this paper employs an infectious model, assuming that a potential patient utilizes Internet at the first sign of flu. The proposed model improves two types of the state-of-the-art systems, Google based system (from 0.837 correlation to 0.928) and Twitter based system (from 0.898 correlation to 0.918). This study demonstrated that a simple model could easily improve the web-based surveillance.