[Bug]: The TEXT_MATCH method did not achieve the expected results as described in the documentation.

[cnzackliu]

Is there an existing issue for this?

• I have searched the existing issues

Environment

- Milvus version: 2.5.0
- Deployment mode(standalone or cluster): standalone
- MQ type(rocksmq, pulsar or kafka): default   
- SDK version(e.g. pymilvus v2.0.0rc2): 2.5.3
- OS(Ubuntu or CentOS): Rocky Linux
- CPU/Memory: 
- GPU: 
- Others:

Current Behavior

The code shown in the documentation

Use text match

The code I wrote

    for keyword in keywords:
        query_parts.append(f"TEXT_MATCH(text, '{keyword}')")
    filter = ' and '.join(query_parts)
    print(filter)
    query_vectors = model.encode([keywords])
    print(f'search {milvus_collection_name}')
    res = client.search(
        collection_name=milvus_collection_name,
        data=query_vectors,  # Query vectors
        anns_field='vector', # Vector field name
        # limit=10,  # number of returned entities
        filter=filter,
        output_fields=["text"],  # specifies fields to be returned
        search_params={'params':{'radius':0.8}}
    )

My code result

I want to search for documents that contain both "2栋-仓库213" and "乙方租客" simultaneously.
But I got some other content in addition to what I wanted when I searched（"11栋-108A"）.

Is this a bug? If not, how should I deal with it?

Expected Behavior

No response

Steps To Reproduce

Milvus Log

No response

Anything else?

No response

[yanliang567]

@cnzackliu I think this is expected as the search text "2栋-仓库213" will be parsed to ["2", "栋"， ”仓库“，”213“], so any text matched to either of the parsed text will be returned. I guess you want to do phrase match, which is in the roadmap of next release. please stay tune.

/assign @cnzackliu
/unassign

[cnzackliu]

@cnzackliu I think this is expected as the search text "2栋-仓库213" will be parsed to ["2", "栋"， ”仓库“，”213“], so any text matched to either of the parsed text will be returned. I guess you want to do phrase match, which is in the roadmap of next release. please stay tune.

/assign @cnzackliu /unassign

I have already processed the string and generated a new expression, which is the one on the second line of the picture： TEXT_MATCH(text, "2栋-仓库231") and TEXT_MATCH(text, "租客") and TEXT_MATCH(text, "乙方")，Will it still be split into ["2", "栋", "仓库", "213"] like this?