Hello, I am new to ElasticSearch
I am following this tutorial trying to implement Semantic Textual Similarity between a query and set of questions/documents in the Stackoverflow dataset.
I found that Pearson/Spearman correlation are not pre-implemented in painless functions of ES,
- How can I implement them in a custom query script? I am looking for guidance and/or code snippets
- How can I sort the results
desc? or are they sorted
descby default, I noticed they were already sorted
- This is a general STS question, why most of the research paper showing their results in Pearson & Spearman Correlation, sometimes Cosine Similarity, while most of the online tutorials/materials have scores reported in Cosine Similarity? Does this mean for STS cosine similarity is better than the others?