1. Home
  2. Services
  3. IP Data Intelligence
  4. IP Expert AI

IP Expert AI

Overwhelming AI innovating IP ecosystem
It provides AI service specialized in IP data, from data planning to labeling and model building.

The only AI in Korea completely specializes in IP data is Wert Intelligence's IP Expert AI.

An AI leader revolutionizing the IP ecosystem

If you are considering AI optimized for IP data, choose IP Expert AI from Wert Intelligence.
The first
By launching the world’s first AI patent language model 'PAT-BERT', it has verified AI technology optimized for IP data.
The largest
By utilizing the largest global patent data in Korea, containing over 300 million cases, it can create an AI model optimized for the IP.
The best
The best experts in Korea, including patent data engineers, AI data scientists, and patent attorneys, participate in the design and construction.

AI optimized for IP data

It fully comprehends the complicated business requirements based on IP data and delivers the best results.

Data building

The whole process of learning language models (LLM, BERT, etc.) specialized for IP data, from planning, data collection, and preprocessing to labeling, processing, and verification, is carried out by experts.

Building a model

It can increase the usability of IP data by designing and training AI models that are suitable for your customer’s needs and building customized AI models specialized for IP data.
IP Expert AI

Verification and evaluation

AI model optimized to IP data is thoroughly verified in the real-use environment, enabling reliability and performance evaluation of the model. Then, it determines whether the constructed AI model can solve business problems from an IP perspective.

Operations and maintenance

It additionally learns the latest IP data to keep up with the changing business environment and supports stable operation and continuous model maintenance for task processing that meets customers' needs.

High-quality IP learning data that determines AI performance

We provide high-quality IP training data to enable IP-based AI models that handle various tasks to perform at their best.

300 million+

Number of patent data
held

6 billion+

Number of patent images
held

300 billion+

Number of AI model parameters

Data planning

It models and plans to learn data according to the purpose of IP-based AI modeling, utilizing the application, operation environment, end-user, and service type.

Data labeling

It ensures accuracy and efficiency at a high level by collecting and preprocessing raw data in order to build IP training data set, and can be used for various purposes, including Pre-Training.

Data preprocessing

It can process high-quality answer sets for the tasks(classification, summary, generation, Q/A, etc.) supported by IP-based AI models and can be utilized for fine-tuning LLM.

Data validation

In addition to data construction, it ensures the reliability of the data, repeatedly checks the semantic accuracy to meet the quality standards of the desired AI model, and considers diversity to ensure that the data is not biased toward specific matters.

Building and maintaining perfect AI models

We carry out all processes flawlessly so that the AI model can establish itself as a core asset of the business.

Build

Select the optimal TASK model
Test and select the model
Utilize open source models (BERT, LLM)
Direct parameter optimization

Model training and building

Identify customer needs
Build learning data reflecting technology/TASK
Model pre-training
Model fine-tuning
Check training environment information
Check available licenses

Model validation and evaluation

Validate test dataset
Judge the TASK appropriateness of the learning model
Set model-specific validity judgment standard
Human verification of model inference results
Verify performance indicators (F1-score, PPL, BLEU, ROUGE, etc.)

Model serving and maintenance

Establish customer assessment criteria according to TASK
Apply sLLM service based on RAG
AI model training and updating
Service maintenance
Establish a monitoring system
Reflect user feedback

Experience new innovations with IP Expert AI

From patent summaries to technology predictions, the IP Expert AI has limitless applications.

Company N

We supplied AI learning data, '560,000 industrial classification, 300,000 science technology standard classifications, 620,000 national key technology classifications, and 200,000 LLMs in civil law/intellectual property law' to the N hub.

Company I

We built a RAG-based sLLM utilizing IP news articles and report data for I, a media specializing in intellectual property.

Company IN

We supplied purpose-solution extraction and summary AI models to IN Corporation, a leading smart device manufacturer in Taiwan. The company's IP team and R&D team were able to greatly reduce the time taken for patent review and analysis through the summary model.

keywert

We provided AI technology classification model, which can classify tens of thousands of patents within seconds by recognizing the user's purpose, to Keywert.
It is equipped with multiple AI classification models covering various classification standards and can identify more than 100 million classification cases.
Experience the new future of business brought to you by AI, right now.