Sub-document Level Research Articles

Combining contextualized word representation and sub-document level analysis through Bi-LSTM+CRF architecture for clinical de-identification

Clinical de-identification aims to identify Protected Health Information in clinical data, enabling data sharing and publication. First automatic de-identification systems were based on rules or on machine learning methods, limited by language changes, lack of context awareness and time consuming feature engineering. Newer deep learning techniques for sequence labeling have shown better results with a reduction in feature engineering efforts and the use of word representation techniques in vector space. However, they are not able to jointly represent the polysemic and context-dependent nature of words, as well as their morpho-syntactic mutations characteristic of handwriting. To address these limitations, a new de-identification approach based on deep learning techniques for Named Entity Recognition has been proposed, whose key factors are: (i) a Bidirectional Long Short-Term Memory + Conditional Random Field architecture for sequence labeling that takes advantage of the widest possible representation context; (ii) a contextualized language model, working at character level, to capture the polysemy of words and manage the morpho-syntactic variations typical of handwritten notes; (iii) more word representations stacked to better capture latent syntactic and semantic similarities. This approach has been tested on the official Informatics for Integrating Biology & the Bedside 2014 de-identification dataset, showing similar or higher performance than state of the art with respect to category and binary recognition, but without any feature engineering or handcrafted rules. The experiments demonstrate the effectiveness of the proposed approach, in particular with regard to category level recognition which is essential to correctly replace entities with surrogates for anonymization purposes.

Understanding and Supporting Cross-Device Web Search for Exploratory Tasks with Mobile Touch Interactions

Mobile devices enable people to look for information at the moment when their information needs are triggered. While experiencing complex information needs that require multiple search sessions, users may utilize desktop computers to fulfill information needs started on mobile devices. Under the context of mobile-to-desktop web search, this article analyzes users’ behavioral patterns and compares them to the patterns in desktop-to-desktop web search. Then, we examine several approaches of using Mobile Touch Interactions (MTIs) to infer relevant content so that such content can be used for supporting subsequent search queries on desktop computers. The experimental data used in this article was collected through a user study involving 24 participants and six properly designed cross-device web search tasks. Our experimental results show that (1) users’ mobile-to-desktop search behaviors do significantly differ from desktop-to-desktop search behaviors in terms of information exploration, sense-making and repeated behaviors. (2) MTIs can be employed to predict the relevance of click-through documents, but applying document-level relevant content based on the predicted relevance does not improve search performance. (3) MTIs can also be used to identify the relevant text chunks at a fine-grained subdocument level. Such relevant information can achieve better search performance than the document-level relevant content. In addition, such subdocument relevant information can be combined with document-level relevance to further improve the search performance. However, the effectiveness of these methods relies on the sufficiency of click-through documents. (4) MTIs can also be obtained from the Search Engine Results Pages (SERPs). The subdocument feedbacks inferred from this set of MTIs even outperform the MTI-based subdocument feedback from the click-through documents.

Exploring Co‐training strategies for opinion detection

For the last decade or so, sentiment analysis, which aims to automatically identify opinions, polarities, or emotions from user‐generated content (e.g., blogs, tweets), has attracted interest from both academic and industrial communities. Most sentiment analysis strategies fall into 2 categories: lexicon‐based and corpus‐based approaches. While the latter often requires sentiment‐labeled data to build a machine learning model, both approaches need sentiment‐labeled data for evaluation. Unfortunately, most data domains lack sufficient quantities of labeled data, especially at the subdocument level. Semisupervised learning (SSL), a machine learning technique that requires only a few labeled examples and can automatically label unlabeled data, is a promising strategy to deal with the issue of insufficient labeled data. Although previous studies have shown promising results of applying various SSL algorithms to solve a sentiment‐analysis problem, co‐training, an SSL algorithm, has not attracted much attention for sentiment analysis largely due to its restricted assumptions. Therefore, this study focuses on revisiting co‐training in depth and discusses several co‐training strategies for sentiment analysis following a loose assumption. Results suggest that co‐training can be more effective than can other currently adopted SSL methods for sentiment analysis.

Modeling object characteristics of dynamic Web content

Requests for dynamic and personalized content have increasingly become a significant part of Internet traffic, driven both by a growth in dynamic Web services and a “trickle-down” effect stemming from the effectiveness of caches and content-distribution networks at serving static content. To efficiently serve this trend, several server-side and cache-side techniques have recently been proposed. Although such techniques, which exploit different forms of reuse at the sub-document level, appear promising, a significant impediment to their widespread deployment is (1) the absence of good models describing characteristics of dynamic Web content and (2) the lack of effective synthetic content generators, which reduce the effort involved in verifying the effectiveness of a proposed solution. This paper addresses both of these shortcomings. Its primary contribution is a set of models that capture the characteristics of dynamic content both in terms of independent parameters such as the distributions of object sizes and their freshness times, as well as derived parameters such as content reusability across time and linked documents. These models are derived from an analysis of the content from six representative news and e-commerce sites, using both size-based and level-based splitting techniques to infer document objects. A secondary contribution is a Java-based dynamic content emulator, which uses these models to generate edge-side include-based dynamic content and serve requests for whole documents as well as separate objects.