首页 Natural Language Processing NLP self-learning platform User Guide Pre-trained models (platform-provided models)

Pre-trained models (platform-provided models)

更新时间: 2026-01-15 16:23:09

Pre-trained models are ready-to-use models provided by the platform. You can call them directly using API operations.

On March 24, 2023, the pre-trained models on the NLP Self-learning Platform were upgraded, categorized by industry, and fully migrated to Natural Language Processing (NLP). The NLP Self-learning Platform provides methods and interfaces for model invocation. For more information about the API operations, see the following documents.

Service invocation

For more information about how to call a model, see Model invocation.

For software development kit (SDK) examples, see SDK examples.

API operation list

E-commerce industry capabilities

Model name

Model description

Default QPS

Maximum text length

Product review parsing for e-commerce

Analyzes product review text from multiple dimensions. This model supports 55 industries in the e-commerce realm and 192 review properties.

10

500 characters

Product review parsing for local services

Analyzes product review text from multiple dimensions. This model supports two local service industries (beauty/hair/nails and food/dining) and 11 review properties.

10

500 characters

Product review parsing for the automotive realm

Analyzes product review text from multiple dimensions. This model supports 68 review properties in the automotive realm.

10

500 characters

Purchase decision analysis from product reviews for e-commerce

Analyzes purchase decision information, such as user motivations, application scenarios, feature requirements, and questions. Use this model to improve products, enhance user experience, segment user profiles, and run targeted marketing campaigns.

10

500 characters

Purchase decision analysis from product reviews for the automotive realm

Analyzes purchase decision information, such as user motivations, application scenarios, feature requirements, and questions. Use this model to improve products, enhance user experience, segment user profiles, and run targeted marketing campaigns.

10

500 characters

Online customer service scenario parsing

Parses consumer conversations in online chat scenarios for e-commerce and other industries to determine consumer intent, sentiment, and emotion.

10

/

Dialogue knowledge extraction

Extracts customer service scripts and user questions from online chat conversations. Use this model to analyze hot spot issues or build a script library.

10

/

Product description generation (Chinese)

Generates product descriptions related to specified selling points for a given product.

10

/

General industry capabilities

Model name

Model description

Default QPS

Maximum text length

Bidding announcement classification service

Classifies bidding announcements. This model currently supports two types: "tender invitation" and "bid award".

10

/

Bidding information extraction - Basic Plan

Extracts 13 fields from bidding information, such as project name, project number, bidder name, and winning bid amount.

10

/

Tender information extraction - Pro

Parses tender invitations and bid awards separately. Extracts 22 fields from tender information.

10

/

Bid award information extraction - Pro

Parses tender invitations and bid awards separately. Extracts 29 fields from bid award information.

10

/

Contract element extraction - General realm

Extracts common elements from contracts. This model supports 26 general element fields.

10

/

Document structuring - Key-value information extraction

Extracts information that follows a key-value pattern from documents such as resumes, contracts, and reports.

10

/

Telemarketing dialogue - Industry classification

Classifies telemarketing outbound call dialogues by industry and scenario for applications such as voice quality inspection. This model supports over 30 industries and over 170 scenarios.

10

/

Profanity detection service

Supports application scenarios such as customer service quality inspection for telemarketing dialogues and streamer monitoring for live streaming.

10

/

Telemarketing dialogue - User intent recognition

Recognizes user intent (reaction) in manual or intelligent telemarketing outbound call scenarios.

10

/

Telemarketing dialogue - Suspected fraud detection

Identifies conversations with fraud risks in telemarketing outbound call scenarios. You can use this for voice quality inspection.

10

/

Resume extraction - English

Extracts 10 resume fields from English resumes, such as name, contact information, degree, company, and position.

10

/

Resume extraction - Chinese

Extracts 33 resume fields from Chinese resumes, such as name, gender, age, education, and employer.

10

/

Event extraction (English)

Extracts events from English news articles. This model includes 33 event categories.

10

/

Product title category prediction

Predicts the category of a product based on its title in e-commerce scenarios. The category system is consistent with e-commerce platforms such as Taobao.

10

/

Pornography detection service for novels

Detects pornographic or erotic content in Chinese novels for content moderation scenarios. The model outputs a confidence level for pornography and the related text.

10

600 characters

Sentiment analysis (Russian)

Predicts the sentiment of Russian text from social media (short text) in e-commerce scenarios. Sentiments are classified as positive, neutral, or negative.

10

/

Sentiment analysis (English)

Predicts the sentiment of English text from social media (short text) in e-commerce scenarios. Sentiments are classified as positive, neutral, or negative.

10

/

Sentiment analysis (Spanish)

Predicts the sentiment of Spanish text from social media (short text) in e-commerce scenarios. Sentiments are classified as positive, neutral, or negative.

10

/

Emotion analysis service

Recognizes customer or agent emotions in application scenarios such as telemarketing and online support. This model supports eight common emotions and three business-specific emotions.

10

1000 characters

News text classification

Classifies one or more news texts.

10

/

Live streaming ASR garbled text detection

Detects poorly readable text caused by multiple people speaking at the same time in live streaming scenarios where speech is converted to text using Automatic Speech Recognition (ASR).

10

600 characters

Legal judgment document extraction

Parses documents for 10 causes of action and extracts 38 fields.

10

/

Keyword extraction and text summarization (extractive)

Extracts keywords or summaries from documents.

10

500 characters

Text summarization (generative)

Generates text summaries or article titles. This model is designed for common text generation needs in real-world scenarios.

10

500 characters

Weather report welcome message generation (Chinese)

Generates an in-vehicle startup welcome message based on given weather information fields.

10

500 characters

Text embedding generation

Takes Chinese text as input and outputs the corresponding vector representation.

10

/

Sarcasm detection

Identifies irony and sarcasm in social media content.

10

/

Humor classification

Analyzes humorous content to predict its type (such as pun, convention, or inversion).

10

/

上一篇: Resume extraction 下一篇: Developer Reference
阿里云首页 自然语言处理 相关技术圈