Tagged articles
text classification
60 articles · Page 1 of 1
ScaleDown and Jev are reviving classic NLP tasks with task-specific small language models that achieve higher accuracy, 200x faster speeds, and 467x lower costs than frontier LLMs, signaling a shift toward specialized architectures for production AI.
JevNLP industry trendsScaleDown
0 likes · 11 min read

The article analyzes the shift from single‑loop engineering to graph engineering, exposing the pitfalls of optimizing a single metric such as overfitting and Goodhart’s law, and demonstrates how multi‑loop supervision, anchored by anchors, frozen nodes, and external judgment, can produce more reliable AI agents, illustrated with a text‑classification case study.
AI agentsGoodhart's LawGraph Engineering
0 likes · 18 min read

This article explains how deep learning models such as RNN, LSTM, GRU and Transformer enable NLP tasks like machine translation, text classification, question answering and text generation, outlines their advantages over traditional methods, and provides a Keras code example for text classification.
KerasNLPPython
0 likes · 8 min read

This article demonstrates how Hologres AI Function enables end‑to‑end text classification directly in the database using a single SQL call, covering data preparation, prompt engineering, batch inference, accuracy evaluation (up to 95%), and cost analysis with KV‑Cache optimization that reduces token charges to as low as 0.11 CNY for 200 reviews.
AI FunctionCost OptimizationHologres
0 likes · 12 min read

This article walks through a real text‑classification project where noisy labels inflated the review workload to over 15,000 samples, and shows how using cleanlab’s confident‑learning framework reduced the manual audit set to 438 items, boosting efficiency by thirty‑four times while improving model performance.
Data-Centric AIcleanlabconfident learning
0 likes · 16 min read

This guide shows how to use the Tongyi Qianwen large‑model API with PHP to automatically classify user feedback into predefined categories, eliminating manual analysis and complex NLP development while providing clear steps, code, and result interpretation for rapid business insights.
APIPHPautomation
0 likes · 7 min read

This guide walks you through the end‑to‑end process of building, training, converting, and deploying a TensorFlow Lite text‑classification model on Android, covering data preparation, model selection, performance trade‑offs, and integration using the TFLite Task Library.
AndroidTensorFlow Litetext classification
0 likes · 19 min read

Bilibili combats personal‑attack and trolling comments by combining sector‑specific keyword databases, user‑group analysis, advanced word‑matching (including pinyin and homophone detection) and multiple NLP/graph models, which has cut personal‑attack reports in entertainment, film and gaming by about 32 % and trolling reports by roughly 25 % between June and December 2023.
BilibiliContent moderationabusive language detection
0 likes · 12 min read

This guide explains how to use ModelScope’s trainer components to fine‑tune a pretrained backbone for text classification, covering dataset loading, configuration modification, trainer construction, training, evaluation, prediction, and checkpoint management with concrete code examples.
ModelScopePyTorchTrainer
0 likes · 11 min read

This tutorial walks through creating a custom iOS SMS spam filter app, covering extraction of personal SMS data from an iPhone backup, training a CoreML text‑classification model with CreateML, implementing a Message Filter Extension in Xcode, and exploring advanced update strategies.
App ExtensionCoreMLMobile App
0 likes · 12 min read

Researchers from Peking University and Huawei present a multiscale positive‑unlabeled learning framework that significantly improves detection of AI‑generated short and long texts, addressing the difficulty of distinguishing AI‑written content from human writing and outperforming existing baselines on multiple benchmarks.
AI detectionPu-Learninglarge language models
0 likes · 8 min read

This article explains how Tubi leverages natural‑language processing, sentence embeddings (USE and BERT), and LightGBM models to automatically categorize large volumes of Net Promoter Score comments and customer‑support tickets, enabling data‑driven product decisions and workflow automation.
LightGBMNLPTubi
0 likes · 11 min read

This article explains how regular expressions can be converted into equivalent neural network models—FA‑RNN for classification and FST‑RNN for slot filling—by leveraging finite‑state automata, tensor decomposition, and pretrained word embeddings, achieving zero‑shot performance and strong results in low‑resource scenarios.
FA-RNNneural networksregular expressions
0 likes · 17 min read

In this interview, Zuoyebang’s NLP lead explains how the company built self‑developed platforms like IQC and FTP to automate text quality inspection and intelligent labeling, outlines their architecture, shares practical deep‑learning applications such as translation and grammar correction, and discusses future research directions in large‑scale multi‑label classification, few‑shot learning, and multimodal models.
AI platformsNLPmachine learning
0 likes · 11 min read

This article reviews recent developments in text classification and intent recognition, comparing industrial practices such as business‑coupled feature engineering with research trends like pretrained language models, and provides references and practical insights for building effective NLP solutions.
NLPPretrained Modelsindustry applications
0 likes · 13 min read

This article describes the end‑to‑end multimodal modeling pipeline—covering text, image, and video understanding, model evolution from shallow to deep networks, key‑frame extraction, fine‑tuning, and multimodal fusion—used in Tencent's game ad exchange platform, along with practical deployment challenges and solutions.
AdvertisingCNNTransformer
0 likes · 22 min read

This talk presents an overview of text‑plus‑speech multimodal emotion analysis, covering background, single‑modal text and audio models, the MSCNN‑SPU multimodal architecture, domain‑adaptation techniques, and future directions, with detailed model explanations, experimental results, and practical deployment insights.
Speech Recognitionaudio processingdeep learning
0 likes · 40 min read

The paper proposes a simple yet effective method that fuses label embeddings into BERT, enhancing text‑classification performance without increasing computational cost, and validates the approach across six benchmark datasets, also exploring tf‑idf‑based label augmentation and the impact of using [SEP] versus no‑[SEP] inputs.
BERTNLPTF-IDF
0 likes · 8 min read

This article presents a comprehensive study on detecting malicious webpages by constructing heterogeneous graphs from URL redirection and textual features, applying Graph Convolutional Networks and Cluster‑Text‑GCN models, detailing optimization techniques for large‑scale deployment, and outlining future research directions.
GCNGraph Neural Networksheterogeneous graph
0 likes · 11 min read


Code DAODec 12, 2021 · Artificial Intelligence
This article explains practical techniques for improving NLP model accuracy on massive corpora, covering challenges of multi‑field text, word‑embedding choices, a fasttext‑based regression demo with book‑review data, feature engineering tricks, and a comparison with tf‑idf + LASSO.
NLPPythonWord2Vec
0 likes · 13 min read

This article introduces a multimodal advertisement detection framework for WeChat KanKan that decomposes the problem into text, image, and article‑structure dimensions, presents novel models for ad text and image recognition, and describes how sequence classification and visualisation are used to filter severe ad‑spam articles.
Content moderationWeChatadvertisement detection
0 likes · 16 min read


58 TechAug 10, 2021 · Artificial Intelligence
This article presents a comprehensive study on extracting semantic tags from 58.com voice data, detailing the use of active learning to address cold‑start problems, comparing keyword matching, XGBoost, TextCNN, CRNN, and an improved Wide&Deep model, and demonstrating significant reductions in labeling effort and superior F1 scores across multiple experiments.
Active LearningCRNNmodel comparison
0 likes · 15 min read

This article presents the background, problem analysis, data preprocessing, modeling approaches and optimization results of applying various NLP methods—including statistical models, word embeddings, attention mechanisms and pretrained language models such as BERT—to improve the accuracy of classifying Ctrip ticket customer service dialogues.
BERTCustomer ServiceNLP
0 likes · 13 min read

The article describes how NetEase adapts a pre‑trained BERT model to classify news articles into ultra‑short, short, or long timeliness categories by combining rule‑based strong and weak time cues, key‑sentence extraction, domain‑embedding fusion and multi‑layer semantic aggregation, achieving accurate and interpretable predictions for its platform.
BERTNLPNews Timeliness
0 likes · 12 min read


58 TechMar 1, 2021 · Artificial Intelligence
This article describes how 58.com’s AI Lab built and continuously improved the QABot intelligent customer‑service system by designing classification and retrieval models, evaluating FastText, LSTM‑DSSM, BERT and a self‑developed SPTM framework, and finally fusing them to boost answer rates and user experience.
AI chatbotBERTQABot
0 likes · 9 min read


58 TechSep 21, 2020 · Artificial Intelligence
The 58.com AI Algorithm Competition showcased intelligent customer‑service technology, with 158 teams competing on text classification and matching tasks, and the top five teams presenting detailed BERT, ELECTRA, focal‑loss and multi‑model fusion solutions along with award ceremonies, video recordings and PPT resources.
AIBERTCompetition
0 likes · 9 min read

This comprehensive guide examines real‑world text classification projects, covering label taxonomy design, data scarcity solutions, efficient annotation, new‑class discovery, algorithm selection, evaluation metrics, OOV handling, model evolution, rule‑model integration, performance‑boosting tricks, and inference under resource constraints.
Algorithm SelectionFew-Shot LearningModel Evaluation
0 likes · 15 min read


58 TechAug 14, 2020 · Artificial Intelligence
This article provides a step‑by‑step guide on preparing data, pre‑training the SPTM lightweight model, fine‑tuning a text‑classification model with qa_match, and generating competition‑ready predictions for the 58 City AI Algorithm Contest, including all required shell commands and parameter explanations.
AICompetitionPretraining
0 likes · 9 min read


58 TechAug 12, 2020 · Artificial Intelligence
This article provides a step‑by‑step tutorial on preparing data, pre‑training the SPTM language model, fine‑tuning a text‑classification model, generating predictions, and creating a submission file for the 58.com AI algorithm competition using the open‑source qa_match toolkit.
AICompetitionModel Training
0 likes · 9 min read

The article reviews the principles of popular pretrained language models, compares their architectures, and details how NetEase Yanxuan applied BERT, GPT and ELMo to classification, matching, sequence labeling and generation tasks, presenting experimental results and deployment insights.
BERTNLPpretrained language models
0 likes · 20 min read

The article details the Tencent Advertising Algorithm competition final, explains the chizhu team's approach of converting ad IDs into word sequences for text classification using large‑scale word2vec embeddings and a dual BiLSTM architecture, presents custom loss functions, training tricks, and shares full Python model code, achieving an overall rank of 11.
AdvertisingBiLSTMCompetition
0 likes · 9 min read

This article investigates the practical performance of Google’s 2019 Unsupervised Data Augmentation (UDA) framework on real‑world financial text classification tasks, detailing experiments with limited labeled data, domain‑out‑of‑distribution samples, noisy labels, and comparisons between BERT and lightweight TextCNN models.
BERTData AugmentationFinancial NLP
0 likes · 21 min read

This article shares a competition winner's step‑by‑step solution for predicting user age and gender from ad click sequences, treating IDs as words, using word2vec embeddings, a custom transformer‑LSTM model, dual‑task loss, and weight‑search post‑processing.
AdvertisingCompetitionNLP
0 likes · 7 min read

This article describes a WeChat hotspot mining platform that integrates multiple data sources, builds quality and prediction models, employs advanced clustering and multi‑granular text matching techniques, and uses generative active learning to efficiently discover, predict, and present news hotspots for users.
Active LearningWeChathotspot detection
0 likes · 17 min read

This article describes how a BERT‑based model is fine‑tuned to compute sentence‑pair similarity for improving recommendation accuracy in an online school, detailing the architecture, training mechanisms, code implementation, experimental results, and future extensions such as sentiment analysis.
BERTChinese NLPdeep learning
0 likes · 20 min read


Amap TechJan 3, 2020 · Artificial Intelligence
Amap replaced its rule‑based feedback pipeline with a three‑stage, LSTM‑driven system that combines word2vec embeddings and structured fields, achieving over 96% classification accuracy, cutting manual workload by 80%, and slashing per‑task costs while enabling scalable, data‑driven map quality improvements.
Gaode MapsLSTMNLP
0 likes · 14 min read

This article reviews the evolution of language models from bag‑of‑words to BERT, compares ELMo, GPT, and BERT architectures, and details how NetEase Yanxuan applies pre‑trained models to classification, text matching, sequence labeling, and generative tasks in production.
BERTELMoGPT
0 likes · 19 min read

This article describes the end‑to‑end development of a multi‑class NLP system for Ctrip airline ticket customer service, covering problem analysis, data preprocessing, sample balancing, model architecture (TextCNN and Bi‑GRU), training strategies, performance evaluation, and online customization to achieve high accuracy in intent recognition.
Bi-GRUCustomer ServiceNLP
0 likes · 16 min read

This article introduces the Huajiao platform's text anti‑spam architecture, analyzes spam categories and challenges, compares rule‑based and machine‑learning approaches, details traditional NLP methods and the TextCNN deep‑learning model, provides its TensorFlow implementation, and describes the online deployment workflow.
CNNNLPTensorFlow
0 likes · 14 min read

This article introduces the challenges of text‑based spam on the Huajiao platform, reviews traditional rule‑based and machine‑learning classification methods, explains the TextCNN architecture for robust character‑level detection, and details its TensorFlow Serving deployment for real‑time anti‑spam services.
CNNTensorFlowanti-spam
0 likes · 16 min read

This article presents a comprehensive solution to the absolute semantic recognition competition, detailing the problem background, dataset, evaluation metrics, feature engineering, model architecture—including Attention, Capsule, Bi‑GRU, and BERT—and analysis of results and lessons learned.
AttentionBERTCapsule Networks
0 likes · 11 min read

The article details a multi‑turn dialogue intent classification pipeline that extracts and expands labeled utterances, preprocesses text with custom tokenization, trains a two‑layer CNN‑Highway and a multi‑head self‑attention model, analyzes errors, and achieves up to 98.7% accuracy on a large, balanced dataset.
BERTCNNdialogue system
0 likes · 15 min read

Kui is iQiyi’s AI‑driven anti‑spam platform that protects online communities through a three‑layer architecture—service, algorithm strategy, and auxiliary modules—and employs keyword, rule‑based, machine‑learning, and risk‑control strategies to detect advertising, pornographic, abusive and other malicious content while continuously adapting to evolving threats.
AI systemContent moderationRisk Control
0 likes · 10 min read


58 TechFeb 22, 2019 · Artificial Intelligence
The article details the design and iterative improvement of 58.com’s intelligent QABot, covering knowledge‑base construction, feature engineering, three generations of classification models—including FastText, Bi‑LSTM, and deep semantic matching—and evaluation metrics that achieve high accuracy and automation rates.
AIQABotdeep learning
0 likes · 12 min read

The paper presents an integrated multimodal quality assessment system for short‑video platforms that evaluates cover images, video content, and accompanying text using deep‑learning and handcrafted features—combining ResNet‑50, NetVLAD, TSN, VGGish, and XGBoost—to improve user experience, recommendation accuracy, and operational efficiency, with plans for optimization and modular deployment.
image analysismultimodal learningtext classification
0 likes · 11 min read

This article explains BERT’s architecture—from its bidirectional Transformer encoder and attention mechanisms to its pre‑training tasks—and presents extensive experiments showing its superior performance on various Chinese and English text‑classification benchmarks across multiple datasets.
BERTNLPPretraining
0 likes · 22 min read

This article introduces reinforcement learning fundamentals, contrasts it with supervised learning, and explores its challenges and advantages in natural language processing, including applications such as text classification, relation extraction from noisy data, and weakly supervised topic segmentation, while summarizing key insights and experimental results.
Weak Supervisionnatural language processingreinforcement learning
0 likes · 11 min read


AntTechAug 16, 2018 · Artificial Intelligence
This article reviews deep‑learning‑based text classification techniques—including TextCNN, BiGRU, Capsule Networks, Attention mechanisms, and the novel cw2vec embedding—applied to Alipay complaint fraud data, presents experimental comparisons, and discusses their advantages, challenges, and future directions.
AlipayAttentioncapsule network
0 likes · 18 min read

This article explains the principles of Naive Bayes text classification, detailing feature representation, model selection, training and testing procedures, probability calculations, code implementation in Python, and evaluation metrics such as accuracy, precision, recall, PR and ROC curves.
Evaluation MetricsNaive BayesPython
0 likes · 22 min read

The article proposes a hybrid approach that first filters blacklist words and then classifies suspicious comments with a character-level TextCNN, achieving around 89% precision and 87% recall, demonstrating that simple convolutional networks outperform keyword filters and RNNs for short, noisy abusive Chinese text.
Abusive Comment DetectionContent moderationNLP
0 likes · 10 min read

This article introduces the Naive Bayes classifier, explains its underlying probability formulas—including conditional probability, total probability, and the Bayes theorem—covers the feature independence assumption, Laplace smoothing, and demonstrates both manual and scikit‑learn implementations for email and text classification with Python code.
Naive Bayesprobabilityscikit-learn
0 likes · 11 min read

This article provides a comprehensive guide to text classification, covering traditional pipelines, bag‑of‑words and TF‑IDF features, dimensionality‑reduction techniques, word‑embedding models such as GloVe, word2vec and fastText, and modern deep‑learning architectures like CNN, RCNN and HAN.
CNNNLPRNN
0 likes · 9 min read

This article walks through constructing a Spark-powered text sentiment analysis pipeline—from crawling movie reviews, preprocessing and feature extraction with jieba and TF‑IDF, to training Naive Bayes and SVM classifiers—while discussing Spark's advantages and ways to improve model accuracy.
NLPPythonSpark
0 likes · 19 min read

The team built a lightweight three‑class sentiment classifier for iQIYI user comments using a linear‑kernel SVM with high‑dimensional bag‑of‑words features and an expanded ~100k word lexicon, achieving over 96% accuracy across domains, and deployed it as a Spring Boot PMML service with zero‑downtime refresh, while planning GBDT‑enhanced features and word‑embedding optimizations.
NLPSVMdeployment
0 likes · 13 min read

The article answers ten machine‑learning questions, explaining how to impute missing behavior data, extract and select features, describe Meituan‑Dianping’s recommendation pipeline, suggest a beginner learning path, clarify L1 sparsity, recommend TextCNN for text, discuss search‑ranking sample bias, label generation for wide‑deep models, the shift to deep‑learning video detection, and the use of factorization machines for CTR with open‑source examples.
L1 Regularizationdata imputationdeep learning
0 likes · 7 min read

Alibaba’s Tmall Genie X1 showcases a new semantic encoding model called Conv‑RNN that improves intelligent question answering and text classification, achieving state‑of‑the‑art results on benchmark datasets while illustrating the broader impact of semantic encoding on NLP applications.
NLPQuestion Answeringconv-RNN
0 likes · 8 min read

At Qunar, the data center algorithm team developed an automatic ticket classification system that combines Support Vector Machine with word2vec embeddings to handle high‑dimensional, low‑sample text data, achieving 89% accuracy and 80% recall while outlining the full machine‑learning pipeline from feature extraction to deployment.
QunarSVMTicket Classification
0 likes · 7 min read


21CTOFeb 12, 2016 · Artificial Intelligence
This article uses machine learning to compare lexical patterns between the first 80 and last 40 chapters of 'Dream of the Red Chamber', demonstrating distinct stylistic differences that support the scholarly view that the final chapters were not authored by Cao Xueqin.
Red MansionsSupport Vector Machinefeature engineering
0 likes · 6 min read

This article explains Suning's implementation of a Naive Bayes‑based product classification system, detailing its basic theory, formal definition, step‑by‑step training process, three implementation phases, evaluation results, and error analysis to improve classification accuracy.
AlgorithmNaive BayesSuning
0 likes · 6 min read

This article details a step‑by‑step machine‑learning pipeline that transforms over one million calibrated POI records into feature vectors, selects discriminative terms via information‑gain and domain rules, trains a Naive Bayes classifier, and achieves 91% accuracy with 84% coverage on unseen POI data.
Chinese NLPNaive BayesPOI classification
0 likes · 12 min read
