This article provides a comprehensive overview of automatic text summarization, covering extractive, abstractive, and hybrid methods, system classifications, applications, datasets, evaluation metrics, and future research directions within the field of artificial intelligence.
This presentation details the XiaoAi intelligent assistant’s QA system, covering its application scenarios, information extraction techniques (including relation and event extraction with SPO/PSO models), graph‑based question answering methods, cross‑domain slot extraction, path retrieval, and practical Q&A insights.
This article explains the importance of text generation, introduces the PEGASUS model’s gap‑sentence pre‑training for abstractive summarization, and shows how the EasyNLP framework integrates PEGASUS and other Chinese and English summarization models with step‑by‑step installation, data preparation, and training commands.
Graph4NLP is a PyTorch‑ and DGL‑based open‑source library that provides a full pipeline—from static and dynamic graph construction to embedding, learning, prediction, and inference—for applying graph neural networks to a wide range of NLP tasks, with extensive documentation, demos, and future scalability plans.
This article describes the production challenges of using BERT for large‑scale text classification at Zuoyebang, explores lightweight alternatives such as knowledge distillation, pruning and quantization, and details a teacher‑student‑active‑learning pipeline that trains a TextCNN model to match BERT performance while dramatically reducing GPU consumption and improving throughput.
This article presents a comprehensive overview of AI-driven music comment moderation and ranking systems, detailing business scenarios, model architectures, data processing techniques, performance improvements, and future directions for both QQ Music and K‑Song platforms.
This article introduces GAST, a graph‑adaptive semantic transfer framework that combines POS‑based Transformers and hybrid graph attention to improve cross‑domain sentiment analysis, presents related work, details the model architecture, reports extensive experiments showing state‑of‑the‑art results, and discusses future directions.
This article presents a comprehensive overview of concept tag mining for recommendation systems, describing unsupervised pattern‑matching, semi‑supervised AutoPhase, and supervised NER approaches, analyzing their advantages and drawbacks, and offering practical solutions to tag duplication and quality issues.
This article explains the role of Youzan’s Query Parser (QP) in search, walks through its overall and layered architecture, details each algorithmic plugin—from preprocessing to synonym handling—and shows concrete code examples and results that improve search relevance across multiple retail scenarios.
The article details Snowball's large‑scale financial knowledge graph, covering its background challenges, two‑layer ontology and data design, data sourcing and pipeline, graph database selection, search and NLP services, domain‑specific pre‑training models, and future industrial considerations.
This guide walks through setting up a Java GraalVM 17 environment, installing Nginx to serve static dictionary files, configuring a HanLP‑based Elasticsearch analyzer plugin, packaging and deploying it, and testing the analyzer with JUnit5 and curl commands.
The DataFun 2022 NLP Summit, held on July 30, brings together top researchers and industry leaders from Alibaba, Baidu, Microsoft, Amazon, and more to present the latest advances in large‑scale pre‑training, multimodal perception, information extraction, dialogue interaction, machine translation, and practical AI deployments, with live streaming and free registration via QR code.
This article reviews the recent NLPCC 2022 NER competition, explains the evolution of named entity recognition, details the five major modeling paradigms, and describes the winning team’s relation‑classification approach, data‑augmentation strategy, experimental results, and its practical deployment in NetEase Cloud Commerce services.
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.
The CCL2022‑DQAB competition, organized by Beihang University and AVIC Mobile Technology, invites participants to develop reading‑comprehension models for extracting accurate question‑answer pairs from civil aviation texts, offering detailed task definitions, evaluation criteria, dataset statistics, a prize structure, and a competition schedule.
Ant Group's insurance technology team secured the top spot in the fine‑grained dialogue social bias detection task at the 11th CCF NLPCC conference, showcasing their AI‑driven bias‑mitigation methods, a proprietary pre‑trained model AntInsBert, and a claim‑automation system that boosts insurance service fairness and efficiency.
This article provides a comprehensive overview of natural language processing, covering its definition, historical development from one‑hot encoding to modern models such as word2vec, ELMo, GPT, and BERT, and discusses the advantages, limitations, and key concepts of each technique.
The article outlines the evolution, architecture, and core value of intelligent customer service systems—detailing their GUI‑based chatbot interface, triage and dialogue modes, knowledge‑base management, and operator benefits—while highlighting future trends such as richer human‑like interactions, 5G‑enabled channels, and continuous feedback‑driven improvement.
The OPPO XiaoBu team introduced OBERT, a series of 100M‑, 300M‑, and 1B‑parameter pretrained language models that leverage massive TB‑scale corpora, multi‑granular masking, retrieval‑augmented training, and distributed acceleration to achieve state‑of‑the‑art results on CLUE and KgCLUE benchmarks while enabling efficient industrial deployment.
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.
Ant Risk AI’s team secured the FinQA competition champion by presenting a comprehensive model that combines a retriever and program generator, detailed dataset analysis, domain-specific language design, and extensive experiments demonstrating superior execution and program accuracy on financial numerical reasoning tasks.
The 2022 Global AI Technology Innovation Competition – Algorithm Challenge, co‑hosted by JD Retail and academic partners, brought together 12 finalist teams from over 3,000 entrants to tackle e‑commerce‑focused AI problems such as multimodal image‑text matching and product‑title entity recognition, highlighting real‑world business impact and fostering talent exchange.
The article presents Ctrip's travel AI team's end‑to‑end entity linking solution built on a large‑scale tourism knowledge graph, detailing its background, technical architecture, core modules—including mention detection, candidate generation, and disambiguation using BERT and prefix‑tree techniques—and real‑world applications such as search, intelligent客服, and POI data maintenance.
The article describes the CCL2022 Video Highlight Extraction Challenge, a competition at the 21st China Conference on Computational Linguistics organized by Baidu, inviting participants worldwide to generate timestamped concise summaries of video segments, with registration details, eligibility, task description, example inputs/outputs, and evaluation metrics based on timing accuracy and ROUGE-L.
This article presents JD Technology's recent research on financial causal knowledge graphs, detailing the overall knowledge‑graph architecture, data layers, causal relation extraction, argument extraction, and graph‑alignment methods, and discusses their applications in finance, intelligent research reports, and industry‑leader recommendation.
FSL++—a RoBERTa‑large‑based few‑shot model enhanced with domain‑adaptive pre‑training, prompt learning, diverse embedding‑level augmentations, and ensemble self‑training—topped the Chinese FewCLUE benchmark, beating human accuracy on news and scientific classification tasks and delivering measurable gains across multiple Meituan product scenarios.
Alibaba’s advertising team introduces CREATER, a CTR‑driven text generator that leverages user reviews, aspect control codes, and contrastive fine‑tuning, and CONNA, a non‑autoregressive bundle creator that predicts heterogeneous ad elements with set‑based loss, both delivering substantial online CTR gains and CPC reductions through dynamic creative optimization.
This article explains how PaddleNLP's Universal Information Extraction (UIE) model can dramatically reduce labeling effort and improve accuracy for logistics parcel data extraction, showcasing a five‑sample experiment that boosts F1 by 18 points to 93% and providing a zero‑shot Python example.
The article analyzes the rising threat of open‑source components, explains Software Composition Analysis (SCA) and SBOM generation, outlines the three‑stage process for building an in‑house SCA capability, discusses practical challenges such as data quality and integration, and looks ahead to future standards and open‑source tools.
In a TVP forum hosted by Li Kaifu and Shen Chunhua, experts trace AI’s 70‑year journey, discuss the origins of the book “AI Future in Progress,” analyze investment stages, AI‑cloud synergy, NLP breakthroughs, medical applications, societal impacts, data privacy, and the challenges facing traditional enterprises.
This article explains an end‑to‑end semi‑supervised training pipeline for Transformer‑based NLP models, detailing the unsupervised language‑model pre‑training, supervised fine‑tuning, and the internal architecture of embeddings, encoder layers, and downstream tasks such as text classification and NER.
This article describes the motivation, micro‑service architecture, code implementation, V1.0 browser‑based NLP prototype, V2.0 AI‑enhanced version with BM25 and BERT, integration with ChatUI, DingTalk bot, command‑based automation, and future plans for the internal intelligent QA chatbot named Jarvis.
This article explains how JD's intelligent recommendation selling point system leverages NLP, BERT, Transformer and pointer‑generator models to automatically create short, personalized product highlights, describing the technical background, system architecture, model training pipeline, online/offline monitoring, and the resulting business benefits.
The article reviews the history and current state of NLP, compares symbolic deep‑parsing and neural pre‑trained approaches, discusses the knowledge‑bottleneck and low‑code trend, and illustrates semi‑automated, low‑code NLP deployment in the financial domain while pondering future integration of symbolic and neural methods.
This talk introduces graph deep learning techniques for natural language processing, covering the motivation for graph representations, traditional graph-based NLP methods, fundamentals of graph neural networks, static and dynamic graph construction, representation learning, and showcases the open‑source Graph4NLP Python library with example applications.
Tencent’s Hunyuan AI large model shattered records by scoring 80.888 to claim first place on the CLUE benchmark, showcasing its advanced natural language processing, multimodal abilities, curriculum‑learning training approach, and real‑world deployments in WeChat Search and advertising.
This article explains how TSDAE, a transformer‑based denoising auto‑encoder, converts noisy job description sentences into robust vector embeddings, details its training process, loss function, dataset preparation, and demonstrates using FAISS for similarity search on the resulting Job2Vec representations.
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.
This article proposes a dual‑mode detection solution for call‑recording quality inspection that combines rule‑based semantic similarity matching with BERT‑based sentence segmentation and RoBERTa multi‑label classification to achieve high accuracy, fast task adaptation, and strong generalization for customer‑service scenarios.
This article describes the end‑to‑end design, architecture, code implementation, and deployment steps for an internal intelligent QA chatbot named “Jarvis”, covering its V1.0 browser‑based prototype, V2.0 AI‑enhanced version, DingTalk integration, automation features, and future roadmap.
ArcCSE introduces an angular‑margin contrastive loss and both pairwise (dropout‑augmented) and triple‑wise (span‑masked) relationship modeling to self‑supervise text embeddings, yielding tighter decision boundaries, higher alignment and uniformity, and superior performance on unsupervised STS, SentEval, and Alibaba’s retrieval and recommendation systems.
This article surveys relation extraction, defining the task, categorizing its five main forms, and detailing key approaches such as entity position encoding, dependency‑tree methods like shortest dependency path and BRCNN, as well as distant supervision with multi‑instance learning and selective attention.
This article details the design, architecture, key technologies, and deployment of Xiaosi, an AI‑powered intelligent customer service chatbot built on the internal IM platform, highlighting its problem‑diagnosis accuracy, manpower savings, adapter mechanism, high‑availability backend, and practical use cases.
At SIGIR 2022, the authors present a constrained Seq2Tree model that transforms hierarchical label taxonomies into preorder sequences and applies dynamic‑dictionary decoding to ensure label consistency, achieving superior hierarchical text classification performance on benchmark datasets and real‑world deployment within Alibaba Entertainment’s AI Brain.
This article explores how to automatically discover new words in Chinese web novels by combining n‑gram statistics, pointwise mutual information, information entropy, and TF‑IDF filtering, presenting a practical, unsupervised pipeline that improves tokenization and search recall without manual labeling.
The 2021 Taobao Push project introduced an AI‑driven copy‑generation platform that combines template extraction and fine‑tuned Unilm models with diverse beam search, creating diverse, high‑quality push messages, cutting manual costs, and delivering a 10 % click‑through lift and higher material adoption.
This article explains how product researchers can tap into massive user comment data using NLP tools to identify pain points, sentiment, and opportunities, offering practical scenarios and tool recommendations for e‑commerce, offline retail, and app platforms.
The article explains how rapid internet growth created information overload, leading to concise summary services, and how recent AI advances—especially large language models like GPT‑3—enable platforms such as JD.com to automatically generate high‑quality, multimodal product copy that drives sales and supports diverse creative tasks.
The Tsinghua‑Meituan Digital Life Joint Research Institute’s Academic Salon on March 23 featured Associate Professor Liu Zhiyuan presenting the latest advances and ten key challenges in large‑model technologies, aiming to foster industry‑academia collaboration and drive innovation in representation learning, knowledge graphs, and social computing.
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.
This article introduces the open‑source WantWords reverse‑dictionary project, explains its token‑based processing pipeline, walks through Python implementation and model invocation with Hugging Face’s Transformers, reviews NLP’s historical evolution, and shows how front‑end developers can quickly integrate NLP models into products.
The article analyzes a recent Kaggle essay‑segmentation competition, explains why standard BERT‑based models plateau, and shows how a two‑stage pipeline that combines coarse BERT filtering with a feature‑rich tree model and post‑processing scaling can push scores well beyond the 70‑point barrier.
This article reviews the seven most notable Python libraries and packages of 2018 for data scientists and AI practitioners, including AdaNet, TPOT, SHAP, Optimus, spaCy, Jupytext, and Chartify, with descriptions, installation commands, and usage examples.
This article breaks down the evolution of Kaggle’s Jigsaw toxic comment competitions and presents a three‑step solution—training on historic data, using a genetic algorithm to weight multi‑label predictions, and ensembling fifteen models—to achieve high‑accuracy zero‑shot text classification.
This article traces the history of deep‑learning pre‑training techniques, comparing the parallel developments in natural‑language processing and computer vision—from early word2vec and bag‑of‑words models through ELMo and BERT to recent transformer‑based vision models like iGPT, ViT, BEiT and MAE—highlighting key innovations, challenges, and the convergence of the two fields.
This article presents a comprehensive overview of OPPO's XiaoBu Assistant, detailing its research background, chat skill architecture, evolution from retrieval and rule‑based methods to generative models, industry model comparisons, decoding and ranking strategies, safety mechanisms, performance optimizations, and evaluation results.
This article shows how the compress‑fasttext Python library can shrink a 7 GB fastText word‑embedding model to about 21 MB—a 300‑fold reduction—while preserving almost the same accuracy on downstream NLP tasks, and explains the underlying compression techniques, usage examples, and evaluation results.
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.
This guide introduces BertViz, an interactive visualization tool for transformer models such as BERT, GPT‑2 and T5, explains how to install it via pip along with required dependencies, and demonstrates head, model, and neuron view visualizations with code examples in Jupyter.
The Beike Real Estate NLP team secured first place in the CCIR Cup 2021 Intelligent Human‑Computer Interaction track by applying semi‑supervised and transfer learning techniques to small‑sample intent recognition and slot filling, and also presented the large‑scale Mandarin dialect speech benchmark KeSpeech at NeurIPS 2021.
JD Explore’s Vega v1 model topped the GLUE benchmark with a 91.3 average score, outperforming Microsoft, Facebook, and Stanford across multiple NLP tasks, including first‑ever human‑level results on sentiment analysis and coreference, showcasing JD’s leading position in deep‑learning research.
This article reviews Baidu's knowledge‑enhanced semantic understanding models, detailing the evolution from early semantic techniques to ERNIE 1.0, 2.0 and the large‑scale ERNIE 3.0, its architecture, training strategies, performance benchmarks, and real‑world applications across industry.
This article presents Ctrip's intelligent customer service system, detailing the evolution of semantic matching methods from traditional lexical models to deep learning approaches such as BERT and ESIM, and describing multi‑stage retrieval, multilingual transfer learning, and KBQA techniques for improving query understanding and response accuracy.
This tutorial walks through Chinese text tokenization with Jieba, optimizes the token list using stop‑words and part‑of‑speech filtering, visualises word frequencies, and applies SnowNLP to perform sentiment analysis on Weibo comments, complete with code examples and result charts.
This article traces the evolution of deep‑learning pre‑training techniques, starting with word2vec in NLP, moving through ELMo and BERT, then shifting to computer‑vision models such as iGPT, ViT, BEiT, and MAE, highlighting key innovations, challenges, and the convergence of NLP and CV paradigms.
This talk presents Alibaba DAMO Academy’s recent work on compressing large pretrained language models, covering task‑adaptive AdaBERT, data‑augmented L2A, and meta‑knowledge distillation Meta‑KD, describing their motivations, architectures, NAS‑based search, loss designs, and experimental results across multiple NLP tasks.
This article explains how to perform Chinese part-of-speech tagging using the jieba.posseg library in Python, including loading stop words, extracting article content via Newspaper3k, applying precise mode segmentation, filtering, and presenting results in a pandas DataFrame.
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.
This article details Laiye Technology’s AI R&D team’s end‑to‑end approach to designing and optimizing a custom entity‑relationship annotation tool, covering data‑labeling challenges, shortcomings of Excel and off‑the‑shelf solutions, architectural requirements, line‑breaking and mark‑position algorithms, performance improvements, and real‑world results.
To enhance decision‑making for users and quality monitoring for merchants, Meituan’s to‑restaurant platform implements fine‑grained aspect‑based sentiment analysis that extracts dish, attribute, opinion and polarity tuples from reviews, employing both a BERT‑CRF pipeline and a joint Dual‑MRC model which raise F1 scores from 0.61 to 0.68, and are deployed in dashboards and review‑management tools, with future work targeting efficiency and broader four‑tuple extraction.
This article walks through a complete Python workflow that loads the 20 Newsgroups dataset, preprocesses the documents, vectorizes them with TF‑IDF, groups them using KMeans, reduces dimensions with PCA, and visualizes the resulting clusters, illustrating each step with code and plots.
This article explains how Shannon's information entropy and NLP are combined in Alibaba Cloud's ARMS intelligent noise reduction to quantify alert uncertainty, filter redundant notifications, and automatically prioritize critical incidents, offering a practical, self‑learning solution for modern monitoring environments.
The presentation explains how JD.com leverages large‑scale text mining and NLP techniques—including data cleaning, multi‑level labeling, sentiment classification with models such as TextCNN, RoBERTa, and USE—to transform unstructured customer feedback into actionable product insights across various e‑commerce scenarios.
This article presents a step‑by‑step methodology for diagnosing and fixing model prediction errors—especially bad cases—in NLP and search systems, covering sample bias, threshold selection, preprocessing, post‑processing, validation cycles, and guidance on when to replace the model.
This article introduces several pandemic‑related open‑source resources—including a nationwide COVID‑19 vaccine record lookup, a tracker for ineffective vaccine distribution, and Baidu's DDParser NLP tool—detailing their purpose, usage, and installation to help developers build better vaccine‑related applications.
Baidu’s Fast Editing tool automates video trimming by using NLP to recognize subtitles, tone markers and duplicate sentences, then aligns them with the timeline for one‑click removal, employing character, Levenshtein and cosine similarity algorithms within a three‑module architecture (Plugin, Window, Caption) and planning on‑device PaddlePaddle analysis to cut latency and cost.
The KuaiSearch research team introduced PERKS, a large‑scale Chinese pre‑trained language model that achieved an 80.618 score on the CLUE 1.1 language classification task, narrowing the gap to human annotation and demonstrating significant advances in multi‑stage training, model optimization, and real‑world search applications.
This article introduces the importance of pre‑training in natural language processing, reviews classic pre‑training models such as Skip‑thoughts, BERT, GPT‑2 and T5, presents the modular UER‑py framework and its Chinese resources, compares it with Huggingface Transformers, and outlines practical deployment steps in industry.
This article presents a comprehensive overview of Xiaomi's Xiao AI intelligent QA system, detailing its background, three core answering modules—knowledge‑graph QA, retrieval‑based FAQ, and reading‑comprehension—and the underlying methods such as template matching, cross‑domain semantic parsing, path‑based reasoning, semantic retrieval, and neural matching, while also discussing performance results and practical trade‑offs.
This article explores how table‑based question answering can be integrated into B2B intelligent customer service by converting table queries into entity‑attribute recognition and multi‑turn dialogue, comparing end‑to‑end NL2SQL and slot‑filling approaches, and presenting NetEase Qiyu's practical implementation with its benefits and use cases.
Meituan’s knowledge‑based question answering system tackles diverse, constraint‑rich, multi‑hop queries across pre‑sale, in‑sale and post‑sale scenarios by integrating fine‑grained query understanding, relation recognition, sub‑graph retrieval and answer ranking, employing optimized BERT models, pre‑training tasks, and domain‑specific enhancements to boost response speed, conversion rates, and benchmark performance, while acknowledging remaining challenges in long‑tail and complex queries.
Meituan’s NLP Center introduced the ASAP dataset—the largest real‑world Chinese attribute‑level sentiment corpus—to date, and the article traces the progression from document‑level regression models upgraded with MT‑BERT, through multi‑task attribute‑level ABSA and opinion‑triplet extraction, to scalable real‑time and batch services, while outlining future transfer‑learning and few‑shot research.
This article presents a comprehensive overview of constructing an industry‑chain knowledge graph for the financial sector, covering its theoretical background, architectural design, automated building pipeline, key NLP techniques, and practical applications such as visualization, IPO review, and investment analysis.
This article explains how Yuedu's TTS synthesis platform tackles the booming audiobook market by using AI‑driven text preprocessing, role graph construction, content structuring, emotion and effect recognition, and a streamlined post‑processing workflow to efficiently generate multi‑character, emotionally rich audio books at scale.
The 9th CCF Big Data & Computing Intelligence Contest partners with Gaode Map to launch a POI Name Generation challenge, requiring participants to fuse image, signboard detection, and OCR data to automatically produce accurate, fluent place names, with a ¥50,000 prize pool, weekly vouchers, and recruitment opportunities for global teams.
This presentation by Tencent senior researcher Li Chao outlines the background, typical challenges, and multi‑layer technical solutions for intelligent grading in education, covering AI‑driven classroom, homework, review, and exam scenarios, multimodal spell‑checking, essay evaluation, and adaptive learning pipelines.
PaddleNLP v2.1 introduces an industrial‑grade Taskflow for eight NLP scenarios, a three‑line few‑shot learning paradigm that boosts small‑sample performance, and a FasterTransformer‑based inference engine that delivers up to 28‑fold speedup for text generation, all backed by extensive model and algorithm integrations.
This article presents a comprehensive overview of AI‑driven intelligent grading technologies, covering background, typical educational challenges, multimodal NLP solutions for essay, spelling and grammar correction, adaptive learning, and related research, illustrating how deep learning and multimodal models improve automated assessment across K‑12 scenarios.
This article presents a practical study of extracting phone numbers from two‑speaker voice dialogues using Seq2Seq models—including LSTM, GRU with attention and feature fusion, and Transformer—detailing data characteristics, model architectures, training strategies, experimental results, and comparative analysis showing the GRU‑Attention approach achieving the best performance.
This presentation reviews the importance of entity‑relation extraction for knowledge‑graph construction, compares sentence‑level and complex contexts, and surveys joint extraction techniques—including sequence labeling, table filling, and seq2seq models—as well as document‑level graph‑based methods and future research directions.
This article explains the fundamentals of contrastive learning, reviews typical models such as SimCLR, MoCo, SwAV, BYOL, SimSiam and Barlow Twins, and demonstrates how these methods are applied to Weibo text and multimodal (text‑image) representation tasks like hashtag generation and image‑text matching.
This article reviews the background and evolution of pre‑training models in NLP, introduces classic models such as Skip‑thoughts, BERT, and T5, and details the modular UER‑py framework, its comparison with HuggingFace Transformers, available Chinese pre‑trained weights, and practical deployment workflows.
This article details the authors' approach to the 2021 Sohu Text Matching competition, describing the task definition, data splits, model architectures (cross‑encoder and bi‑encoder), pretrained language models used, various training tricks, ensemble strategies, and the resulting evaluation scores.
The article explains how Ctrip’s hotel customer‑service team uses the Snorkel weak‑supervision framework to generate large‑scale labeled data for training models that automatically produce structured event summaries, detailing the workflow, labeling functions, generative and discriminative model training, and performance improvements.
This article recounts the author's six‑month Text2SQL research experience, detailing how systematic paper reading, leveraging existing engineering solutions, and fully utilizing academic, human, and hardware resources led to a successful thesis, a patent, a paper, and a second‑place finish in Yale's global Text2SQL competition.
Meituan’s NLP team leverages few‑shot learning—using data‑augmentation, semi‑supervised, ensemble/self‑training, and domain‑adaptation techniques—to cut annotation costs, achieving 1–2 percentage‑point accuracy gains on internal benchmarks and deploying high‑performing models for tasks such as topic classification, fake‑review detection, and sentiment analysis, while planning broader platform and model extensions.
Meituan’s 2021 ACL contributions comprise seven accepted papers—six long and one short—introducing novel approaches to event argument decoding, cross‑domain slot transfer, contrastive out‑of‑domain detection, novel slot discovery, self‑supervised sentence representation, unsupervised semantic parsing, and pseudo‑query‑enhanced dense retrieval, inviting further research and collaboration.
This article presents a comprehensive technical summary of the 2021 Sohu Campus Text Matching Algorithm Competition, detailing data characteristics, preprocessing strategies, tokenization choices, positional encoding methods, model architectures using relative encodings such as WoBERT and RoFormer, experimental results, and reflections on future improvements.
The article explains how Tencent addresses the limitations of traditional content understanding methods in personalized recommendation by introducing an interest‑graph framework that combines classification, concept, entity, and event layers, and details the associated mining, matching, and online evaluation techniques.
This article provides a comprehensive overview of natural language processing fundamentals, covering the challenges of language modeling, N‑gram and Markov assumptions, smoothing techniques such as discounting and add‑one, evaluation via perplexity, basic tasks like Chinese word segmentation, subword tokenization, POS tagging, syntactic and semantic parsing, and a range of downstream applications including information extraction, sentiment analysis, question answering, machine translation, and dialogue systems.
