0 views collected around this technical thread.
This article explains how Gaode Map leverages lightweight edge TTS models, dual‑autoregressive large‑model data augmentation, and a configurable audio‑processing DAG to enable users to create highly realistic personalized voice packs from just three recorded sentences.
This guide walks readers through every stage of building high‑quality AI training datasets—from locating open‑source data and defining goals, through collection, annotation, cleaning, large‑scale processing, optional augmentation, and splitting, to validation—using a medical QA example for fine‑tuning DeepSeek‑R1.
This article details Akulaku’s deployment of large language models across multimodal financial risk‑control scenarios—covering business background, a three‑module intelligent‑agent architecture, concrete tool‑ and planning‑enhancement case studies, and future outlook—demonstrating how LLMs boost efficiency, reduce labeling effort, and enable copilot‑style assistance.
This article describes a low‑cost, privacy‑preserving chatbot for logistics that combines data cleaning, large‑model‑based data augmentation, BM25 and vector retrieval, a DNN rerank model, and LLM‑driven answer rewriting to deliver accurate, compliant automated responses.
This article provides a detailed, English-language walkthrough of the HRNet source code, covering how the COCO keypoint dataset is built, the online data‑augmentation techniques applied during training, and the end‑to‑end training and inference procedures for human pose estimation.
The article explains CNN fundamentals—convolution, pooling, and fully‑connected layers—illustrates their implementation for American Sign Language letter recognition, details parameter calculations, demonstrates data augmentation and transfer learning techniques, and highlights how these methods boost image‑classification accuracy to around 92%.
This article introduces ten popular Python libraries—Augmentor, imgaug, albumentations, nlpaug, textaugment, pytorch‑geometric, audiomentations, nlpaugment, keras‑augment, and OpenCV—that provide powerful image, text, audio, and graph data augmentation techniques to improve model generalization and robustness.
This article introduces a continuous semantic enhancement approach for neural machine translation that overcomes the limitations of discrete data‑augmentation techniques, details the neighbor risk minimization training objective, presents benchmark improvements on ACL‑2022 datasets, and describes practical deployment and fine‑tuning workflows in the Modu community.
This article introduces various ChatGPT‑based data‑augmentation techniques for natural language processing, explains how to use prompts for synonym, antonym, homophone, random insertion, deletion, and swapping transformations, and provides concrete example prompts and outputs to illustrate each method.
This article surveys recent Ant Group research on explainable fraud detection, including constraint‑adaptive rule‑set learning (CRSL), meta‑path guided rule generation (MetaRule), biased sampling for imbalanced ranking, and a semi‑supervised data‑augmentation framework (SDAT) for tabular data, highlighting their motivations, methodologies, deployments, and experimental results.
In this AI Live Stream, two Taobao Technology engineers present how causal representation learning enables unbiased data augmentation and factor‑controllable generation to boost fine‑grained image classification, while also unveiling a real‑time color‑enhancement technique that merges cascaded lookup tables with dynamic neural networks, illustrating modern AI trends and practical deployment strategies.
58.com’s NLP senior engineer explains how a recruitment knowledge graph is built—through multi‑dimensional tag systems, tag mining, and relation extraction—and how it enhances bidirectional matching and recommendation efficiency, addressing challenges such as weak expression, cold start, and supply‑demand imbalance.
The talk explains how to maximize the value of training data by exploring background on model generalization, automatic data augmentation techniques, denoising strategies, active learning for selecting unlabeled samples, and robust data splitting methods, offering practical guidelines for AI practitioners.
RapNeRF enhances Neural Radiance Fields for extreme view extrapolation by introducing Random Ray Casting and a Ray Atlas, which together augment training data and store view‑dependent surface features, enabling robust, high‑quality novel‑view synthesis from sparse images and outperforming prior methods on synthetic and real datasets.
This article documents the author’s step‑by‑step journey of building an OCR system for recognizing Chinese characters in a card‑game UI, covering game selection, technical background, data generation, deep‑learning model training with CRNN, real‑image data collection, optimization attempts, and final performance evaluation.
After the BERT era, this article reviews the limitations of pre‑trained language models for semantic matching, discusses negative‑sample sampling, data‑augmentation techniques, contrastive learning methods such as ConSERT and SimCSE, and practical deployment considerations in vector‑based retrieval systems.
This article presents a comprehensive case study of applying named‑entity‑recognition (NER) techniques to the smart chat assistant of 58.com’s yellow‑page service, covering business background, model selection (BiLSTM‑CRF, IDCNN‑CRF, BERT), data‑augmentation, focal loss, fusion of rule‑based and neural methods, context modeling, online performance, and future research directions.
The article summarizes two INTERSPEECH 2021 papers from Beike's voice technology team, detailing a grammar‑based semantic data augmentation method that improves end‑to‑end Chinese speech recognition and introducing GigaSpeech, a massive 10,000‑hour multilingual English speech dataset for robust ASR research.
This article presents Didi's exploration of few‑shot learning, data‑augmentation, semi‑supervised self‑training and multi‑task learning techniques to address the scarcity of labeled samples in safety and governance scenarios, demonstrating practical solutions and performance gains across various risk‑detection tasks.
To overcome scarce labeled data for safety and governance, Didi combines few‑shot learning with systematic data augmentation, self‑training semi‑supervised labeling, and multi‑task neural architectures, cutting labeling costs and reducing log‑loss by over 20% while boosting ROC‑AUC and PR‑AUC across harassment detection, expense‑complaint, and route‑intercept use cases.