Tuong Tri Nguyen

-Spear: A New Method for Expert Based Recommendation Systems

Recommendation systems are based on a fast and effective personalized mechanism to provide items relevant to users. In this article, an expert-based approach for recommendation is proposed. We extend the spamming-resistant expertise analysis and ranking (SPEAR) algorithm to determine a set of experts from a set of attributes and values, calling the modification the -SPEAR algorithm. This system can recommend a set of items to users using expert opinions. In this approach, we use ontology to build profiles of users. The experimental results are implemented in the movie domain as a case study. Our data set was collected from IMDB and MovieLens data sets.

The Perceived Value of Passive Animal Health Surveillance: The Case of Highly Pathogenic Avian Influenza in Vietnam

Economic evaluations are critical for the assessment of the efficiency and sustainability of animal health surveillance systems and the improvement of their efficiency. Methods identifying and quantifying costs and benefits incurred by public and private actors of passive surveillance systems (i.e. actors of veterinary authorities and private actors who may report clinical signs) are needed. This study presents the evaluation of perceived costs and benefits of highly pathogenic avian influenza (HPAI) passive surveillance in Vietnam. Surveys based on participatory epidemiology methods were conducted in three provinces in Vietnam to collect data on costs and benefits resulting from the reporting of HPAI suspicions to veterinary authorities. A quantitative tool based on stated preference methods and participatory techniques was developed and applied to assess the non‐monetary costs and benefits. The study showed that poultry farmers are facing several options regarding the management of HPAI suspicions, besides reporting the following: treatment, sale or destruction of animals. The option of reporting was associated with uncertain outcome and transaction costs. Besides, actors anticipated the release of health information to cause a drop of markets prices. This cost was relevant at all levels, including farmers, veterinary authorities and private actors of the upstream sector (feed, chicks and medicine supply). One benefit associated with passive surveillance was the intervention of public services to clean farms and the environment to limit the disease spread. Private actors of the poultry sector valued information on HPAI suspicions (perceived as a non‐monetary benefit) which was mainly obtained from other private actors and media.

When private actors matter: Information-sharing network and surveillance of Highly Pathogenic Avian Influenza in Vietnam

The effectiveness of animal health surveillance systems depends on their capacity to gather sanitary information from the animal production sector. In order to assess this capacity we analyzed the flow of sanitary information regarding Highly Pathogenic Avian Influenza (HPAI) suspicions in poultry in Vietnam. Participatory methods were applied to assess the type of actors and likelihood of information sharing between actors in case of HPAI suspicion in poultry. While the reporting of HPAI suspicions is mandatory, private actors had more access to information than public actors. Actors of the upstream sector (medicine and feed sellers) played a key role in the diffusion of information. The central role of these actors and the influence of the information flow on the adoption by poultry production stakeholders of behaviors limiting (e.g. prevention measures) or promoting disease transmission (e.g. increased animal movements) should be accounted for in the design of surveillance and control programs.

Social Tagging Analytics for Processing Unlabeled Resources:A Case Study on Non-geotagged Photos

Social networking services (SNS) have been an important sources of geotagged resources. This paper proposes Naive Bayes method-based framework to predict the locations of non-geotagged resources on SNS. By computing TF-ICF weights (Term Frequency and Inverse Class Frequency) of tags, we discover meaningful associations between the tags and the classes (which refer to sets of locations of the resources). As the experimental result, we found that the proposed method has shown around 75% of accuracy, with respect to F1 measurement.

PageRank-based approach on ranking social events: A case study with Flickr

Exploring social events from Social Network Services (SNSs) (known as detecting events) has been studied in many researches because of its challenges. Most of researches focus on detecting events based on textual context. In this paper, we propose a novel framework using media data for not only systematically identifying events but also ranking these events. Firstly, we detect events from the photos textual annotations as well as visual features (e.g., timestamp, location); and then effectively identify events by considering the spreading effect of events in the spatio-temporal space. Secondly, we use these relationships among events (e.g., event spatial, temporal and content) for enhancing the precision of the algorithm. Finally, we rank events by analyzing relationships between them (e.g., locations, timestamps, tags) at different period of time. The experiments are conducted with two different approaches: (i) using a collected dataset (offline approach), and (ii) using a realtime dataset (online approach).

Extending HITS Algorithm for Ranking Locations by Using Geotagged Resources

The paper focuses on using geotagged resources from the social network service (SNS) for searching the famous places from keyword. We extend the HITS[9] algorithm in order to rank locations which are collected from geotagged resources on SNS. Our approach not only uses the similarity measurement between locations’tags for computing the value of locations but also calculate the term frequency of tags which occur in each location to modify the value of tags for ranking. We implement and show the experimental results with the set of locations from the geotagged resources.

A Consensus-Based Method to Enhance a Recommendation System for Research Collaboration

With the development of scientific societies, research problems are increasingly complex, requiring scientists to collaborate to solve them. The quality of collaboration between researchers is a major factor in determining their achievements. This study proposes a collaboration recommendation method that takes into account previous research collaboration and research similarities. Research collaboration is measured by combining the collaboration time and the number of co-authors who already collaborated with an author. Research similarity is based on authors’ previous publications and academic events they attended. In addition, a consensus-based algorithm is proposed to integrate bibliography data from different sources, such as the DBLP Computer Science Bibliography, ResearchGate, CiteSeer, and Google Scholar. The experimental results show that this proposal improves the accuracy of the recommendation systems, in comparison with other methods.

Active Learning-Based Approach for Named Entity Recognition on Short Text Streams

The named entity recognition (NER) problem has an important role in many natural language processing (NLP) applications and is one of the fundamental tasks for building NLP systems. Supervised learning methods can achieve high performance but they require a large amount of training data that is time-consuming and expensive to obtain. Active learning (AL) is well-suited to many problems in NLP, where unlabeled data may be abundant but labeled data is limited. The AL method aims to minimize annotation costs while maximizing the desired performance from the model. This study proposes a method to classify named entities from Tweet streams on Twitter by using an AL method with different query strategies. The samples were queried for labeling by human annotators based on query by committee and diversity-based querying. The experiments evaluated the proposed method on Tweet data and achieved promising results that proved better than the baseline.

User Timeline and Interest-Based Collaborative Filtering on Social Network

A lot of users and large amount of information have been posted and shared through on-line systems. User timeline and interest are important features on recommendation systems (e.g., user likes watching action movies in the morning, and likes watching drama movies in the afternoon however he/she likes watching thriller movies in the evening) and also on social network. There are some recommendation applications have been developed on social network to support users selecting what kind of wanted items based on user timeline and interest. However, there is not any approaches based on user timeline and interest have been proposed that user interest have been separated into partitions of user interest. Thus, a recommendation mechanism will be applied on social networks based on extracting user timeline and user interest that is necessary. In this paper, we propose a new approach that user interest will be determined on a set of time partitions.

Discovering Co-author Relationship in Bibliographic Data Using Similarity Measures and Random Walk Model

Discovering the research communities to bring techniques to the world is an interesting topic. In this paper we use the DBLP data to investigate the co-author relationship in a real bibliographic network and predict the interactions between co-authors. We analysis the research trend of authors and conferences based on extracted keywords from paper titles. We can understand research fields and change of research trend to find appropriate co-authors and conferences to submit our work. We also find potential co-authors for an existing author in DBLP data by using a variety of similarity measures and a random walk model. It can be useful for building a recommendation system.