that significant text doesn't work on nested objects. I have nested objects on es 6.0.0 and my significant text aggregation 'works' in that I get back words that make sense, but the bg_count figures are well in excess of my total document count. I've not deleted any documents, so I'm assuming that this bg_count is actually the count of the nested docs and so the scores and stats are incorrect (because nested objects aren't supported).
Can someone confirm whether my hypothesis holds weight ?
The bg_count lookup needs to be fast to eliminate many candidates. We use the cheapest way of doing this which is to use the pre-calculated counts held in the Lucene index for ranking purposes. These counts can include nested docs and deleted docs.
However, you can apply a custom background_filter which will take longer to compute but be more accurate.
Apache, Apache Lucene, Apache Hadoop, Hadoop, HDFS and the yellow elephant
logo are trademarks of the
Apache Software Foundation
in the United States and/or other countries.