[LEGAL-276] SOLR-8542 contribution and potential patent concerns

• Type: Question
• Status: Closed
• Priority: Major
• Resolution: Not A Bug
• Assignee: Unassigned
• Reporter: Christine Poerschke
• Created: 10/Oct/16 14:38
• Updated: 20/Oct/16 01:06
• Resolved: 20/Oct/16 01:06

Description

(Had just posted this question to the legal-discuss@ mailing list and then discovered that there is this JIRA project also, it seems a more accessible alternative and so creating/repeating here.)

Hello Everyone!

The ticket and code:

The Learning-To-Rank (LTR) contribution https://issues.apache.org/jira/browse/SOLR-8542 includes an implementation of the LambdaMART algorithm (https://github.com/apache/lucene-solr/blob/jira/solr-8542/solr/contrib/ltr/src/java/org/apache/solr/ltr/model/LambdaMARTModel.java) and also an implementation of the RankSVM algorithm (https://github.com/apache/lucene-solr/blob/jira/solr-8542/solr/contrib/ltr/src/java/org/apache/solr/ltr/model/RankSVMModel.java) with both implementations extending the abstract LTRScoringModel class (https://github.com/apache/lucene-solr/blob/jira/solr-8542/solr/contrib/ltr/src/java/org/apache/solr/ltr/model/LTRScoringModel.java).

The RankSVM implementation class' javadoc provides a background reading link to the http://www.cs.cornell.edu/people/tj/publications/joachims_02c.pdf conference paper.

The LambdaMART implementation class' javadoc provides a background reading link to the http://research.microsoft.com/pubs/132652/MSR-TR-2010-82.pdf Microsoft Research Technical Report.

The contribution's README file (https://github.com/apache/lucene-solr/blob/jira/solr-8542/solr/contrib/ltr/README.md) suggests use of the liblinear library (https://www.csie.ntu.edu.tw/%7Ecjlin/liblinear/) for training of a machine learning model but the contribution itself is about use of a model and not about training a model.

The question:

It has been brought to my attention that there is https://www.google.com/patents/US20110208735 and also several other LTR patents held by Google, Microsoft, Yahoo, and others.

Has anyone come across a similar situation before? What are the recommended and/or required steps that should be taken before the contribution is committed to the project repo and included in an official release? (I am the committer person currently assigned to the SOLR-8542 ticket and have reached out to the project PMC also already.)

Thanks in advance for your input.

Regards,

Christine

Issue Links

• is related to: SOLR-8542 Integrate Learning to Rank into Solr