CIKM AnalytiCup 2017: Lazada Product Title Quality Challenge Forum

Go back to competition Back to thread list Post in this thread

> If your score is 3.141593 ...

... that means there is an error with your submission. You would notice that is the value of Pi. You would also realize that the RMSE cannot be more than 1. So this score value indicates an error.

If this happens to you, to get a clue of what might have gone wrong, please do the following:
-> Participate
-> Submit / View Results
-> Go to submission table
-> Go to + Sign next to a specific submission file
-> View scoring output log

I'm going to document various cases in this thread. In case you fall into one of the cases, then you would know how to fix the problem.

Posted by: hadylauw @ May 6, 2017, 12:18 a.m.

Case #1: Missing file

The scoring program expects two submission files: "clarity_valid.predict" and "conciseness_valid.predict" respectively. Both must be present in the zip file.

Posted by: hadylauw @ May 6, 2017, 12:20 a.m.

Case #2: Wrong filename

During the Phase 1, the two submission files must be named: "clarity_valid.predict" and "conciseness_valid.predict" respectively. Please use these exact names, without further extensions or modifications.

Posted by: hadylauw @ May 6, 2017, 12:21 a.m.

Case #3: Compression problem

The two submission files: "clarity_valid.predict" and "conciseness_valid.predict" should be compressed directly. The zip file name is not important. For example, we expect the following structure:
"submission.zip"
-> "clarity_valid.predict"
-> "conciseness_valid.predict"

DO NOT put the two files into a folder first, and then zip the folder.

DO select or highlight the two files, and then zip.

Posted by: hadylauw @ May 6, 2017, 12:23 a.m.

Case #4: Missing lines in a file

In each submission file, we expect the same number of lines as the number of lines in the feature file. There is one prediction of clarity and conciseness each for every line in the feature file.For Phase 1, data_valid.csv has 11838 lines. Therefore, "clarity_valid.predict" and "conciseness_valid.predict" must also have 11838 lines each.

Posted by: hadylauw @ May 6, 2017, 12:29 a.m.

Case #5: Illegal entries in the prediction file

Each prediction value is supposed to be a probability value, between 0 and 1 inclusive. It can be 0. It can be 1. It can be a value in between, such as 0.54321.

Your prediction file therefore should NOT contain the following values:
* something that is not a number
* a number less than 0
* a number more than 1

Posted by: hadylauw @ May 6, 2017, 12:33 a.m.

Case #6: Wrongly-formatted decimal values

Each prediction value is supposed to be a probability value, between 0 and 1 inclusive.

To write a decimal value, our convention is to use a dot or period. The following value is acceptable: 0.12345.

Some countries have a different convention of using comma separation for decimal values. We do not adopt this convention. The following is NOT acceptable: 0,12345.

Posted by: hadylauw @ May 6, 2017, 12:36 a.m.
Post in this thread