Articles tagged with Natural Language Processing
A breakthrough in optical neural networks accelerates computing speed and processing power to over 1000 times that of previous processors. The system can process record-sized images and achieve full facial image recognition.
Researchers have discovered subnetworks within BERT that can complete the same task more efficiently, reducing computing costs and increasing accessibility to state-of-the-art natural language processing. The 'lottery ticket hypothesis' identifies these leaner subnetworks, which can be repurposed for multiple tasks.
A new study examines the 'word-of-machine' effect, where consumer preference for AI recommenders is influenced by the importance of utilitarian versus hedonic attributes. When utilitarian features are emphasized, consumers prefer AI over human assistance, while hedonic features lead to a preference for humans.
A new study uses artificial intelligence technologies to analyze natural language patterns to discern degrees of loneliness in older adults. The study found that NLP-based tools could help reconcile the differences between research assessments and an individual's subjective experience of loneliness, improving how clinicians and familie...
A joint laboratory is being developed by George Mason University and Dartmouth researchers to extract knowledge bases from technical documents. The U.S. Department of the Navy has allocated $529,194 for this project, which aims to create a distributed computing system to achieve large-scale document generation.
Researchers developed a recipe for creating ideal hybrid memristive-CMOS neuromorphic computing systems, exploiting the advantages of low-precision, noisy, and variable neurons. This work aims to enable compact and efficient real-time processing for applications such as bio-signal processing and brain-machine interfaces.
Regenstrief Institute's Eneida Mendonca and Umberto Tachinardi will share their expertise on artificial intelligence and data infrastructure at a Brazilian workshop. They discuss AI in natural language processing and building data infrastructure for precision medicine.
Researchers at Carnegie Mellon University have created a computer model that can translate text describing physical movements into simple computer-generated animations. The Joint Language-to-Pose model enables sentences and physical motions to be jointly embedded, learning how language is related to action, gestures, and movement.
Studies in linguistics and experimental psychology demonstrate that our minds can assign linguistic structure to non-linguistic content on the fly. Gestures and visual animations help researchers understand how inferences are generated, suggesting that all inferential types result from general, productive processes.
Researchers used AI to analyze electronic health records and identified 16 medical concepts predicting alcohol misuse in 78% of cases. This method could overcome staffing and patient barriers for screening and referral programs at trauma centers.
A team of researchers identified an algorithm called 'nearest-neighbor chaining' that describes how word senses accumulate over time. The study analyzed 1,000 years of English development and found that new word senses emerge through an efficient mechanism that expresses new ideas via a compact set of words.
A new KIT Motion-Language Dataset has been created to support the development of robot activities based on natural language input. The dataset, which includes over 4,000 motions and 6,200 annotations in natural language, aims to unify and standardize research linking human motion and natural language.
The University of Massachusetts Amherst has unveiled a powerful new GPU cluster, housing 400 specialized graphics processing units, to accelerate deep learning research and attract top talent in the field. The cluster is expected to significantly boost the university's capabilities in artificial intelligence, computer vision, and natur...
A new analysis system developed by Carnegie Mellon University found that nearly half of popular free apps lack a privacy policy, despite processing personally identifiable information. The system also identified inconsistencies in the policies of those that do exist.
Researchers developed a new AI system that uses web search to extract data from plain text, outperforming traditional machine learning methods by up to 10%. The system learns to generate search queries and gauge relevance, then extracts relevant data from online texts.
A study by MIT linguistics professors and Carleton University researchers found that children between 3-6 years old use a similar logical approach to adults when evaluating ambiguous sentences, but fail to consider the implications of substituting 'and' for 'or',
A researcher from Michigan State University is using a fellowship to study the role of sound in modern South Asian poetry, exploring rhyme, stress and patterns of repetition. The project aims to develop best practices for encoding linguistic data and preserve cultural forms in the region.
A study by Hanna Marno and colleagues found that languages with SVO structures prefer simplicity due to cognitive load constraints. In experiments where participants were not required to invent their own gestures, a preference for the SVO form emerged in both Italian-speaking and Farsi-speaking groups.
Natural language processing can help track quality of care at individual provider, service line, and health-care systems levels. The Regenstrief Institute's Center for Biomedical Informatics is utilizing natural language processing to improve care and provide information to those who need it.
A Regenstrief study found that natural language processing can accurately track colonoscopy quality, outperforming human review in accuracy and cost. The system analyzed over 90,000 colonoscopies and achieved 95% accuracy in detecting adenomas.
Researchers aim to enable scientists to ask a computer a question and receive an answer in an easy-to-understand picture. The project, funded by the National Science Foundation, seeks to develop an interactive computer system that can produce meaningful visualizations based on natural language requests.
A new $3.75 million project aims to teach computers to read and evaluate lengthy privacy policies using crowdsourcing, machine learning and natural language processing. The project will create user-friendly digests that highlight policy elements most relevant to users.
Researchers at MIT create systems that can translate natural-language descriptions into working code in a few narrow cases. The method involves mapping regular expressions to equivalent natural-language expressions, enabling programmers to specify complex searches using simple language.
Researchers found that men use heavy terms like 'hammered' for moderate intoxication, while women use lighter terms like 'buzzed.' These findings suggest tailored measures for prevention and intervention could be effective. The study's results have implications for how we perceive and communicate intoxicated states, particularly among ...
A Georgia Tech study identifies specific words and phrases that predict whether workplace emails are sent to someone higher or lower in the corporate hierarchy. Certain phrases consistently signal a power relationship between two people, with an estimated probability of chance occurring less than 1 in 1,000.
A team of MIT cognitive scientists found that ambiguity in language allows for the reuse of short, efficient sounds that listeners can easily disambiguate with context. This theory contradicts previous arguments that ambiguity is a problem for communication, suggesting it's actually a key to making language more efficient.
Linguistic researcher Marianne Pouplier aims to clarify why certain sound patterns are rarely encountered in languages, proposing an empirical approach using ultrasound imaging and perceptual tests.
A natural language processing-based approach outperformed traditional administrative data codes in detecting postoperative surgical complications. The study found the new method correctly identified 82% of acute renal failure cases, while also being highly specific for diagnoses like pneumonia and sepsis.
Research shows that certain types of metaphors are difficult to convey in sign language due to the iconicity of a form constraining possible metaphorical extensions. Sign languages exhibit rich metaphorical use, but clashes between form and meaning components can render some metaphors impossible.
Psychologist Neal J. Roese and computer scientist Eyal Amir predict that by 2060, androids will be able to mimic human-like voices, facial expressions, and emotions, but still struggle with understanding natural language and forming conclusions from visual input.
A USC Information Sciences Institute system identifies alpha chatterers in online communities, enabling the extraction of statistically correct answers. The method integrates speech act analysis, lexical similarity, and poster trustworthiness to create links for interpretation.
A new algorithm called Automatic Distillation of Structure (ADIOS) has been developed to learn complex patterns in raw texts, including transcripts of parents' speech directed at young children. The algorithm can generate grammatical new sentences and prove useful in fields like bioinformatics.