Grapheme-to-phoneme conversion (G2P) is the task of converting letters (grapheme sequences) into their pronunciations (phoneme sequences). It plays a crucial role in natural language processing, text-to-speech synthesis, and automatic speech recognition systems. This paper provides a systematical overview of the G2P conversion from different perspectives. The conversion methods are first presented in the paper; detailed discussions are conducted on methods based on deep learning technology. For each method, the key ideas, advantages, disadvantages, and representative models are summarized. This paper then mentioned the learning strategies and multilingual G2P conversions. Finally, this paper summarized the commonly used monolingual and multilingual datasets, including Mandarin, Japanese, Arabic, etc. Two tables illustrated the performance of various methods with relative datasets. After making a general overall of G2P conversion, this paper concluded with the current issues and the future directions of deep learning-based G2P conversion.

Face tracking serves as the crucial initial step in mobile applications trying to analyse target faces over time in mobile settings. However, this problem has received little attention, mainly due to the scarcity of dedicated face tracking benchmarks. In this work, we introduce MobiFace, the first dataset for single face tracking in mobile situations. It consists of 80 unedited live-streaming mobile videos captured by 70 different smartphone users in fully unconstrained environments. Over 95K bounding boxes are manually labelled. The videos are carefully selected to cover typical smartphone usage. The videos are also annotated with 14 attributes, including 6 newly proposed attributes and 8 commonly seen in object tracking. 36 state-of-the-art trackers, including facial landmark trackers, generic object trackers and trackers that we have fine-tuned or improved, are evaluated. The results suggest that mobile face tracking cannot be solved through existing approaches. In addition, we show that fine-tuning on the MobiFace training data significantly boosts the performance of deep learning-based trackers, suggesting that MobiFace captures the unique characteristics of mobile face tracking. Our goal is to offer the community a diverse dataset to enable the design and evaluation of mobile face trackers. The dataset, annotations and the evaluation server will be on https://mobiface.github.io/.

The uncommon metabolic pathways of organic pollutants are easily overlooked, potentially leading to idiosyncratic toxicity. Prediction of their biotransformation associated with the toxic effects is the very purpose that this work focuses, to develop a de novo method to mechanistically predict the reactive toxicity pathways of uncommon metabolites from start aliphatic amine molecules, which employed sertraline triggered by CYP450 enzymes as a model system, as there are growing concerns about the effects on human health posed by antidepressants in the aquatic environment. This de novo prediction strategy combines computational and experimental methods, involving DFT calculations upon sequential growth, in vitro and in vivo assays, dissecting chemically reactive mechanism relevant to toxicity, and rationalizing the fundamental factors. Significantly, desaturation and debenzylation-aromatization as the emerging metabolic pathways of sertraline have been elucidated, with the detection of DNA adducts of oxaziridine metabolite in mice, highlighting the potential reactive toxicity. Molecular orbital analysis supports the reactivity preference for toxicological-relevant C-N desaturation over N-hydroxylation of sertraline, possibly extended to several other aliphatic amines based on the Bell-Evans-Polanyi principle. It was further validated toward some other wide-concerned aliphatic amine pollutants involving atrazine, epsilon-caprolactam, 6PPD via in silico and in vitro assays, thereby constituting a complete path for de novo prediction from case study to general applications.

Furan-containing compounds distribute widely in food, herbal medicines, industrial synthetic products, and environmental media. These compounds can undergo oxidative metabolism catalyzed by cytochrome P450 enzymes (CYP450) within organisms, which may produce reactive products, possibly reacting with biomolecules to induce toxic effects. In this work, we performed DFT calculations to investigate the CYP450-mediated metabolic mechanism of furan-ring oxidation using 2-methylfuran as a model substrate, meanwhile, we studied the regioselective competition of another hydroxylation reaction involving methyl group of 2-methylfuran. As a result, we found the toxicological-relevant cis-enedione product can be produced from O-addition directly via a concerted manner without formation of an epoxide intermediate as traditionally believed. Moreover, our calculations demonstrate the kinetic and thermodynamic feasibility of both furan-ring oxidation and methyl hydroxylation pathways, although the former pathway is a bit more favorable. We then constructed a linear model to predict the rate-limiting activation energies (ΔE*) of O-addition with 11 diverse furan substates based on their adiabatic ionization potentials (AIPs) and condensation Fukui functions (CFFs). The results show a good predictive ability (R²=0.94, Q²_CV=0.87). Therefore, AIP and CFF with clear physichem meanings relevant to the mechanism, emerge as pivotal molecular descriptors to enable the fast prediction of furan-ring oxidation reactivities for quick insight into the toxicological risk of furans, using just ground-state calculations.