The document contains charts and graphs showing the results of experiments comparing different systems for trustworthiness analysis of web search results. One chart shows that a combination of the authors' system and Google performed better than Google alone across 10 different query categories. Another chart shows the average precision of 4 different algorithms for determining the credibility of a data pair.
This study was presented in the 20th International Conference on Web Information Systems Engineering (WISE 2019), 19-22 January 2020.
-----
Fumiaki Saito, Yoshiyuki Shoji and Yusuke YamamotoHighlighting Weasel Sentences for Promoting Critical Information Seeking on the WebProceedings of the 20th International Conference on Web Information Systems Engineering (WISE 2019), pp.424-440, Hong Kong, China, November 2019.
This document discusses link analysis and PageRank, an algorithm for identifying important nodes in large network graphs. It begins with an overview of graph data structures and the goal of identifying influential nodes. It then introduces PageRank, explaining its basic assumptions and showing examples of how it calculates node importance scores. The document discusses problems with the initial PageRank approach and how it was improved with the Complete PageRank algorithm. Finally, it briefly introduces Topic-sensitive PageRank, which aims to identify important nodes related to specific topics.
Matrix factorization techniques can be used to address some of the limitations of traditional collaborative filtering approaches for recommender systems. Matrix factorization decomposes the user-item rating matrix into the product of two lower-dimensional matrices, one representing latent factors for users and the other for items. This reduced dimensionality addresses data sparsity and scalability issues. Specifically, singular value decomposition is often used to perform this matrix factorization, which can approximate the original rating matrix while ignoring less important singular values and factor vectors. The decomposed matrices can then be multiplied to predict unknown user ratings.
This document discusses item-based collaborative filtering for recommender systems. It describes how item-based collaborative filtering works by predicting a target user's rating for an item based on the ratings of similar items. It highlights advantages over user-based filtering like lower computational cost and more stable similarity computations. Key aspects covered include using cosine similarity to calculate item similarities, adjusting for individual rating biases, selecting the top K similar items, and predicting ratings based on similar items' ratings.
This document discusses recommender systems and collaborative filtering. It introduces user-based collaborative filtering, which predicts a user's rating for an item based on the ratings from similar users. Similarity between users is calculated using the Pearson correlation coefficient. The ratings of the top K most similar users are then averaged to predict the target user's rating.
This document presents the development of a Web Access Literacy Scale to measure users' abilities to critically evaluate information found online. The researchers conducted a study with 534 participants to develop and validate the scale. Factor analysis resulted in a 7-factor scale measuring logical approach, content verification strategies, inquisitiveness, bias tolerance, search skills, author verification, and objectivity. Scores were higher for those with information literacy experience. The scale can help identify weaknesses and inform the development of literacy training and search tools.
The document contains charts and graphs showing the results of experiments comparing different systems for trustworthiness analysis of web search results. One chart shows that a combination of the authors' system and Google performed better than Google alone across 10 different query categories. Another chart shows the average precision of 4 different algorithms for determining the credibility of a data pair.
This study was presented in the 20th International Conference on Web Information Systems Engineering (WISE 2019), 19-22 January 2020.
-----
Fumiaki Saito, Yoshiyuki Shoji and Yusuke YamamotoHighlighting Weasel Sentences for Promoting Critical Information Seeking on the WebProceedings of the 20th International Conference on Web Information Systems Engineering (WISE 2019), pp.424-440, Hong Kong, China, November 2019.
This document discusses link analysis and PageRank, an algorithm for identifying important nodes in large network graphs. It begins with an overview of graph data structures and the goal of identifying influential nodes. It then introduces PageRank, explaining its basic assumptions and showing examples of how it calculates node importance scores. The document discusses problems with the initial PageRank approach and how it was improved with the Complete PageRank algorithm. Finally, it briefly introduces Topic-sensitive PageRank, which aims to identify important nodes related to specific topics.
Matrix factorization techniques can be used to address some of the limitations of traditional collaborative filtering approaches for recommender systems. Matrix factorization decomposes the user-item rating matrix into the product of two lower-dimensional matrices, one representing latent factors for users and the other for items. This reduced dimensionality addresses data sparsity and scalability issues. Specifically, singular value decomposition is often used to perform this matrix factorization, which can approximate the original rating matrix while ignoring less important singular values and factor vectors. The decomposed matrices can then be multiplied to predict unknown user ratings.
This document discusses item-based collaborative filtering for recommender systems. It describes how item-based collaborative filtering works by predicting a target user's rating for an item based on the ratings of similar items. It highlights advantages over user-based filtering like lower computational cost and more stable similarity computations. Key aspects covered include using cosine similarity to calculate item similarities, adjusting for individual rating biases, selecting the top K similar items, and predicting ratings based on similar items' ratings.
This document discusses recommender systems and collaborative filtering. It introduces user-based collaborative filtering, which predicts a user's rating for an item based on the ratings from similar users. Similarity between users is calculated using the Pearson correlation coefficient. The ratings of the top K most similar users are then averaged to predict the target user's rating.
This document presents the development of a Web Access Literacy Scale to measure users' abilities to critically evaluate information found online. The researchers conducted a study with 534 participants to develop and validate the scale. Factor analysis resulted in a 7-factor scale measuring logical approach, content verification strategies, inquisitiveness, bias tolerance, search skills, author verification, and objectivity. Scores were higher for those with information literacy experience. The scale can help identify weaknesses and inform the development of literacy training and search tools.