Are Topics Interesting or Not? An LDA-based Topic-graph Probabilistic Model for Web Search Personalization

In this article, we propose a Latent Dirichlet Allocation– (LDA) based topic-graph probabilistic personalization model for Web search. This model represents a user graph in a latent topic graph and simultaneously estimates the probabilities that the user is interested in the topics, as well as the probabilities that the user is not interested in the topics. For a given query issued by the user, the webpages that have higher relevancy to the interested topics are promoted, and the webpages more relevant to the non-interesting topics are penalized. In particular, we simulate a user’s search intent by building two profiles: A positive user profile for the probabilities of the user is interested in the topics and a corresponding negative user profile for the probabilities of being not interested in the the topics. The profiles are estimated based on the user’s search logs. A clicked webpage is assumed to include interesting topics. A skipped (viewed but not clicked) webpage is assumed to cover some non-interesting topics to the user. Such estimations are performed in the latent topic space generated by LDA. Moreover, a new approach is proposed to estimate the correlation between a given query and the user’s search history so as to determine how much personalization should be considered for the query. We compare our proposed models with several strong baselines including state-of-the-art personalization approaches. Experiments conducted on a large-scale real user search log collection illustrate the effectiveness of the proposed models.

Does More Context Help? Effects of Context Window and Application Source on Retrieval Performance

We study the effect of contextual information obtained from a user’s digital trace on Web search performance. Contextual information is modeled using Dirichlet–Hawkes processes (DHP) and used in augmenting Web search queries. The context is captured by monitoring all naturally occurring user behavior using continuous 24/7 recordings of the screen and associating the context with the queries issued by the users. We report a field study in which 13 participants installed a screen recording and digital activity monitoring system on their laptops for 14 days, resulting in data on all Web search queries and the associated context data. A query augmentation (QAug) model was built to expand the original query with semantically related terms. The effects of context window and source were determined by training context models with temporally varying context windows and varying application sources. The context models were then utilized to re-rank the QAug model. We evaluate the context models by using the Web document rankings of the original query as a control condition compared against various experimental conditions: (1) a search context condition in which the context was sourced from search history; (2) a non-search context condition in which the context was sourced from all interactions excluding search history; (3) a comprehensive context condition in which the context was sourced from both search and non-search histories; and (4) an application-specific condition in which the context was sourced from interaction histories captured on a specific application type. Our results indicated that incorporating more contextual information significantly improved Web search rankings as measured by the positions of the documents on which users clicked in the search result pages. The effects and importance of different context windows and application sources, along with different query types are analyzed, and their impact on Web search performance is discussed.

Browsers’ Private Mode: Is It What We Were Promised?

Web browsers are one of the most used applications on every computational device in our days. Hence, they play a pivotal role in any forensic investigation and help determine if nefarious or suspicious activity has occurred on that device. Our study investigates the usage of private mode and browsing artefacts within four prevalent web browsers and is focused on analyzing both hard disk and random access memory. Forensic analysis on the target device showed that using private mode matched each of the web browser vendors’ claims, such as that browsing activity, search history, cookies and temporary files that are not saved in the device’s hard disks. However, in volatile memory analysis, a majority of artefacts within the test cases were retrieved. Hence, a malicious actor performing a similar approach could potentially retrieve sensitive information left behind on the device without the user’s consent.

A Survey on Blockchain-Based Search Engines

With the increasing growth of different types of data, search engines have become an essential tool on the Internet. Every day, billions of queries are run through few search engines with several privacy violations and monopoly problems. The blockchain, as a trending technology applied in various fields, including banking, IoT, education, etc., can be a beneficial alternative. Blockchain-based search engines, unlike monopolistic ones, do not have centralized controls. With a blockchain-based search system, no company can lay claims to user’s data or access search history and other related information. All these data will be encrypted and stored on a blockchain. Valuing users’ searches and paying them in return is another advantage of a blockchain-based search engine. Additionally, in smart environments, as a trending research field, blockchain-based search engines can provide context-aware and privacy-preserved search results. According to our research, few efforts have been made to develop blockchain use, which include studies generally in the early stages and few white papers. To the best of our knowledge, no research article has been published in this regard thus far. In this paper, a survey on blockchain-based search engines is provided. Additionally, we state that the blockchain is an essential paradigm for the search ecosystem by describing the advantages.

NN-QuPiD Attack: Neural Network-Based Privacy Quantification Model for Private Information Retrieval Protocols

Web search engines usually keep users’ profiles for multiple purposes, such as result ranking and relevancy, market research, and targeted advertisements. However, user web search history may contain sensitive and private information about the user, such as health condition, personal interests, and affiliations that may infringe users’ privacy since a user’s identity may be exposed and misused by third parties. Numerous techniques are available to address privacy infringement, including Private Information Retrieval (PIR) protocols that use peer nodes to preserve privacy. Previously, we have proved that PIR protocols are vulnerable to the QuPiD Attack. In this research, we proposed NN-QuPiD Attack, an improved version of QuPiD Attack that uses an Artificial Neural Network (RNN) based model to associate queries with their original users. The results show that the NN-QuPiD Attack gave 0.512 Recall with the Precision of 0.923, whereas simple QuPiD Attack gave 0.49 Recall with the Precision of 0.934 with the same data.

Praktykować media/Praktykować siebie… O doświadczaniu przestrzeni późnej nowoczesności

In advanced globalization, the digital code, computer, and Internet become tools of cultural change. 2-3-4.0 generation media produce images, artificial events, objects, as well as methods of vision prevailing in design, design processes, and communication. The number of visibility types offered to users, results in media matrices becoming regimes of (for) the eye, of (for) a bodily experience. It also shapes a sense of reality by means of media space (TV stream, cyberspace) and spaces inscribed in media forms (simulations, hybrids, onto-ontological topias). Two issues seem of particular relevance here. Questions about access to reality offered by ‘images’ (satellites, HDMI helmets, combat glasses, etc.) and possibilities of seeing the world from behind the visual media matrices. Media, multi-media communication platforms distribute not only schemes of viewing, but also the right to look (extranet, friends lists, access to archives and libraries, consumer profiles collected by Google, MS, etc.). The excess of artificial forms and spaces, on the other hand, directs attention to participation and participants of culture: individuals and communities (real, imagined, virtual). Their activities in cyberspace, including identity and identity practices, should be the focus of interest. The author of the text reflects upon identity practices of the Praktykować media/Praktykować siebie individual, that is the individual’s participation in culture through the prism of immaterial materiality, self-care, and the need for bonds and integration with the Others. She discusses techniques of advanced audio-vision (their matrices and spatiality), as well as creation of subjective coherence negotiated with others in terms of individual – group (conventional) – universal content (e.g. humanity). Identity practices are located within the framework of data flows and transmissions, information bubbles, heterotopia, and identities of legitimization, resistance, and design implemented there. The author of the text also perceives identity from the perspective of a person and a mask, of users’ tactics articulating their presence in cyber-virtual communities (e.g. avatars, nicks, multiple identities), as well as in the perspective of data visualization: user profiles, metric identifiers, files, selection algorithms, search history, archives of published photos…, in other words – institutional (cyber)surveillance strategies.

Recipe sharing on Pinterest- healthfulness assessment through nutrition information analysis and natural language processing (Preprint)

BACKGROUND Although Pinterest has become a popular platform for distributing influential information that shapes users’ behaviors, the role of recipes pinned on Pinterest has not been well understood. OBJECTIVE To explore patterns of food ingredients and the nutritional content of recipes posted on Pinterest, and examine the factors associated with recipes that engaged more users. METHODS Data were randomly collected from Pinterest between June 28 and July 12, 2020 (207 recipes and 2,818 comments). All samples were collected via two new user accounts with no search history. A codebook was developed with a raw agreement rate of 0.97 across all variables. Content analysis and a novel natural language processing (NLP) sentiment analysis technique were employed. RESULTS Recipes using seafood or vegetables as the main ingredient had on average fewer calories and less sodium, sugar, and cholesterol compared to meat- or poultry-based recipes. For recipes using meat as the main ingredient, more energy was from fat (56.6%). Although the most followed pinners tended to post recipes containing more poultry/seafood and less meat, recipes serving higher fat or providing more calories per serving were more popular, having more shared photos/videos and comments. The NLP-based sentiment analysis suggested that Pinterest users weighted “taste” more heavily than “complexity” (less than 8% of comments) and “health” (less than 3% of comments). CONCLUSIONS While popular pinners tended to post recipes with more seafood/poultry/vegetables and less meat, recipes with higher fat and sugar content were more user-engaging, with more photo/video shares and comments. Data on Pinterest behaviors can inform developing and implementing nutrition health interventions on promoting healthy recipes on social media platforms.

A Search History-Driven Offspring Generation Method for the Real-Coded Genetic Algorithm

In evolutionary algorithms, genetic operators iteratively generate new offspring which constitute a potentially valuable set of search history. To boost the performance of offspring generation in the real-coded genetic algorithm (RCGA), in this paper, we propose to exploit the search history cached so far in an online style during the iteration. Specifically, survivor individuals over the past few generations are collected and stored in the archive to form the search history. We introduce a simple yet effective crossover model driven by the search history (abbreviated as SHX). In particular, the search history is clustered, and each cluster is assigned a score for SHX. In essence, the proposed SHX is a data-driven method which exploits the search history to perform offspring selection after the offspring generation. Since no additional fitness evaluations are needed, SHX is favorable for the tasks with limited budget or expensive fitness evaluations. We experimentally verify the effectiveness of SHX over 15 benchmark functions. Quantitative results show that our SHX can significantly enhance the performance of RCGA, in terms of both accuracy and convergence speed. Also, the induced additional runtime is negligible compared to the total processing time.