Researchers from the Instituto Valenciano de Investigaciones Agrarias (IVIA) in Moncada, Valencia, led by corresponding author Isabel Rodríguez-Carretero, spent two consecutive growing seasons testing composts in a commercial orchard of adult Tango mandarins grafted onto a hybrid rootstock and irrigated by traditional surface flooding. The team compared two industrially produced composts: one made from pruning residues and sewage sludge at a 1:3 fresh-weight ratio, the facility’s usual recipe, and a second made from rice straw and sewage sludge at a 1:8 ratio, using the rice straw as a locally available bulking agent. Composts were surface-applied each June within the tree canopy projection at two rates, 10 and 20 tonnes per hectare, alongside unfertilised-with-compost control plots, in a randomized complete block design with three replicates.

The chemical characterisation of the two composts revealed a striking difference. The rice straw–sewage sludge compost contained significantly higher total nitrogen, largely as organic nitrogen, and a lower carbon-to-nitrogen ratio of 10.3 compared with 13.8 for the pruning-residue version. Most notably, its phosphorus content, expressed as P2O5, was more than double that of the conventional compost and exceeded average values typically reported for livestock manures. The researchers suggest this makes the rice straw compost a promising phosphorus-rich organic fertiliser, a significant finding at a time when phosphate rock reserves are finite and Europe depends heavily on imported fertiliser raw materials.

Both composts met Spanish legal requirements for fertilising products, including minimum organic matter thresholds and maximum carbon-to-nitrogen ratios. One caveat emerged: the rice straw compost slightly exceeded the 500 milligrams per kilogram zinc limit for Class B classification under the older RD 506/2013 regulation, though it complied with the stricter thresholds of the more recent RD 1051/2022, which governs sustainable soil nutrition. The compost’s slightly acidic pH of 6.71, unusual for sewage sludge-based products, may actually be an advantage in the alkaline, carbonate-rich soils of eastern Spain, where it can enhance the solubility and availability of nutrients that are otherwise locked up by calcium carbonates.

In the field, the effects on soil chemistry unfolded over time. Soil organic matter and organic nitrogen rose significantly in the first season at both application rates, and in the second season at the double dose. Available phosphorus increased significantly in both seasons regardless of dose, pushing soil levels from normal or high into the high to very high categories used in citrus nutritional diagnostics. Available potassium rose significantly only at the double rate, by 17 percent in the first season and 29 percent in the second. Interestingly, exchangeable sodium decreased in composted plots in the second season, while soil pH dropped significantly under the double dose, a cumulative effect the authors attribute to repeated organic matter inputs and the release of organic acids during decomposition.

Salinity was the one indicator demanding caution. Electrical conductivity rose by 21 percent in the first season and 105 percent in the second under the double application rate, reflecting the salt load carried by successive compost additions. The values nonetheless remained below both the 4 dS per metre threshold that defines a saline soil and the 1.7 dS per metre level at which citrus yields begin to decline. The authors recommend that long-term management might involve reduced annual rates or biennial applications, combined with irrigation practices that promote salt leaching through the improved soil permeability that organic matter confers.

Soil biology responded quickly but transiently. In the first season, microbial biomass carbon jumped 65 percent under the double compost dose, and dehydrogenase activity, a standard proxy for overall microbial metabolic activity, increased under both rates. By the second season, however, no significant differences among treatments remained. The researchers suggest the first-year compost, richer in readily oxidisable organic carbon, offered microorganisms a more accessible substrate, and note that biological indicators are highly sensitive to soil moisture, temperature, and sampling timing. They call for future studies with multiple sampling dates across crop phenological stages to disentangle these dynamics.

Heavy metal concentrations in the soil, including copper, zinc, nickel, lead, cadmium, and chromium, stayed within permissible limits under both older and current Spanish regulations throughout the trial, although the double compost dose raised total soil zinc by roughly 17 to 18 percent in both seasons. Foliar nutrient analysis told a similarly reassuring story: macronutrient concentrations remained within optimal ranges in all treatments, with the sole exception of potassium in the second season. Compost significantly increased leaf zinc only in the first season. Encouragingly, because soil and foliar phosphorus were already elevated after the first compost application, the team cut mineral phosphorus fertilisation by 10 percent in the second season, and leaf phosphorus stayed optimal, evidence that compost-derived phosphorus can partially replace mineral P inputs under comparable conditions.

Perhaps the most commercially significant result is what did not change. Yield, fruit weight, diameter, peel thickness, colour index, total soluble solids, titratable acidity, and maturity index showed no agronomically relevant differences between composted and control trees in either season. All fruit met EU marketing standards for mandarins, with diameters above 45 millimetres, juice content above 33 percent, and maturity indices exceeding 7.5. Because mineral fertilisation met crop requirements across all plots, the soil fertility gains from composting simply had no yield gap to fill. The authors caution that their findings come from a flood-irrigated orchard, a system still used on roughly 13 percent of Spanish citrus acreage, and that responses under modern drip irrigation may differ. Still, the study demonstrates that industrial-scale composting of rice straw and sewage sludge is a technically viable, regulation-compliant route to closing nutrient loops in Mediterranean agriculture, turning two disposal headaches into a soil-building resource while keeping the region’s mandarins just as sweet.

Subject of Research: Effects of rice straw–sewage sludge compost on soil fertility, tree nutrition, and fruit quality in Mediterranean citrus orchards

Article Title: Field application of rice straw–sewage sludge compost in Mediterranean citrus orchards: effects on soil properties, nutrient status and fruit quality

Article References: Rodríguez-Carretero, I., Canet, R., Pérez-Piqueres, A., & Quiñones, A. (2026). Field application of rice straw–sewage sludge compost in Mediterranean citrus orchards: effects on soil properties, nutrient status and fruit quality. SOIL, 12(2), 841-853. https://doi.org/10.5194/soil-12-841-2026

Image Credits: AI Generated

DOI: 10.5194/soil-12-841-2026

Keywords: compost, rice straw, sewage sludge, citrus, soil fertility, circular economy, phosphorus, soil organic matter, microbial biomass, Mediterranean agriculture, mandarin, organic fertiliser

The scientific leadership rests with Vicent Arbona Mengual of the Department of Biology, Biochemistry and Natural Sciences at Universitat Jaume I. His team also includes Miguel González Guzmán, coordinator of the GaMBiT research group – Genetic and Metabolic Integration of Biotic and Abiotic Interactions – together with Rosario Vidal and Beatriz Julián from the Institute of Advanced Materials (INAM-UJI). The wider consortium spans 11 partners from five countries: Spain, Italy, France, Germany and Sweden. Spanish participants include the Valencian Institute for Agricultural Research and the Galician University–Business Foundation. Italy contributes the Council for Agricultural Research and Economics (CREA) and UNITEC S.p.A., while France is represented by the French Agricultural Research Centre for International Development (CIRAD), the French National Research Institute for Agriculture, Food and Environment (INRAE) and Doriane SAS. Germany’s Computomics GmbH and CYBRES GmbH join the Swedish University of Agricultural Sciences to complete the partnership.

The project’s primary biological targets are sweet orange (Citrus sinensis) and mandarin (Citrus reticulata), two species that dominate global citrus production and consumption. The central objective is to generate genetically modified lines with greater resilience to the adverse conditions caused by climate change and its exacerbating effects on pests and diseases. Rising temperatures, altered rainfall patterns and shifting pest distributions are already straining citrus orchards across the Mediterranean and beyond, making the development of hardier varieties an increasingly urgent agricultural priority. By combining stress physiology expertise with advanced genetic engineering, the consortium aims to produce trees that can maintain productivity under conditions that would compromise conventional varieties.

Resilience is only half of the project’s ambition. CitrusAld also pursues the biofortification of citrus fruits to enhance their nutritional properties. Citrus fruits are widely regarded as functional foods that combine nutritional value with health-promoting properties, thanks to their antioxidant, anti-inflammatory and cardioprotective effects. The project seeks to increase naturally occurring beneficial compounds such as flavonoids and furanocoumarins, while also introducing a compound that citrus does not normally produce: resveratrol. This stilbene, best known as a constituent of grapes and cocoa, has been selected for its cardioprotective and anti-ageing properties. If successful, the work could transform an everyday fruit into an even richer source of bioactive molecules, blurring the line between conventional nutrition and functional food design.

Identifying which genes to modify in a genome as complex as citrus is a formidable challenge, and this is where artificial intelligence enters the project. The consortium will design and implement AI-based systems capable of analysing and integrating large volumes of biological data to pinpoint the most suitable target genes for each desired modification. Rather than relying on laborious trial-and-error approaches, the researchers intend to use computational pipelines that can sift through genomic, transcriptomic and metabolic datasets to prioritise candidate genes whose manipulation is most likely to yield stress tolerance or enhanced nutritional profiles. This data-driven gene discovery methodology represents one of the project’s key methodological innovations and reflects a broader trend of machine learning reshaping plant biotechnology.

Once target genes are identified, the actual genetic modifications will be carried out using New Genomic Techniques, or NGTs, which are now permitted in Europe under the newly updated regulatory framework. These techniques allow precise edits to a plant’s genome without introducing foreign DNA in the way older transgenic methods did, and their recent regulatory acceptance in the European Union has opened the door to projects of this kind. The consortium will also integrate novel materials capable of overcoming existing biological barriers that have historically limited the genetic efficiency of citrus species. Citrus is notoriously difficult to transform genetically, with recalcitrant tissues, long generation times and barriers to regeneration, so the development of new delivery materials could prove as consequential as the gene edits themselves.

In parallel with genome editing, the consortium will explore complementary strategies based on RNA interference technology. RNAi allows researchers to slow down or silence the expression of specific genes without permanently altering the genome, offering a reversible and highly targeted means of modulating plant traits. A central technical hurdle for RNAi in plants is delivering the interfering molecules into plant tissues, where cell walls and other barriers block conventional approaches. The CitrusAld team plans to use novel materials as delivery vehicles into plant tissues, an approach that draws on the materials science expertise within the consortium and could establish new protocols for transient gene silencing in woody perennial crops.

The funding mechanism behind the project is as distinctive as its science. The EIC Pathfinder Challenges programme, part of the European Innovation Council, supports highly innovative, high-risk scientific research with the potential to generate entirely new technologies in the future. It deliberately funds the early stages of research, when ideas are still being explored in the laboratory and there is not yet a clear commercial application. The programme’s objective is to establish the scientific and technological foundations for breakthroughs that, in the long term, can transform entire sectors or help address major societal challenges. For a crop that underpins the economies of Mediterranean regions and provides vitamin-rich food to billions of people, the societal stakes of climate-resilient citrus are considerable.

The four-year project will begin in October 2026 under proposal number 101306995, giving the consortium a defined window in which to deliver its gene discovery pipelines, transformation protocols and biofortified citrus lines. The breadth of the partnership – spanning academic plant science institutes, agricultural research organisations, materials specialists and companies – suggests a deliberate effort to cover the full pipeline from fundamental gene discovery to applied field-relevant varieties. The involvement of computational firms such as Computomics and CYBRES alongside plant science heavyweights like INRAE and CIRAD signals that the AI and data integration components are treated as core infrastructure rather than an afterthought.

For consumers and growers alike, the project’s promise is twofold: citrus trees better equipped to endure a warming, pest-pressured world, and fruit with an enhanced portfolio of health-promoting compounds. Whether engineered resilience and resveratrol-enriched oranges will reach commercial orchards remains a question for the years beyond the project’s horizon, but CitrusAld represents one of the most comprehensive European attempts to marry genomics, artificial intelligence, materials science and plant breeding in service of a single iconic crop. As climate change accelerates and the tools of biotechnology mature, the humble orange may become a showcase for how 21st-century science re-engineers the foods we have cultivated for millennia.

Subject of Research: Genetic modification of citrus crops for climate resilience and nutritional biofortification

Article Title: Universitat Jaume I leads European consortium to develop genetically modified citrus lines with greater resilience to climate change, pests and diseases

Article References: Universitat Jaume I leads European consortium to develop genetically modified citrus lines with greater resilience to climate change, pests and diseases. (n.d.). Original publication

Image Credits: AI Generated

DOI: Not provided

Keywords: citrus, genetic modification, climate change, EIC Pathfinder, Universitat Jaume I, New Genomic Techniques, RNA interference, biofortification, resveratrol, artificial intelligence, sweet orange, mandarin

The study, published in the Journal of Agriculture and Food Research, was led by Luciano Quezada and Eduardo Holzapfel together with colleagues at Chilean institutions. Unlike most optimization work in agriculture, which focuses on annual crops that can be replanted each season, the model was built specifically for perennial systems, where decisions made in one year ripple through the entire lifespan of the orchard. The team formulated a nonlinear programming problem that maximizes net profits over two decades by simultaneously allocating land, irrigation water, and labor among more than a dozen fruit crops, including wine grapes, oranges, avocados, mandarins, lemons, pears, peaches, and kiwifruit.

At the heart of the model lies a set of empirically derived crop-water production functions. These polynomial relationships link the relative yield of each fruit species to its relative evapotranspiration, the ratio of actual to potential water consumption. The curves capture a crucial and often counterintuitive feature of fruit physiology: beyond an optimal point, additional irrigation actually reduces yields. Because the functions are expressed in relative terms, they can be transferred to other regions with comparable crop and irrigation conditions, making the framework adaptable well beyond the original Chilean case study.

The objective function aggregates revenues from fruit sales minus a detailed accounting of production costs, including labor, pruning, harvesting, fertilizers, machinery, pesticides, and the separate costs of surface water and groundwater. Establishment expenses during the first three unproductive years of each orchard are incorporated, along with amortization of irrigation infrastructure distributed over a seventeen-year capital recovery period. Constraints encode the farm’s physical reality: total cultivated area cannot exceed available land, seasonal water deliveries from two surface sources and groundwater rights must cover gross irrigation demand, each crop must receive at least 55 percent of its potential evapotranspiration to avoid catastrophic stress, and annual labor availability caps the sum of person-days demanded across all orchards. The model was implemented in the General Algebraic Modeling System and solved with the MINOS nonlinear solver.

The test bed was a commercial enterprise in the O’Higgins Region, where a Mediterranean climate delivers roughly 652 millimeters of rain concentrated in winter while summers are parched. The researchers reconstructed the farm’s water availability from official records covering 2000 to 2020, a period that tells a sobering story about the region’s hydrology. Surface water supplies peaked in the early 2000s, with more than 28 million cubic meters available in some seasons, then declined sharply from 2008 onward. The 2013–2014 and 2014–2015 seasons brought severe scarcity, and deficits recurred through the end of the decade, a pattern consistent with the megadrought that has gripped central Chile.

A central innovation of the study is its explicit treatment of irrigation efficiency, measured as total distribution efficiency, or TDE, which reflects how uniformly and effectively water reaches the crop. The team compared a high-efficiency scenario of 90 percent TDE against a 70 percent scenario representing the imperfect practices commonly observed on real farms, such as poor irrigation scheduling and excessive application. The difference proved economically decisive. Under 90 percent efficiency, profit losses from water deficits ranged from just 1.0 to 2.2 percent across the farm’s historical crop patterns. At 70 percent efficiency, those losses ballooned to between 4.7 and 8.5 percent, because lower efficiency forces farmers to apply far more water to meet the same crop demand, intensifying shortages in dry years.

The deficit years revealed how the model allocates scarce water rationally. In the worst season, 2014–2015, wine grapes received only 62 percent of their required water under high efficiency, while avocados and kiwifruit received 71 and 80 percent respectively. When efficiency dropped to 70 percent, oranges, avocados, grapefruits, kiwifruit, mandarins, and apples all fell to the 55 percent floor, the minimum the model allows before severe yield penalties set in. These allocations reflect both the shape of each crop’s production function and its economic value, effectively teaching the model to triage water where it does the most good.

The centerpiece of the work is the optimal crop pattern derived from year-2000 conditions. The solution allocated the maximum permitted 30 percent of the farm to avocados and the same to mandarins, with cherries, oranges, lemons, peaches, and wine grapes filling the remainder. Notably, some of the most profitable crops, including cherries, blueberries, and apples, were reduced or excluded entirely, because their intensive seasonal labor demands collided with the farm’s annual labor ceiling. Labor availability, the researchers found, was the single most restrictive constraint shaping the optimal configuration, a finding with broad implications as labor shortages and rising wages squeeze fruit producers worldwide. The optimized pattern required about 7 percent more water per season than the historical 2000 configuration but stayed within the farm’s actual supply, and it generated over one million million Chilean pesos in cumulative profit, a 32.7 percent gain. The advantage was not immediate: the optimized pattern underperformed in the first two seasons, but from 2005–2006 onward it consistently out-earned the historical pattern, peaking at a 50.6 percent annual advantage in 2015–2016.

Sensitivity analysis probed the plan’s robustness. Moderate shocks, such as a 500 percent increase in water costs or a doubling of operational costs, reduced profits by less than 3 percent without changing the crop mix. Even a 300 percent surge in labor costs cut profits by 13.7 percent yet left the allocation untouched. The picture changed with severe water cuts: a 40 percent reduction in availability forced high-demand crops off the land and shaved 16.2 percent from profits. Market shocks proved equally potent, as halving the export price of any single crop removed it from the optimal pattern, with cherries producing the largest single loss at 8.1 percent. Interviews with the farm’s management board added a human dimension: the team praised the model’s long-term scenario capability but noted that directors’ preferences and practical constraints sometimes override purely optimal solutions, and they identified labor shortages and water fluctuations as their top operational challenges.

The authors acknowledge limitations, including the absence of an explicit soil water balance, a static crop pattern over the horizon, and the difficulty of validating against a farm whose plantings changed continuously. Even so, the study marks a rare demonstration that integrated land, water, and labor optimization can be deployed under genuine farm conditions in perennial systems. As climate change tightens water supplies across Mediterranean climates globally, and as labor costs climb, the Chilean results suggest that the most valuable harvest a fruit grower can plan for may be the one computed decades in advance.

Subject of Research: Long-term optimization of crop pattern, irrigation water, and labor allocation in perennial fruit orchards

Article Title: An integrated optimization framework for long-term crop pattern and water resource planning in fruit orchards

Article References: Quezada, L., Holzapfel, E., Kuschel-Otárola, M., Lillo-Saavedra, M., Rivera, D., Garcia-Vila, M., Rivera-Ruiz, D., & Pérez, A. (2026). An integrated optimization framework for long-term crop pattern and water resource planning in fruit orchards. Journal of Agriculture and Food Research, 31, Article 103330. https://doi.org/10.1016/j.jafr.2026.103330

Image Credits: AI Generated

DOI: 10.1016/j.jafr.2026.103330

Keywords: fruit orchards, irrigation optimization, water resources, crop pattern planning, nonlinear programming, Chile, avocado, mandarin, labor constraints, irrigation efficiency, drought, agricultural economics

The dataset, described in the Journal of Ambient Intelligence and Humanized Computing, contains 18.54 hours of audio spanning 9300 recordings produced by 34 bilingual participants, all of them fluent in both Mandarin and English. Unlike many earlier corpora, every single utterance in the collection involves code-switching. That design choice matters. Older resources such as SEAME, which stretches across roughly 190 hours, and TALCS, which covers 587 hours, contain substantial proportions of monolingual speech, meaning their effective supply of genuinely code-switched material is far smaller than their total length suggests. Some of those datasets, including the ASRU and TALCS corpora, are no longer publicly accessible at all, leaving the field with a shortage of usable, openly available benchmarks.

The construction of DOTA-ME-CS follows an unusual pipeline that blends large language model generation with human recording. The team used GPT-4o with carefully engineered prompts to produce scripted sentences across ten everyday scenarios: education, entertainment, environmental protection, food, health, home, life, pets, travel and work. Each prompt required the model to produce sentences that mimic daily conversational style, contain more English words than Mandarin words, and include at least one Mandarin word, with a dominant language assigned to every sentence. The authors justify this topic-anchored approach with a probabilistic argument: when a specific category is given, the probability of generating a relevant, high-quality sentence is higher than when the model is left to roam across all possible topics, which also reduces hidden cultural bias in the resulting scripts.

Human evaluators checked the generated scripts for grammatical problems and confirmed the presence of genuine switching, and a post-hoc naturalness study asked eight bilingual raters to score 100 randomly sampled stimuli on a five-point Likert scale. The average rating came in at 4.12 with a standard deviation of 0.61, and the median was 4.0, indicating that bilingual listeners generally perceive the generated sentences as natural, though the authors acknowledge that occasional awkwardness is an inherent limitation of LLM-based text generation. Bilingual volunteers, mostly college students with academic backgrounds in China and the United Kingdom, including roughly eighteen participants based at Imperial College London, then recorded the scripts on their own laptops or smartphones as 16-bit WAV files in quiet indoor settings. Recordings that failed basic quality checks were rejected and re-recorded, and accepted files were peak-normalised to 3 dBFS for consistent loudness.

The participants’ linguistic profiles were deliberately diverse. Twelve reported English dominance, eighteen reported Mandarin dominance and four identified as balanced bilinguals, and the mix of speakers from China and the United Kingdom ensures that the corpus spans multiple varieties of English and second-language accents. An average of 2.22 switching points per utterance, combined with a broad part-of-speech distribution across both languages, means the recordings capture switching at varied syntactic positions and grammatical categories rather than concentrating it in a single predictable spot. The dataset also includes longer recordings of roughly 10 to 15 seconds and about 100 words, an intentional response to the weakness current ASR models show on extended speech, where dependencies and critical information can be lost.

What sets the corpus apart most sharply is its AI-driven augmentation. Because human recordings were captured in quiet conditions at a normal pace, the researchers used the Librosa audio library to modify a random subset of clips. Playback speed was adjusted to 0.75x, 0.5x, 1.25x, 1.5x and 2x, with 200 recordings modified at each setting. Five categories of background noise, drawn from highways, war, natural sounds, white noise and playground environments, were added at 200 recordings per type, with the noise pitch scaled to 0.8x to account for the Lombard effect, the well-documented tendency of speakers to raise their vocal intensity in noisy surroundings. In addition, five AI-generated timbres, two male and three female, replaced the original voices in 200 recordings each, simulating both voice conversion deepfake scenarios and privacy-preserving speech processing conditions.

The accompanying data analysis is unusually thorough. Phoneme distributions catalogued with the International Phonetic Alphabet reveal that Mandarin contributes more multisyllabic pronunciations, tones, aspirated consonants and palatalised sounds, while English favours single-syllable vowels; shared features such as the sounds t and a suggest cues that future switching-detection models could exploit. Physical measurements show a frame rate of 46.7 kHz and a sample width of 2.0 bytes, meeting professional audio standards, with an average maximum pitch of 3617.13 Hz and an average per-recording pitch of 535.04 Hz. Formant frequencies computed through Linear Predictive Coding yield averages of 804.02 Hz, 4419.15 Hz and 7549.48 Hz for F1, F2 and F3, pointing to mid-to-low and front vowels, reduced lip rounding and retroflex consonants. The measured speaking rate of 2.05 words per second, with a standard deviation of 0.49, sits squarely within the typical range for conversational read speech, supporting the corpus’s ecological validity.

To establish baseline performance, the team benchmarked three pre-trained ASR models on the full dataset without fine-tuning: Whisper large-v3, a 1.5-billion-parameter Transformer encoder-decoder trained on 680,000 hours of multilingual audio; SenseVoice Small, a lightweight hybrid CTC-attention model of roughly 200 million parameters that also performs language identification, emotion detection and audio event detection; and Paraformer, a non-autoregressive Transformer with 220 million parameters optimised for Mandarin and built on continuous integrate-and-fire alignment. Paraformer and SenseVoice achieved the best results, with Whisper slightly behind, and a word error rate of 0.177 alongside a character error rate of 0.176. Compared against the publicly available ASCEND corpus, the models produced higher error rates on DOTA-ME-CS, which the authors read not as a defect but as evidence that their dataset poses a more demanding and therefore more useful benchmark.

Case studies expose exactly where today’s systems fail. In one example embedding the Chinese noun 菠萝包, meaning pineapple bun, inside an English sentence, Paraformer and SenseVoice produced garbled outputs such as bullleball and boobao, while Whisper translated the phrase correctly but destroyed the mixed-language structure by refusing to preserve the code-switch itself, a translation bias that also appeared on noise-free recordings. At double playback speed, Paraformer’s output strayed entirely from the intended meaning, SenseVoice descended into repetitions such as Miamiami, and Whisper mangled Florida lifestyle into Forensic Lifestyle. Altered timbres degraded English recognition across the board, and all three models performed worst at 2.0x speed. Whisper proved the most resilient to background noise, while Paraformer was the most sensitive to voice changes.

The implications reach beyond engineering. Reliable code-switching recognition would make automated transcription, voice assistants and translation systems far more useful for the hundreds of millions of people who live their linguistic lives between two languages, reducing a bias that currently disadvantages non-monolingual speakers. The authors stress that the study received ethical approval from Imperial College London’s ethical board, that all participants gave informed consent, and that privacy protections were maintained throughout. They are candid about limitations: funding constrained the scale of the corpus, and scripted reading, while reducing transcription errors and ethical risks, sacrifices the spontaneity of natural conversation. Even so, their deliberate trade-off of quality and coverage over raw scale gives the community something it has lacked, a fully public, exclusively code-switching Mandarin-English benchmark with baseline scores, rigorous acoustic analysis and data, code and recordings all released through a public repository. As speech technology races toward multilingual ubiquity, datasets like this one may determine whether the next generation of voice interfaces can finally follow the way people actually talk.

Subject of Research: A Mandarin-English code-switching speech dataset for advancing automatic speech recognition research.

Article Title: DOTA-ME-CS: daily oriented text audio-Mandarin English-Code switching dataset

Article References: Li, Y., Wei, Z., Yu, H., Xue, J., Zhou, H., & Schuller, B. W. (2026). DOTA-ME-CS: daily oriented text audio-Mandarin English-Code switching dataset. Journal of Ambient Intelligence and Humanized Computing. https://doi.org/10.1007/s12652-026-05119-x

Image Credits: AI Generated

DOI: 10.1007/s12652-026-05119-x

Keywords: code-switching, automatic speech recognition, Mandarin, bilingual speech, speech dataset, Whisper, Paraformer, SenseVoice, data augmentation, phonetics, large language models, voice conversion