Articles tagged with Search Engines
The SciNet search engine outperforms current ones by displaying keywords related to user searches in a topic radar, enabling faster and more diverse search results. This technology helps people find information more efficiently, even when they're unsure of their search query.
Dartmouth researchers have created an AI software that uses images to locate documents on the internet with greater accuracy than ever before. The system, tested on over 600 search queries, shows a 30% improvement in precision over traditional text-based search engines.
A new study by Columbia Business School found that consumers using less popular search terms have higher click activity and are more targetable for paid advertising. The research also showed that day of the week has no impact on click likelihood, with consumers investing more effort in products they're searching for.
A new INFORMS study finds that buying competitors' brand names as keywords on search engines can backfire if the advertising brand is inferior to the rival. The authors provide a framework for choosing keywords among brand names, offering advertisers a way to optimize their search ads.
Researchers created a prototype automated system to check millions of lines of code for consistency with privacy promises. They employed a new language, Legalease, and used Grok to annotate existing programs, allowing for high accuracy in compliance checks.
A survey by Xiaojing Sheng and Penny Simpson found that reading privacy policies can lead to decreased trust in providers, with over a quarter of respondents citing complexity as the main reason for not reading them. Almost one third of respondents expressed concern about their private information being shared or sold.
Researchers developed AIDA software that disambiguates named entities by analyzing mentions in texts and their connections to Wikipedia. The approach enables precise search results, including categorization and object-based searches.
Researchers at Tecnalia developed a new approach to improve web page search results by analyzing image content. The method relies on the subject matter of web pages inferred from images and analyzed by search engines to decide relevance.
Researchers at NC State have developed a technique for personalized searches that can handle complex queries and scale up to support thousands of concurrent users. The new approach uses ambient query context to determine user intent in real-time, allowing for more accurate results and better handling of ambiguous searches.
A University of Oregon study reveals how librarians adapted to the Internet by redefining their occupational identity from expert knowledge to interpreting and connecting patrons with relevant information
A new technique developed by researchers at Concordia University has slashed data-crunching time for law enforcement, allowing investigators to extract hidden knowledge from large volumes of text. The new method automatically identifies criminal topics and provides visualization of social networks among participants.
A Northwestern University study found that many college students cannot track down reliable information on emergency contraception pills (ECPs) online. Only two-thirds of the participants were able to conclude that their friend should seek ECPs, with fewer than half providing the ideal response.
Researchers found that brand names in online search engine advertising campaigns attract more attention and encourage more sales. A study by Penn State researchers discovered that branded keyword phrases combined with branded advertisements generated the highest average sales, 15 times higher than generic ads.
Oren Etzioni proposes a future where computers can find answers by analyzing data on the internet, going beyond keyword searches. The University of Washington's Turing Center has developed an open-source tool called ReVerb to achieve this goal.
A new study published in INFORMS Journal found that online consumers are willing to pay a premium for purchases from vendors with clear and protective privacy policies. In an experimental study, participants were more likely to buy from high-privacy sites than low-privacy sites, with an average additional payment of $0.59 for batteries.
Researchers found that real-time search terms have significant economic value, with unique searches generating $33 million in revenue daily. The study, published in Information Processing and Management, suggests a growing market for real-time search, driven by the increasing number of Twitter updates.
Researchers develop three-step approach to identify clusters of like-minded bloggers using social annotations. The method calculates similarities between bloggers based on their blog posts' keywords and ratings.
Academic users must be aware of how search engines operate to ensure quality sources are chosen. José van Dijck argues that Google Scholar is a co-producer of academic knowledge, not just a neutral tool. Its algorithm prioritizes quantitative citation analysis over relevance and truth-value.
Researchers created a machine-learning model called DiaTM that learns vernacular terms for health problems and symptoms, improving medical website performance. The system achieves a 25% improvement in nDCG, a scientific term referring to the relevance of information retrieval.
A Northwestern University study found that many college students trust websites and search engines ranked high on Google, prioritizing brand recognition over factual accuracy. The study involved 102 students who showed a strong preference for Google and other top brands, even when given hypothetical information-seeking tasks.
The Fraunhofer Institute developed FlashFind to enable swift, intuitive full-text search on mobile devices for large datasets. Users can quickly locate streets, cities, and points of interest without network access or limited keyboards.
Researchers found that religious searching behaviors remained mainstream and technologically savvy, despite secularization trends. Search engines poorly delivered relevant content, likely due to localized religious websites.
Researchers found search engines are used for fact-checking and supporting higher-level information needs, revealing a more complex role in learning. Their study suggests design improvements for search engines can better address users' cognitive and affective aspects.
Researchers developed a protocol to distort internet search engine user profiles, preserving privacy from profiling and data exploitation. The new system offers an improvement in response time while delaying searches slightly.
Researchers found that readers better understood, remembered and emotionally responded to material found through searching compared to surfing. Information was also better retained when individuals conducted specific searches for information.
Researchers created a music recommender system that generates playlists as good as or better than Genius, but with less user data required. The system uses auto-tagging algorithms to label songs based on their acoustic content, leading to high-quality playlists including lesser-known songs.
The OpenRide mobile ridesharing service uses intelligent route matching to connect drivers with passengers in real-time, providing a convenient solution for last-minute journeys. The system's use of open interfaces enables integration with additional partners, expanding its potential market reach.
A Penn State researcher analyzed nearly 1 million Web searches to detect patterns of query reformulation and create models to predict them. The research found that users reformulated their search terms in 22% of queries to more precisely convey the information they were searching for.
Researchers developed a system to extract subjective expressions from web pages, scoring them for subjectivity and indicating whether they express personal opinions or marketing speak. The method outperformed general search engines in detecting personal opinion pages across four categories.
A new music search engine developed by UC San Diego engineers uses artificial intelligence to analyze songs and provide accurate song suggestions based on user input. The system uses machine learning algorithms to identify patterns in music genres, emotions, and instruments, enabling users to find specific songs through descriptive words.
Webscalers' technology integrates multiple small search engines to retrieve a vast amount of information, including deep Web content. The system aims to provide fresher results and overcome limitations of existing search engines like Google and Yahoo.
A study by Penn State researchers identified nine factors that can help predict future click-through rates, including number of records in a search and browser type. The positive factors had five effects, while four had negative effects, with user intent having no significant impact on predicting click-throughs.
Researchers found ad clicks on Dogpile.com to be only 15% of total interactions, with 35% of queries yielding no clicks. The study challenges previous reports of a 30% ad click-through rate and suggests advertisers can leverage alternative forms of keyword advertising.
A study published in the International Journal of Internet Marketing and Advertising suggests that non-profits can exploit online marketing strategies to gain traffic and raise awareness. By focusing on relevant keywords and optimizing their websites, non-profits can improve their search engine rankings and increase donations.
A researcher at Penn State is examining how users choose a search engine, evaluate its performance, and react to individual results. The study aims to build a multi-layer model of branding in search engine selection and final web page evaluation.
Researchers analyzed over 1.5 million queries to find that 80% are informational, while 10% are navigational or transactional in nature. An algorithm developed by Penn State researchers achieved a 74% accuracy rate in classifying these searches.
A study from Tufts University reveals a digital skills divide among socio-economic groups, affecting parents' ability to access trustworthy child development information. Parents in higher SES groups outperform lower SES counterparts in evaluating search engines and building searches.
Researchers found that people tend to focus on information supporting their pre-existing beliefs, even if it's accurate. The study suggests designing search engine interfaces to help users organize information and overcome decision biases.
A study of over 7,500 websites found that webmasters intentionally favor Google over other search engines in robots.txt files, resulting in a strong correlation between the favored robots and market share. This bias may disadvantage consumers who prefer other search engines, as Google can index some information not available to them.
The system uses speech recognition and language processing technology to create transcripts of MIT lectures, allowing users to search for specific topics. The prototype web site has drawn increased traffic, primarily from international students accessing MIT's OpenCourseWare initiative.
A transderivational search engine will be developed to help people discover connections between text, audio, image, geometry, and motion data. The researchers aim to enhance the synaesthetic effect in analogy generation and advance artificial intelligence through this technology.
The UC San Diego researchers have created an online game called Listen Game that allows players to label songs with words, generating crucial data for building a music search engine. The system uses machine learning to train computers to annotate songs accurately.
TableSeer automates table extraction and searching across documents, improving efficiency for researchers. The engine's innovative ranking algorithm, TableRank, also identifies frequently cited tables in search results.
ChemxSeer is a search engine designed specifically for chemical formulae, allowing scientists to find relevant research on CH4 or methane without filtering out results about Channel 4. The tool identifies related chemicals with different formula representations and substructures, providing more accurate search results.
A study by MU researchers found that people pay more attention to pictures selected from a small array of choices than from a large array. Participants who chose from six thumbnail pictures remembered the pictures with 99 percent accuracy, while those who chose from 24 only remembered with 89 percent accuracy.
A new search engine mashup uses intelligent virtual robots to scan data quickly, assessing keywords and context, and combining it with meta search tools to retrieve relevant results. This approach improves the precision rate and recall rate of traditional search engines.
A study by Penn State found that branding matters when searching online, with participants favoring Yahoo! and Google over MSN Live Search and an in-house engine. The researchers believe emotional branding has an impact on the evaluation of search engine results, leading to a 15% higher score for Yahoo!.
A new Carnegie Mellon University study found that people are more likely to buy from online merchants with good privacy policies, which were identified by a search engine called Privacy Finder. Participants were willing to pay about 60 cents extra on a $15 purchase when buying from a site with a privacy policy they liked.
The UCSD system uses a clever image indexing technique that allows it to cover larger collections of images at a lower computational cost. It outperforms existing approaches in terms of annotation and retrieval accuracy, as well as efficiency.
The Digital Locke Project has made John Locke's first texts available online, with a digital critical edition and search functionality. The project aims to produce a definitive text of Locke's work, building on previously rediscovered manuscripts.
A recent study by IU informatics researchers challenged the notion of a Web-dominating 'Googlearchy' where search engines direct most traffic to established mainstream sites. The study found that popular sites receive significantly less traffic than predicted, indicating a more even distribution.
WebIC uses machine learning to predict users' information needs, filtering out irrelevant results and providing suggested sites. Developed by the University of Alberta, the software can be incorporated with search engines or downloaded directly onto individual computers.
The new tool uses natural language processing and historical knowledge to extract entities from historical documents, allowing users to quickly locate specific individuals or locations. The system will enable users to browse lists of entities and their frequencies within individual documents and the collection as a whole.
A new web site compares search results of Google and Yahoo across four countries, highlighting the differences between Western and Eastern censored versions. The research project CenSEARCHip reveals starkly different outcomes for political topics like human rights and democracy.
A recent study by Dr. Jim Jansen found that five popular search engines return nearly identical results in terms of effectiveness, with little benefit for occasional online shoppers. Niche e-commerce search engines may provide a slight increase in relevant results for frequent online consumers.
Researchers at institutions in developing countries can now access the scientific subset of the web using Scirus and TEK, reducing connectivity charges. The collaboration between Elsevier and MIT aims to provide a robust and user-friendly service to researchers globally.
The Next Generation CiteSeer project will expand the database, add services, and provide opportunities for joint authoring. The new architecture will be open source, enabling easier use and greater access to metadata.
A study by Penn State researchers found that 80% of participants preferred organic search results over sponsored links. Despite identical content, participants rated sponsored links as less relevant due to biased perception. To address this, businesses should prioritize organic visibility alongside paid advertising.
A new search engine, Unintended Information Revelation (UIR), has been developed to uncover hidden information within documents. It uses concept chain graphs to find connections between seemingly unrelated concepts and can reveal sensitive data.
A new image search technique developed by Mirela Tanase recognizes visual patterns and significantly increases the number of correctly identified objects in a search. The method provides a useful addition to existing methods, such as MPEG7, and can improve image recognition rates in certain cases.