In this competition, you who can understand the topic have great potential

International Annual top-level conference in the field of artificial intelligence NeurIPS, And ICML It is also called the two most difficult fields in artificial intelligence , The highest level , The most influential meeting , Every year, it attracts a large number of top leaders in industry and academia AI Experts and doctoral students participate .

This year, , Huawei cloud Vancouver big data and intelligent platform laboratory and two top universities in Canada （University of British Columbia and Western University） Together , In artificial intelligence AI International Summit NeurIPS2022 Jointly held The Natural Language for Optimization (NL4Opt) competition .

Mathematical modeling is the key to solve practical application problems by using mathematical optimization solver , This time NL4Opt Competition launched The first data set for mathematical modeling based on natural language , We are committed to reducing the threshold of using mathematical optimization solvers , Promote the research of mathematical optimization modeling using natural language technology .

At present , The use of industry solvers requires professional operations research optimization engineers to deeply communicate business problems with domain experts , Mathematical modeling , This tedious process greatly limits the scope of application of the general tool mathematical optimization solver . With the rapid development of natural language processing technology , Automatic programming through natural language 、 The technology of solving mathematical application problems has become a research hotspot , These related technologies provide the possibility for mathematical modeling of natural language processing technology . The research on mathematical optimization modeling of natural language technology is expected to become another important application breakthrough of natural language processing technology .

However, the research of this technology requires both

Deep understanding of natural language processing technology

You also need to have professional knowledge of mathematical optimization

If you can challenge the following questions

congratulations ！ friend , You have a lot to do ！

Competition questions

This competition includes Named entity recognition And Mathematical model generation Two tracks , Contestants can choose any track to submit a plan .

▌ Mission 1— Named entity recognition

This task aims to detect and mark entities in mathematical optimization problem description , Such as target name 、 Decision variable names or constraints , To help build a more accurate mathematical model . Complete the task , It will effectively identify the key information related to mathematical optimization problems in natural language .

Pictured above is an example , English problem description , Contains the following key information ：

• The customer's budget in investment is 60,000 element
• Customers want to invest in trust and savings, two financial products with different returns , The rewards are 7% and 2%
• Customers have their own investment constraints on the two financial products
• The customer hopes to find the maximum return investment plan that meets the above constraints

Task 1 hopes to use the named entity recognition technology in natural language processing , Get the above information accurately .

▌ Mission 2— Mathematical model generation

The goal of this task is to describe the mathematical optimization problem 、 Marked entity （ Such as target name 、 Decision variable names or constraints ） As input , Generate a mathematical model that can be solved by the mathematical optimization solver . This task is based on the first task , Learn the relationship between the key information of mathematical optimization problems through natural language to complete mathematical modeling , So that the model can call the mathematical solver to solve .

Pictured above is an example , When the key information in task 1 is extracted , If you want to solve with a mathematical solver , We need to further clarify the logical relationship between these information , And mathematical modeling （ As shown in the red box ）. This also makes task two more challenging .

Schedule

The competition time is 7 month 1 Japan —10 month 15 Japan , The application was submitted on 7 month 15 Open day .

• 07-01： Competition start , Contestants register
• 07-15： Submit scheme , Leaderboards and forums are open
• 10-08： Competition registration deadline
• 10-15： Proposal submission deadline
• 10-31： The results are announced

prize

Take each of the two tracks Top five The contestant who submitted the best plan ：

ranking	Bonus
The first name	6,000 dollar
proxime accessit	3,000 dollar
The third	1,000 dollar
A fourth	500 dollar
fifth	500 dollar

After the competition , Depending on the NeurIPS2022 Hold seminars , We will invite the winners to share the solutions at the meeting , And have in-depth exchanges with authoritative experts in the field of natural language processing and operations research .

NL4Opt, I look forward to your participation ！

More details about the competition 、 Schedule

Qualifications 、 Rules and other information

Scan the code to learn more ~~