Semantic search across multiple fields in the index

Elastic Stack Elasticsearch
1

I am reviewing the tutorial here to setup semantic search on my index. The tutorial helps to do this on one field. Lets say i want to do the semantic search across multiple fields, how do I achieve that? For example, I have this data:

Employee name: sally, Age: 35yrs, Education: MS, description: Sally is an experienced product manager …, Location: San Diego

I want to do NLP search on this index and get answers to questions like : Are there any young employees in West coast?

If it was just description field, we could have added the semantic_Search on that field. But this is across multiple fields in the index and with context.

Any help is appreciated.

Thanks,

1 Like
2

Hi @SunnyMentions , welcome!

Initially, I will define a new field with values for age, country, and description to be used for semantic search. In this new field, I will enrich the data to generate something like this: 'Sally is an experienced product manager; she is 35 years old and lives in San Diego.'

I’m not sure if this is the best approach, Initially, I will define a new field with values for age, country, and description to be used for semantic search. In this new field, I will enrich the data to generate something like this: 'Sally is an experienced product manager; she is 35 years old and lives in San Diego.'

I’m not sure if this is the best approach, but I would start with it and see the results.

3

i am having similar use case and i implemented a similar kind of approach where i combined all fields into single field and created a parsed vector.
when writing search query i combined sparse vector search with multi search with different fields having thier own boosting and working quite well.
But now words like stopwords such as 'in' 'and' are highly affecting my search result counts and i want to resctrict search results to only relevent results