COVID rekindled an appreciation of nature for many

An opportunity to rediscover why the great outdoors are so great in the first place

Peer-Reviewed Publication

UNIVERSITY OF CONNECTICUT

The pandemic has impacted our lives in a multitude of ways, many of which will no doubt be felt for years to come. While many of those effects are clearly negative, UConn researchers have identified at least one positive impact – our perception of natural spaces changed. The findings are published in Nature Scientific Reports.

As people flocked to outdoor spaces for recreation in the spring of 2020, Sohyun Park, assistant professor in UConn’s College of Agriculture, Health and Natural Resources Department of Plant Science and Landscape Architecture, noticed some interesting trends: more people were on the trails, and many of those people had traveled from far away to enjoy nature.

Park was also part of the team for the Connecticut Trail Census and co-wrote a paper about the trends.

Sohyun Park of the Department of Plant Science and Landscape Architecture in her office in the W.B. Young Building. Mar. 8, 2022. (Jason Sheldon/UConn Photo)

“What’s interesting was rural trail use increased compared to urban trails,” Park says. “I wanted to try to find out how people were changing their mindset or their attitudes or perceptions.”

To do this, Park and co-authors Seungman Kim and Jaehoon Lee of Texas Tech University, and Biyoung Heo of James Corner Field Operations, looked to social media data and machine learning techniques to help make sense of the vast quantities of information on Twitter and try to find a pattern among those data, says Park. They utilized Twitter’s Application Programming Interface (API), which allows researchers to obtain aggregated data from Twitter posts.

“We wanted to know from the people who have been in greenspaces, what were they doing and what they were writing about on Twitter. We utilized very innovative and advanced levels of machine learning methods,” says Park. “The machine categorized the keywords and classified them into several human-recognizable groups. One group was nature related. The second group is all traditional park-related activities and the third one is obviously the COVID-related one, so mask wearing and social distancing, and things like that.”

A newfound appreciation for nature

Though what people were doing in the parks did not change significantly, the researchers noticed some significant changes in keyword usage between pre-pandemic and pandemic Twitter, with people frequently referencing nature, and their experiences within it.

“Users began to put the word ‘nature’ and nature-related activities or nature-related pictures on their Twitter,” Park says. “It was very interesting because, in the past, there were not many keywords like that, but people used keywords like ‘playing’, ‘walking the dog’, ‘baseball’, and other traditional, active park activities that they were enjoying while they were there in the parks. If you look at post- 2020 Twitter, you can also see keywords thanking God or showing appreciation for nature, describing the birds singing or water sounds.”

What the researchers found was people seemed to have realized a newfound appreciation for nature and greenspaces, especially true for those in urban settings. As a landscape architect, Park was interested to see if there was something more to the findings and perhaps if there might be some design implications for the results.

Common features in modern parks include benches, walking paths, and sports fields, for instance. However, Park says the results of the paper seem to suggest that natural greenspaces appeal to people differently.

“We might want to go back to the origin of public parks like those Olmsted designed,” Park says. “In his time in 19th century, there was a lot of hustle and bustle in the city, and they wanted to have space for people to find respite and peace. We might want to go back to that era living with a 21st century health crisis and try to rethink about the design principles.”

Park says to imagine Manhattan’s Central Park, which hosts ponds, wooded areas, and meadows. Most of those natural features were introduced artificially and were not there in the first place. Compare this with contemporary parks:

“Modern parks may be well managed, maintained, and manicured, everything is clean and tidy,” Park says. “There are some seating areas, paved surfaces, and structures where you can play something with your friends and family members, but not really in a naturalistic style. People can feel that in public spaces.”

The outdoors as essential resource for overall well-being

Greenspaces impact mental, physical, and spiritual health, and Park reasons that these natural elements might be essential in public spaces, particularly for those who have less access to the public parks, or marginalized communities that don’t have any green areas at their residences.

“I’m arguing that parks are not only recreational spaces; greenspaces and parks serve as essential amenities for all including those with low incomes or disabilities, and the elderly,” Park says. “Parks need inclusive planning approaches that might be added to the current principles for park development.”

Park explains that many people are involved in the planning, design, and management of parks and greenspaces. The tricky part is that the more naturalistic, garden concept for public parks may require more planning and maintenance, and therefore these design features rely more heavily on resources and budgets.

“In the long term, I think that will be the direction that we need to go and now officials and park managers need to work together with those who are living nearby so that we can have some kind of co-managing type of approaches to the future,” Park says.

This study highlights the importance of those design features and their roles in our emotional and spiritual well-being, and Park says it is important for the public to advocate for our greenspaces. Research like this can inform decision-making.

“It is important for the public and decision-makers to understand that ultimately, we need to have a budget to have more natural features and nature-oriented programs in the park.  We all need to be more active in terms of the things that towns are doing. That can start with joining your town’s Conservation Commission or attending monthly meetings. Participate and make your voice heard. That makes a huge change and can impact big decisions. Sometimes these decision makers are really grounded by how the stakeholders are feeling so giving some input and feedback on the public decision-making should be the first step. It is empowering and more people need to be involved in public planning.”

---

JOURNAL

Scientific Reports

DOI

10.1038/s41598-022-17077-3 

METHOD OF RESEARCH

Data/statistical analysis

SUBJECT OF RESEARCH

People

ARTICLE TITLE

Evolving norms: social media data analysis on parks and greenspaces perception changes before and after the COVID 19 pandemic using a machine learning approach

Did primitive cetaceans feed like marine reptiles?

Research carried out at the University of Liège (Belgium) has studied the morphological similarities between the first cetaceans and ancient marine reptiles from the dinosaur era, the mosasaurs

Peer-Reviewed Publication

UNIVERSITY OF LIEGE

 

IMAGE: ANALYZES OF THE DIFFERENCES BETWEEN SKULLS. view more 

CREDIT: @ULIÈGE/EDDYLAB

Did the first ancestors of whales pick up where the mosasaurs left off 66 million years ago, after the extinction of all the large predatory marine reptiles? A study conducted by Rebecca Bennion, a PhD student at the EDDyLab of the University of Liège (Belgium), has looked into the possible convergences in morphology and behaviour that may exist between these two groups of large marine predatory animals. This research has been published in the journal Paleobiology.

Many of us are familiar with modern whales and dolphins. However, the extinct ancestors of these modern marine mammals bear an uncanny resemblance to earlier forms of marine life, notably the mosasaurs, a completely extinct group closely related to snakes and lizards. "Superficial similarities have long been noted, but the idea that these two groups might be functionally similar has never been rigorously tested," explains Rebecca Bennion, a doctoral student at ULiège's EDDyLab and first author of the study.

The research, which has just been published in the journal Paleobiology and was carried out by an international team of scientists based in Europe, the USA and New Zealand, investigated the potential for convergent evolution of skull morphology between ancestral cetaceans and mosasaurs. To do this, a range of functional and biomechanical features were recorded from high-resolution three-dimensional (3D) scans of skulls from both groups. "Our laboratory has assembled a vast library of 3D scans of fossils, allowing us to explore in detail questions on large-scale evolution," explains Prof. Valentin Fischer, palaeontologist and director of the EDDy Lab at ULiège.

While cetaceans and mosasaurs initially had quite different ecological characteristics, this study found that several species had nevertheless acquired quite similar morphology and thus showed evolutionary convergence. "This convergence between early cetaceans and mosasaurs tells us more about the physical characteristics necessary for large marine predators to evolve optimally," says EDDyLab researcher Dr Jamie MacLaren, "Many members of these groups become very similar in their ecological characteristics, suggesting similar selective pressures on these animals despite being separated by tens of millions of years. Nevertheless, important differences remain between the two groups despite these examples of convergence. "Our results show what is called 'incomplete convergence', with differences remaining due to the mammalian or reptilian origin of each group," continues Rebecca Bennion.

Modern cetaceans are indeed a group that remains scientifically very interesting to study, it remains to be seen how diverse their morphology and ecology is compared to other fossil marine animals. This research is just the tip of the iceberg; further research into the convergence of marine animals through the fossil record will help us understand the constraints that evolution places on aquatic organisms and how they overcome them.

---

JOURNAL

Paleobiology

DOI

10.1017/pab.2022.27 

METHOD OF RESEARCH

Observational study

SUBJECT OF RESEARCH

Animals

ARTICLE TITLE

Convergence and constraint in the cranial evolution of mosasaurid reptiles and early cetaceans

http://prehistoricpulp.blogspot.com/2007/09/ice-hunt-by-james-rollins-2003.html

Ice Hunt by James Rollins (2003) - Prehistoric Pulp 

Sep 5, 2007 ... In this case, the plot hinges around the survival of the famous "walking whale" Ambulocetus. Ice Hunt starts when an experimental U.S. Navy ...

Hidden acid-base clusters drive rapid formation of atmospheric ultrafine particles

Peer-Reviewed Publication

SCIENCE CHINA PRESS

 

IMAGE: ON SUNNY DAYS IN MEGACITIES SUCH AS BEIJING, GASEOUS ACID AND BASE POLLUTANTS FORM A LARGE AMOUNT OF NEW ULTRAFINE PARTICLES. THE CLUSTERING BETWEEN ONE ACID AND ONE BASE IS THE RATE-LIMITING STEP FOR NEW PARTICLE FORMATION. view more 

CREDIT: ©SCIENCE CHINA PRESS

A joint research team led by Dr. Jingkun Jiang from Tsinghua University and Dr. Markku Kulmala from the University of Helsinki has reported an efficient mechanism for gaseous sulfuric acid and bases to form atmospheric ultrafine particles. The findings explain the rapid formation of secondary ultrafine particles, which could further influence air quality and the climate.

The team found that acid-base reactions are the major driving forces for gaseous precursors to overcome surface tension and form ultrafine particles, and the key mechanism is the formation of hidden acid-base heterodimers. This hidden mechanism explains the high particle formation rate in Chinese megacities.

Their findings were published in National Science Review.

“There are hundreds of thousands of ultrafine particles in per cubic centimeter air in Chinese megacities, and a new particle formation event on a sunny noon can readily elevate their concentration by one order of magnitude within several hours,” Jiang says.

To explain how new particles can be so efficiently converted from gaseous precursors, Jiang and Kulmala, together with Dr. Runlong Cai, are determined to seek the key mechanism for rapid new particle formation. They have known that sulfuric acid is a primary precursor, whereas the challenge is to find the key bases among many candidates. “Urban air is a complex cocktail of chemicals with poorly understood interactions and feedbacks”, Kulmala commented.

The researchers observed highly abundant molecular clusters containing sulfuric acid during new particle formation in Beijing and Shanghai. Some of the measured clusters contain sulfuric acid and amine molecules. These provide strong evidence for amine participation in the formation of stable sulfuric acid clusters, which increases the conversion rate from gaseous sulfuric acid to new particles.

“It is intriguing that we measured fewer bases than acids in a cluster. There must be some key information hidden behind the measured signals,” Cai says. It was previously proposed that the clustering between a base molecule and a sulfuric acid homodimer is the key mechanism for new particle formation, as there were no base molecules in the measured clusters containing one sulfuric acid molecule. The research team, however, found that this was a measurement artifact.

Combining long-term measurements and theory based on quantum chemistry and cluster kinetics, they found that the formation of hidden acid-base heterodimers is the key mechanism. This mechanism is far more efficient than the previously proposed mechanism with acid-acid homodimers, ensuring the rapid formation of sulfuric acid clusters and new particles.

The hidden heterodimers solve the puzzle of why new particles could be frequently formed against a high background particle loading in megacities. The hidden acid-base heterodimers with a considerable fraction in the measured sulfuric acid signals can effectively cluster with each other. This ensures a high particle formation rate approaching the theoretical maximum even at a low ambient amine concentration. The hidden heterodimers also explain the temperature dependency of new particle formation in Beijing and Shanghai. “Atmospheric measurements are often perturbed by many factors. I did not expect such an amazing consistency between the measurements and the new theory,” Cai says.

The team also sought the hidden base molecules using thermodynamic and kinetic analyses. Among the measured gaseous molecules, strong amines such as dimethylamine serve as the key bases in acid-base heterodimers, whereas the highly abundant ammonia and other weak bases are more likely to be involved in the subsequent cluster growth process.

###

See the article:

The missing base molecules in atmospheric acid-base nucleation

https://doi.org/10.1093/nsr/nwac137

---

JOURNAL

National Science Review

DOI

10.1093/nsr/nwac137 

Spatiotemporal variation of mortality burden attributable to heatwaves in China, 1979-2020

Peer-Reviewed Publication

SCIENCE CHINA PRESS

 NEWS RELEASE 2-SEP-2022

 

IMAGE: HEALTH RISKS OF HEATWAVES WERE CHARACTERIZED BY RAPID GROWTH, NONLINEAR TEMPORAL EVOLUTION AND EXTREMITY. NATIONALLY, THE NUMBER OF ANNUAL ATTRIBUTABLE DEATHS WAS AN AVERAGE OF 3,679 IN 1980S BUT INCREASED TO 15,500 IN 2010S. TAKING A 5-YEAR MOVING AVERAGE, IT TOOK 2.8 YEARS FOR EVERY INCREASE OF 1,000 ANNUAL HEATWAVE-RELATED DEATHS FROM 1980S TO 2000S, BUT JUST ONE YEAR FOR THE SAME INCREASE TO OCCUR IN 2010S. ALSO, THE ATTRIBUTABLE DEATHS REACHED A HIGHEST RECORD OF 26,486 IN 2017, FOLLOWED BY THE SECOND 21,219 IN 2019 AND THE THIRD 20,431 IN 2013. REGIONALLY, EAST AND CENTRAL CHINA HAD THE LARGEST NUMBER OF ATTRIBUTABLE DEATHS IN GENERAL, ACCOUNTING FOR MORE THAN 50% OF DEATHS NATIONWIDE. AMONG THE PROVINCES, DEATHS ASCRIBED TO HEATWAVES WERE HIGHEST IN SHANDONG, FOLLOWED BY HENAN, HEBEI AND JIANGSU. A-D, THE NUMBER OF ATTRIBUTABLE DEATHS TO HEATWAVES IN GRID DURING THE PAST FOUR DECADES. E, ATTRIBUTABLE DEATHS TO HEATWAVES IN CHINA FROM 1979 TO 2020. THE SOLID LINE SHOWS THE ESTIMATED NUMBER OF ATTRIBUTABLE DEATHS TO HEATWAVES; GREY AREAS SHOW THE 95% CONFIDENCE INTERVALS; THE HORIZONTAL DASHED LINE SHOWS THE AVERAGE ANNUAL DEATHS IN 1980S, 1990S, 2000S, 2010S. F, CHANGES IN ATTRIBUTABLE DEATHS TO HEATWAVES IN 2010S RELATIVE TO THE BASELINE PERIOD OF 1980-2009. AD, ATTRIBUTABLE DEATHS TO HEATWAVES. THE MAP NO. IS GS(2019)1673. ART BY HUIQI CHEN. view more 

CREDIT: ©SCIENCE CHINA PRESS

This study is led by Dr. Cunrui Huang (Vanke School of Public Health, Tsinghua University). Heatwaves impose heavy disease burden by increasing the risk of mortality and morbidity, which has been exacerbated worldwide under climate change. "In China, evidence documenting the impact of heatwaves on the number of attributable deaths, spatiotemporal variations and their driving factors is still limited, hindering the understanding of dangerous heatwaves. " Huang says.

Huang, together with his group member Chen and meteorological expert Zhao, sought to identify what was the spatial and temporal trends of heatwave-attributable deaths in China over the past four decades. The team performed event-based attributable loss estimation to quantify the gridded attributable deaths.

The team found that health risks of climate change were characterized by rapid growth, nonlinear evolution and extremity. The attributable deaths to heatwaves in China have increased dramatically by four times in the past four decades, with the rising trends becoming more apparent in the recent decade but some fluctuations among individual years. Regionally, east and central China had the largest number of attributable deaths in general, accounting for more than 50% of deaths nationwide. Among the provinces, deaths ascribed to heatwaves were highest in Shandong, followed by Henan, Hebei and Jiangsu.

The researchers also decomposed the driving factors to changes in attributable deaths. The increase in attributable deaths to heatwaves in China over time was primarily due to increased heatwave exposure, followed by population growth, population aging and the mounting baseline mortality. Notably, population aging has played an increasingly important role in attributable deaths over time. This work could serve important information for policy-makers to develop effective climate mitigation and adaptation measures in response to increasing heatwaves, especially for the most vulnerable elderly populations.

See the article:

Spatiotemporal Variation of Mortality Burden Attributable to Heatwaves in China, 1979-2020

https://doi.org/10.1016/j.scib.2022.05.006

---

JOURNAL

Science Bulletin

DOI

10.1016/j.scib.2022.05.006 

CAPTION

The increase in attributable deaths to heatwaves in China was primarily due to increased heatwave exposure, followed by population growth, population aging and the mounting baseline mortality. Notably, population aging has played an increasingly important role in attributable deaths over time, increased from 10.6% in the period between 1980s and 1990s to 15.8% afterward, reaching as high as 20.8% in the period between 2000s and 2010s. a, Decomposition of the contribution of driving factors in seven regions (left) and for the whole nation (right) from 1980s to 1990s. b, Decomposition of the contribution of driving factors in seven regions (left) and for the whole nation (right) from 1990s to 2000s. c, Decomposition of the contribution of driving factors in seven regions (left) and for the whole nation (right) from 2000s to 2010s. AD, attributable deaths to heatwaves. The Map NO. is GS(2019)1673. Art by Huiqi Chen.

CREDIT

©Science China Press

Mortality caused by heatwaves in China has increased since 1979

Peer-Reviewed Publication

INSTITUTE OF ATMOSPHERIC PHYSICS, CHINESE ACADEMY OF SCIENCES

 

IMAGE: HEATWAVES SWEEPS THROUGH NORTHERN CHINA. PHOTO TAKEN IN BEIJING ON JULY 20. view more 

CREDIT: LIANG ZHAO

Since the beginning of the summer in 2022, China has been sweltering under the worst heatwave in decades. A number of people in Zhejiang, Henan, Jiangsu, and Sichuan provinces were diagnosed with thermoplegia, the most severe form of heatstroke, and some even died of this disease.

In a warming world, the threat of heatwaves to human health is increasing. Researchers led by Liang Zhao from the Institute of Atmospheric Physics of the Chinese Academy of Sciences and Cunrui Huang from Tsinghua University investigated the spatial and temporal variation of heatwaves-related human deaths in China from 1979 to 2020.

The study was published in Science Bulletin.

They also explored the relative contributions of drivers such as heatwaves exposure, population growth, population aging, and baseline mortality to changes in attributable mortality.

"We found that the number of deaths caused by heatwaves in China has increased rapidly since 1979, from 3,679 persons per year in the 1980s to 15,500 persons per year in the 2010s," said Zhao. "We also found strong spatial heterogeneity across the country, with more human deaths in East and Central China."

Then what are the main drivers of the large increase in heatwaves-related deaths in China over the past four decades?

"The main drivers are the rapid increase in the frequency of heatwaves, followed by population growth, population aging, and rising baseline mortality. From the 2000s to the 2010s, these four factors accounted for 40.6%, 22.4%, 20.8%, and 16.2% of the change in attributed human deaths, respectively," said Huang.

The research team hopes this study's accurate assessment of the number of heatwaves-related deaths in China and the contribution of their different drivers can help policy makers to fully understand human health hazards of heatwaves and to develop response policies to reduce health losses from increased heatwaves exposure under climate change.

---

DOI

10.1016/j.scib.2022.05.006 

ARTICLE TITLE

Spatiotemporal variation of mortality burden attributable to heatwaves in China, 1979–2020

Dragons and brain evolution

A molecular atlas of an Australian dragon’s brain sheds new light on over 300 million years of brain evolution

Peer-Reviewed Publication

MAX-PLANCK-GESELLSCHAFT

 

IMAGE: THE AUSTRALIAN BEARDED DRAGON POGONA VITTICEPS. view more 

CREDIT: MAX PLANCK INSTITUTE FOR BRAIN RESEARCH / G. LAURENT

These days, dragons are keeping Game of Thrones fans on their toes. But they are also providing important insights into vertebrate brain evolution, as revealed by the work of Max Planck scientists on the brain of the Australian bearded dragon Pogona vitticeps. Vertebrate evolution took a major turn 320 million years ago when early tetrapods (animals with four limbs) transitioned from water to land, eventually giving rise to three major clades: the reptiles, the birds (an offshoot of the reptilian tree) and the mammals. Because of common ancestry, the brains of all tetrapods share a similar basal architecture established during early development. Yet, how variations on this common “Bauplan” contributed to clade-specific attributes remains unclear. Neuroscientists at the Max Planck Institute for Brain Research in Frankfurt tackled this question by generating a molecular atlas of the dragon brain and comparing it with one from mice. Their findings suggest that, contrary to popular belief that a mammalian brain consists of an ancient “reptilian” brain supplemented with new mammalian features, both reptilian and mammalian brains evolved their own clade-specific neuron types and circuits, from a common ancestral set.

“Neurons are the most diverse cell types in the body. Their evolutionary diversification reflects alterations in the developmental processes that produce them and may drive changes in the neural circuits they belong to”, says Prof. Gilles Laurent, Director at the Max Planck Institute for Brain Research who led the new study published in Science. “For example, distinct brain areas do not work in isolation, suggesting that the evolution of interconnected regions, such as the thalamus and cerebral cortex, might in some way be correlated. Also, a brain area in reptiles and mammals that derived from a common ancestral structure might have evolved in such a way that it remains ancestral in one clade today, while it is “modern” in the other. Conversely, it could be that both clades now contain a mix of common (ancient) and specific (novel) neuron types. These are the sorts of questions that our experiments tried to address”, Laurent adds.

While traditional approaches to compare developmental regions and projections in the brain do not have the necessary resolution to reveal these similarities and differences, Laurent and his team took a cellular transcriptomic approach. Using a technique called single-cell RNA sequencing that detects a large fraction of the RNA molecules (transcriptomes) present in single cells, the scientists generated a cell-type atlas of the brain of the Australian bearded dragon Pogona vitticeps and compared it to existing mouse brain datasets.

CAPTION

Reptiles and mammals are evolutionary separated by over 300 million years. Max Planck scientists generated a cell type atlas from the brain of a lizard. Computationally integration of this data with mouse transcriptomics revealed that multiple brain areas contain mixtures of similar and divergent neurons, suggesting ubiquitous neuron diversification in these brain regions.

CREDIT

Max Planck Institute for Brain Research / G. Laurent; Hain et al. Science 2022 DOI: 10.1126/science.abp8202

Transcriptomic comparisons reveal shared classes of neuron types

“We profiled over 280,000 cells from the brain of Pogona and identified 233 distinct types of neurons”, explains David Hain, graduate student in the Laurent Lab and co-first author of the study. “Computational integration of our data with mouse data revealed that these neurons can be grouped transcriptomically in common families, that probably represent ancestral neuron types”, says Hain. In addition, he found that that most areas of the brain contain a mix of common (ancient) and specific (novel) neuron types, as shown in the figure below.

Graduate student Tatiana Gallego-Flores used histological techniques to map these cell types throughout the dragon brain and observed (among other) that neurons in the thalamus could be grouped in two transcriptomic and anatomical domains, defined by their connectivity to other regions of the brain. Because these connected regions have had different fates in mammals and in reptiles, one of these regions being highly divergent, comparing the thalamic transcriptomes of these two domains proved to be very interesting. Indeed, it revealed that transcriptomic divergence matched that of the target regions.

“This suggests that neuronal transcriptomic identity somewhat reflects, at least in part, the long-range connectivity of a region to its targets. Since we do not have the brains of ancient vertebrates, reconstructing the evolution of the brain over the past half billion years will require connecting together very complex molecular, developmental, anatomical and functional data in a way that is self-consistent. We live in very exciting times, because this is becoming possible”, concludes Laurent.

---

JOURNAL

Science

DOI

10.1126/science.abp8202 

METHOD OF RESEARCH

Meta-analysis

SUBJECT OF RESEARCH

Animals

ARTICLE TITLE

Molecular diversity and evolution of neuron types in the amniote brain.

From wound healing to regeneration

Heidelberg scientists demonstrate how injuries are converted to regeneration signals at the molecular level

Peer-Reviewed Publication

HEIDELBERG UNIVERSITY

The phenomenon of regeneration was discovered over 200 years ago in the freshwater polyp Hydra. Until now, however, it was largely unclear how the orderly regeneration of lost tissues or organs is activated after injury. In its investigations of Hydra, an interdisciplinary research team at Heidelberg University was able to show how wound healing signals released upon injury are converted into specific signals of pattern formation and cell differentiation. Essential components are the mitogen-activated protein kinases (MAPK) and the Wnt signalling pathway – molecular mechanisms that have remained relatively unchanged throughout evolution.

The ability to regenerate varies widely in animals. Most mammals and vertebrates have only limited regeneration capacity, while basal and simple animals that emerged early in evolution, like cnidarians and planarians, can regenerate their whole body. In all cases, the process of regeneration begins with wound healing. The cells at the site of injury proliferate and form an undifferentiated mass – a blastema – from which the missing structures are re-patterned. This activates genetic processes that also control embryonic development. To determine the molecular mechanisms involved, the research team led by Prof. Dr Thomas W. Holstein studied the freshwater polyp Hydra to understand the basic features of this activation of regeneration.

The core of their investigations is the doctoral thesis of Anja Tursch. She repeated the key experiment of Geneva naturalist Abraham Trembley (1710 to 1784) which led him to discover the regeneration phenomenon. The Hydra polyp is bisected, prompting the upper half to regenerate a new “head” and the lower half a new “foot” – hence totally different body parts can grow from the exact same tissue at the cut surface in the middle. Building on their previous work on Hydra regeneration, the researchers at the Centre for Organismal Studies (COS) of Heidelberg University have now shown how this is possible.

Regardless of where it occurs, any damage triggers nonspecific signals for an injury response, i.e. wound healing, via calcium ions and the production of reactive oxygen species. The signals are transmitted intracellularly by three mitogen-activated protein kinases – p38, JNKs, and ERK. Activation of these three molecules is required for both head and foot regeneration. Wnt signalling pathways are then activated that are important during embryonic development for the formation of rudimentary organs and the body axis. The generic signals of wound healing are thus transferred into position-specific signals of patterning and cell differentiation for regeneration.

“Our experiments show that the Wnt signalling pathway is a main component of the initially general injury response and, depending on signal strength, directs the tissue toward head or foot development,” explains Prof. Holstein. This is why, in the case of MAPK inhibition, the otherwise absent regeneration can be induced by artificially generated, recombinant Wnt proteins. “It was also surprising that in middle body parts that had both head and foot removed, heads can be induced at both ends in this way,” adds Dr Suat Özbek, a member of Prof. Holstein’s “Molecular Evolution and Genomics” research group at the COS.

Wnt/β-catenin, one piece of the Wnt signalling pathway, was already known to encode positional information for new head structure formation. In collaboration with mathematicians led by Prof. Dr Anna Marciniak-Czochra, the research team of Prof. Holstein and Dr Özbek developed a model that shows how basal positional information in the tissue transforms the initially undifferentiated injury response into a differential patterning process via the Wnt signalling pathway. “Because MAPKs and Wnts are highly evolutionarily conserved, this mechanism is likely deeply embedded in our genome, which is important for regenerative processes in vertebrates and mammals as well,” stresses Thomas Holstein.

The research was done under the auspices of the “Mechanisms and Functions of Wnt Signaling” Collaborative Research Centre (CRC 1324) funded by the German Research Foundation. The results were published in the “Proceedings of the National Academy of Sciences” (PNAS).

---

JOURNAL

Proceedings of the National Academy of Sciences

DOI

10.1073/pnas.2204122119 

ARTICLE TITLE

Injury-induced MAPK activation triggers body axis formation in Hydra by default Wnt signaling

Researchers propose new and more effective model for automatic speech recognition

Consumer-focused new framework improves automatic speech recognition by including context information, an improved solution for video conferencing and live interviews

Peer-Reviewed Publication

TSINGHUA UNIVERSITY PRESS

 

IMAGE: THE PHONETIC-SEMANTIC PRE-TRAINING (PSP) FRAMEWORK USES “NOISE-AWARE CURRICULUM” LEARNING TO EFFECTIVELY IMPROVE THE PERFORMANCE OF ASR IN NOISY ENVIRONMENTS. INTEGRATING WARM-UP, SELF-SUPERVISED LEARNING, AND FINE-TUNING. view more 

CREDIT: CAAI ARTIFICIAL INTELLIGENCE RESEARCH, TSINGHUA UNIVERSITY PRESS

Popular voice assistants like Siri and Amazon Alexa have introduced automatic speech recognition (ASR) to the wider public. Though decades in the making, ASR models struggle with consistency and reliability, especially in noisy environments. Chinese researchers developed a framework that effectively improves the performance of ASR for the chaos of everyday acoustic environments.

 

Researchers from the Hong Kong University of Science and Technology and WeBank proposed a new framework - phonetic-semantic pre-training (PSP) and demonstrated the robustness of their new model against synthetic highly noisy speech datasets.

 

Their study was published in CAAI Artificial Intelligence Research on Aug. 28.

 

“Robustness is a long-standing challenge for ASR,” said Xueyang Wu from the Hong Kong University of Science and Technology Department of Computer Science and Engineering. “We want to increase the robustness of the Chinese ASR system with a low cost.”

 

ASR uses machine-learning and other artificial intelligence techniques to automatically translate speech into text for uses like voice-activated systems and transcription software. But new consumer-focused applications increasingly call for voice recognition to work better — handle more languages and accents, and perform more reliably in real-life situations like video conferencing and live interviews.

 

Traditionally, training the acoustic and language models that comprise ASR requires large amounts of noise-specific data, which can be time- and cost-prohibitive.

 

The acoustic model (AM) turns words into a “phones,” which are sequences of basic sounds. The language model (LM) decodes phones into natural-language sentences, usually with a two-step process: a fast but relatively weak LM generates a set of sentence candidates, and a powerful but computationally expensive LM selects the best sentence from the candidates.

 

“Traditional learning models are not robust against noisy acoustic model outputs, especially for Chinese polyphonic words with identical pronunciation,” Wu said. “If the first pass of the learning model decoding is incorrect, it is extremely hard for the second pass to make it up.”

 

The newly proposed framework PSP makes it easier to recover misclassified words. By pre-training a model that translates the AM outputs directly to sentence along with the full context information, researchers can help the LM efficiently recover from the noisy outputs of the AM.

 

The PSP framework allows the model to improve through a pre-training regime called noise-aware curriculum that gradually introduces new skills, starting easy and gradually moving into more complex tasks.

 

“The most crucial part of our proposed method, Noise-aware Curriculum Learning, simulates the mechanism of how human beings recognize a sentence from noisy speech,” Wu said.  

 

Warm-up is the first stage, where researchers pre-train a phone-to-word transducer on a clean phone sequence, which is translated from unlabeled text data only — to cut back on the annotation time. This stage “warms up” the model, initializing the basic parameters to map phone sequences to words.

 

In the second stage, self-supervised learning, the transducer learns from more complex data generated by self-supervised training techniques and functions. Finally, the resultant phone-to-word transducer is fine-tuned with real-world speech data.

 

The researchers experimentally demonstrated the effectiveness of their framework on two real- life datasets collected from industrial scenarios and synthetic noise. Results showed that the PSP framework effectively improves the traditional ASR pipeline, reducing the relative character error rates by 28.63% for the first dataset and 26.38% for the second.

 

In next steps, researchers will investigate more effective PSP pre-training methods with larger unpaired datasets, seeking to maximize the effectiveness of pretraining for noise-robust LM.

 

Other contributors include Rongzhong Lian, Di Jiang, Yuanfeng Song, Weiwei Zhao, and Qian Xu, and Qiang Yang from WeBank Co. Ltd. Qian Xu and Qiang Yang are also affiliated with The Hong Kong University of Science and Technology.

 

CAAI Artificial Intelligence Research is a new journal jointly sponsored by Chinese Association for Artificial Intelligence (CAAI) and Tsinghua University. This is the first paper published in the journal.

 

About CAAI Artificial Intelligence Research 

 

CAAI Artificial Intelligence Research is a peer-reviewed journal jointly sponsored by Chinese Association for Artificial Intelligence (CAAI) and Tsinghua University. The journal aims to reflect the state-of-the-art achievement in the field of artificial intelligence and its application, including knowledge intelligence, perceptual intelligence, machine learning, behavioral intelligence, brain and cognition, and AI chips and applications, etc. Original research and review articles from all over the world are welcome for rigorous peer-review and professional publishing support.

 

About SciOpen 

 

SciOpen is a professional open access resource for discovery of scientific and technical content published by the Tsinghua University Press and its publishing partners, providing the scholarly publishing community with innovative technology and market-leading capabilities. SciOpen provides end-to-end services across manuscript submission, peer review, content hosting, analytics, and identity management and expert advice to ensure each journal’s development by offering a range of options across all functions as Journal Layout, Production Services, Editorial Services, Marketing and Promotions, Online Functionality, etc. By digitalizing the publishing process, SciOpen widens the reach, deepens the impact, and accelerates the exchange of ideas.

 

---

DOI

10.26599/AIR.2022.9150001 

ARTICLE TITLE

A Phonetic-Semantic Pre-Training Model for Robust Speech Recognition