Search

Finding and recommending suitable services for moving, connected devices are increasingly important due to the popularity of mobile Internet. While recent research has attempted to use role-based approaches to recommend services, role discovery is still an ongoing research topic. This paper proposes a simple and yet efficient, matrix-based, context-aware role mining method to automatically group users according to their interests and habits. Using a question-based approach, we also allow popular mobile services to be recommended to other members in the same group in a context dependent.

Manual knowledge acquisition is extremely laborious and time consuming. In this paper, we propose a new automatic method for domain knowledge acquisition by semantic annotating and pattern mining. This method makes use of Minipar to label sentences and extract structural patterns. Semantic bank is proposed to annotate and represent concepts with semantic labels considering sentence context. The method can further learn and assign relations to previously extracted concepts by pattern matching. The involved concepts and semantic labels with learned relations together construct a domain knowledge base. Preliminary experiments on Yahoo! Data in “heart diseases” category show that this method is feasible for automatic domain knowledge acquisition.

This paper presents two user interfaces for a pattern-based User Interactive Question Answering system, which are designed to encourage users to ask questions by using semantic patterns. One is a Guide-Based User Interface (GBUI), which can guide users with clear instructions. However, it involves many steps and the operation may become tedious. The other is a Recommendation-Based User Interface (RBUI), which recommends a few relevant patterns containing automatically suggested details for each free-text question. However, the recommended patterns may not always be satisfactory and sometimes the user's revision is needed. In comparing these two user interfaces, we propose a new Complexity Evaluation Model (CEM) to evaluate the complexity on the basis of user log study and a realistic focused user study. The results of the study user logs, which cover a test set of 1605 users and 488 semantic patterns, show that RBUI can improve the complexity of GBUI by 39.8% on average. The improvement is also confirmed by the user study. It has thus become clear that RBUI can improve the usability of the UIQA system in terms of helping the system accumulate high quality pattern-based questions.

Non-invasive utility location techniques are vital for maximizing the potential of trenchless technologies, especially within an urban environment. This paper proposes the approach of implementing a knowledge-based system (KBS) for predicting the geophysical properties of the soil by taking into account both the geographical and seasonal variation in soil properties. The key motivation for developing KBS as a prediction methodology is the advantage of being able to gather and integrate both analytical models as well as expert knowledge based upon existing and opportunistic data. The design and implementation of this system are presented as a case study for ground penetrating radar, which is mainly affected by the complex permittivity and electrical conductivity of the soil. Soil electromagnetic semiempirical models are implemented to obtain an estimate of the complex permittivity using geotechnical properties obtained in laboratory tests or found on a national database. Hydrology models are applied to account for variations in soil water content and the associated impact on permittivity values due to changes in precipitation patterns. Results obtained from a ground penetrating radar test site with a long-term monitoring station installed are also presented, showing agreement between measured data and KBS output, and demonstrating potential for improving the effectiveness of utility surveys.

The ground in which utility service pipelines are buried inevitably controls, to a large degree, the structural performance and progressive deterioration of the pipelines themselves. In a parallel programme of research to the UK Mapping the Underworld (MTU) project, a study of the fundamental properties of the ground, and how they change with the seasons and local physical and chemical contexts, is being conducted at the University of Birmingham, UK. While the results of this study feed into both the operational protocols for the MTU multi-sensor location device and the associated knowledge based system (KBS) that is being created to aid its deployment (both topics being the subjects of separate papers to this conference), the suite of complementary research projects on the ground and its properties provide valuable insights to the pipeline engineer. Geophysics is being used by the research team to explore the state of the ground with the aim of highlighting areas of concern for the structural health of pipelines buried in the ground. Studies of cast iron pipeline corrosion mechanisms have focussed on the changes that the reaction products cause to the surrounding soils, with a particular emphasis on clay soils, and one interesting finding is that these clay-iron reaction products can make the pipelines ‘invisible’ to standard geophysical location devices. Moreover there are other features in the ground that are being targeted (voids, ground wetting and softening due to leakage, ground weakening due to progressive erosion), and these features effectively make the ground more or less ‘visible’ to geophysical technologies. Alongside this work, bespoke tests have been developed for use on site to ‘calibrate’ the geophysics, thereby enhancing the signatures of the features. This paper introduces these parallel research projects and draws out the important findings for pipeline engineers charged with establishing the condition of existing buried assets.

Semantic annotation for text is a well-studied topic. However, little contribution has been engaged in the application of short text annotation. In this article, an automatic annotation approach is proposed for such purpose, which annotates short text with semantic labels for question answering systems. In the first step, keywords are extracted from a question and then a semantic label selection module is used to select semantic labels to tag keywords. If there is no appropriate label, WordNet is employed to obtain candidate labels to annotate those keywords by calculating the similarity between each keyword in the question and the concept list in our predefined Tagger Ontology. To improve the accuracy of annotation, we also design a naïve Bayesian based method to distinguish multi-senses and assign best semantic labels by referring to historically annotated questions. Preliminary experiments on 6 categories show our approach achieves the precision of 76% in average. Refereed/Peer-reviewed

A personalized e-learning framework based on a user-interactive question-answering (QA) system is proposed, in which a user-modeling approach is used to capture personal information of students and a personalizedanswer extraction algorithm is proposed for personalizedautomatic answering. In our approach, a topic ontology (or concept hierarchy) of course content defined by an instructor is used for the system to generate the corresponding structure of boards for holding relevant questions. Students can interactively post questions, and also browse, select, and answer others’ questions in their interested boards. A knowledge base is accumulated usinghistorical question/answer (Q/A) pairs for knowledge reuse. The students’ log data are used to build an association space to compute the interest and authority of the students for each board and each topic. The personal information of students can help instructors design suitable teaching materials to enhance instruction efficiency, be used to implement the personalized automatic answering and distribute unsolved questions to relevant students to enhance the learning efficiency. The experiment results show the efficacy of our user-modeling approach.

An automatic method for text categorizing and ranking search engine's results by semantic similarity is proposed in this paper. We first obtain nouns and verbs from snippets obtained from search engine using Name Entity Recognition and part-of speech. A semantic similarity algorithm based on WordNet is proposed to calculate the similarity of each snippet to each of the pre-defined categories. A balanced similarity ranking method combined with Google's rank and timeliness of the pages is proposed to rank these snippets. Preliminary experiments with 500 labeled questions from TREC03 show that 72.7% are correctly categorized. Refereed/Peer-reviewed

An approach of automatic question generation from given learning material of medical text is presented in this paper. The main idea is to generate the questions automatically based on question templates which are created by training on many medical articles. In order to provide interesting questions, our research focuses on medical related concepts. This method can be used for evaluation of learner's comprehension after he/she finished a reading material. Different from traditional learning system the articles and questions are all prepared beforehand; participants can learn whatever new input medical articles with the help of automatic question generation.

A new semantic pattern is proposed in this paper, which can be used by users to post questions and answers in user-interactive question answering (QA) system. The necessary procedures of using semantic pattern in a QA system are also presented, which include question structure analysis, pattern matching, pattern generation, pattern classification and answer extraction. A user interface of using semantic pattern is also implemented in our QA system, which allows users to effectively post and answer questions. It gains good overall user satisfaction.

Display Omitted A feature-based approach identifies clinical trials with similar eligibility text.Identified clusters are effective according to crowd-sourced evaluation.Trial clusters can facilitate similarity-based trial search and systematic review. ObjectivesTo automatically identify and cluster clinical trials with similar eligibility features. MethodsUsing the public repository ClinicalTrials.gov as the data source, we extracted semantic features from the eligibility criteria text of all clinical trials and constructed a trial-feature matrix. We calculated the pairwise similarities for all clinical trials based on their eligibility features. For all trials, by selecting one trial as the center each time, we identified trials whose similarities to the central trial were greater than or equal to a predefined threshold and constructed center-based clusters. Then we identified unique trial sets with distinctive trial membership compositions from center-based clusters by disregarding their structural information. ResultsFrom the 145,745 clinical trials on ClinicalTrials.gov, we extracted 5,508,491 semantic features. Of these, 459,936 were unique and 160,951 were shared by at least one pair of trials. Crowdsourcing the cluster evaluation using Amazon Mechanical Turk (MTurk), we identified the optimal similarity threshold, 0.9. Using this threshold, we generated 8806 center-based clusters. Evaluation of a sample of the clusters by MTurk resulted in a mean score 4.331?0.796 on a scale of 1-5 (5 indicating "strongly agree that the trials in the cluster are similar"). ConclusionsWe contribute an automated approach to clustering clinical trials with similar eligibility features. This approach can be potentially useful for investigating knowledge reuse patterns in clinical trial eligibility criteria designs and for improving clinical trial recruitment. We also contribute an effective crowdsourcing method for evaluating informatics interventions.

An automatic method for building a semantic dictionary from existing questions in a pattern-based question answering system is proposed for question categorization. This dictionary consists of two main parts: Semantic Domain Terms (SDT), which is a domain specific term list, and Semantic Labeled Terms (SLT), which contain common terms tagged with semantic labels. The semantic dictionary is built using the proposed method on a set of 2509 questions with semantic patterns in our system. 3390 questions without semantic patterns are used as ground truth to test its performance. Experimental results show that the precision of question classification is improved by 7.5% in average after using the constructed semantic dictionary compared with the baseline method.