0 views collected around this technical thread.
This article showcases an AI‑generated song that teaches the four core classification metrics—sensitivity, specificity, precision, and recall—by presenting lyrical explanations, a confusion‑matrix overview, Python code for MIDI creation, and a step‑by‑step guide to producing the final video.
We built an LLM‑powered system for Bilibili that automatically creates ad titles from user keywords, employing fluency, style, and quality classifiers, mixed domain data cleaning, and alignment methods such as SFT, DPO and KTO, resulting in a product that now generates about ten percent of daily titles and drives significant ad spend.
The ICDAR 2024 Historical Map Text Recognition competition was won by Bilibili’s DNTextSpotter, a Transformer‑based model built on DeepSolo and ViTAE‑v2 that uses deformable self‑attention, dual‑query decoding and denoising training, combined with mixed‑vocabulary fine‑tuning, advanced loss functions and strict PDQ/PWQ/PCQ metrics to achieve state‑of‑the‑art dense, rotated, arbitrary‑shaped text detection and recognition on historical maps and real‑world multimedia.
AlignRec is a joint‑training framework that synchronizes multimodal encoders with personalized recommendation models through a staged alignment strategy and three specialized loss functions, preserving both content and ID signals, and achieving state‑of‑the‑art performance on multiple datasets while releasing superior Amazon multimodal features.
This article provides a comprehensive guide on building user portrait tags—from basic attribute tags to business and strategy tags—detailing data collection methods, feature engineering techniques such as cleaning, time decay, and smoothing, and evaluation metrics for cohesion and stability, aimed at data product managers and analysts.
The survey traces the evolution of image‑quality evaluation for text‑to‑image generation—from early handcrafted edge and color cues, through GAN‑era similarity scores such as IS, FID and KID, to modern perceptual and CLIP‑based metrics like LPIPS, CLIPScore, TRIQ, IQT and human‑preference models—highlighting a shift toward semantic, aesthetic, and text‑image alignment measures and forecasting domain‑specific metrics for future diffusion models.
This article explains how to build user portrait tags—from basic attribute tags to business and strategy tags—covers methods for data collection, anomaly handling, time decay, smoothing, and evaluates tag quality using cohesion, stability, and AUC-related metrics to support data‑driven product decisions.
This article analyzes challenges in advertising placement, introduces user value modeling practices such as CLTV estimation, discusses data sparsity, multi‑distribution issues, evaluation metrics, and presents future work on budget allocation and iterative model improvement for growth optimization.
This article provides a detailed overview of modern recommendation system technology, covering system architecture, user understanding layers, various recall and ranking techniques, additional algorithmic directions such as cold‑start and bias modeling, and the evolving evaluation metrics used in practice.
This article introduces the evolution of embedding techniques, presents real‑world case studies of embedding‑based FAQ retrieval, explains evaluation metrics such as Recall and MRR, and compares the performance of a proprietary ZhongAn embedding model with OpenAI and Sentence‑BERT models on Chinese FAQ datasets.
This article reviews the fundamentals, feature extraction methods, evaluation metrics, and common datasets of content‑based image retrieval (CBIR), discussing traditional low‑level features, local descriptors, unsupervised and supervised learning approaches, and recent deep‑learning models for improving retrieval performance.
This article reviews common classification evaluation metrics—accuracy, precision, recall, and F1—explains their limitations on imbalanced data, and introduces the Matthews Correlation Coefficient (MCC) with Python implementations to provide a more reliable performance measure.
This article presents a detailed survey of recent advances in recommendation system technology, covering system architecture, user understanding layers, various recall methods, ranking techniques, auxiliary algorithms such as cold-start and bias modeling, and evaluation metrics, with references to industry practices and academic research.
This article presents Tencent AI Lab's intelligent writing assistant, detailing the TexSmart text‑understanding platform, the Effidit writing‑assistant features, a multi‑level controllable unsupervised text‑rewriting method, and a novel ParaScore metric that jointly measures semantic similarity and diversity for paraphrase evaluation.
This interview compiles expert opinions on the end‑to‑end recommendation system pipeline—including architecture, data collection, user profiling, content structuring, feature engineering, recall strategies, ranking algorithms, multi‑objective optimization, multi‑modal fusion, re‑ranking, cold‑start solutions, evaluation metrics and real‑world applications—highlighting the technical challenges and practical solutions.
This interview with three recommendation‑system experts explores the technical architecture, data sources, feature engineering, recall and ranking strategies, evaluation metrics, cold‑start solutions, and practical difficulties, offering actionable insights to avoid common pitfalls in real‑world recommender deployments.
This article presents the QQ Browser search relevance team's experience integrating QQ Browser and Sogou search systems, detailing business overview, relevance system evolution, algorithm architecture, evaluation metrics, deep semantic matching, relevance calibration, and model distillation techniques to improve search relevance performance.
This guide explains essential evaluation metrics for classification and regression models—including confusion matrix, ROC/AUC, R², and main performance indicators—covers model selection strategies such as train‑validation‑test splits, k‑fold cross‑validation, and regularization techniques, and discusses bias‑variance trade‑offs and diagnostic tools.
The QQ Browser search relevance team describes a unified, billion‑scale architecture that combines a main and vertical subsystem, a pyramid‑shaped ranking pipeline (recall, coarse, fine), a dedicated GPU‑accelerated relevance service, and hybrid semantic‑matching models (dual‑tower, BERT, matrix fusion) evaluated with offline and online metrics to deliver accurate, fresh, and authoritative results for diverse content and long‑tail queries.
This article explains data governance fundamentals, outlines common evaluation shortcomings, and introduces observability concepts and the GuanCe Cloud platform as a way to objectively measure and improve governance outcomes across the entire data lifecycle.