span_ruler is not working on ENT patterns

[kamlesh0606]

Text : "A 50-year-old male patient lodges chief complaint pain in lower right side of chest for three days associated with burning in epigastrium. History is cuffed with sputum."

Get Ent in Doc :
0 | 50-year | 2 | 5 | Age
0 | male | 7 | 8 | Sex
0 | pain in lower right side of chest | 12 | 19 | chief complaint
0 | three days | 20 | 22 | time_unit
0 | burning in epigastrium | 24 | 27 | chief complaint

Now ,

Add span_ruler in Pipe.
patterns = srsly.read_jsonl("./drugs_patterns_3.jsonl") span_ruler = nlp.add_pipe("span_ruler" , name= "span_ruler_3" , after="entity_ruler_2" , config= { } ) span_ruler.add_patterns(patterns)

drugs_patterns_3.jsonl File have 1 patterns ,
{"label":"cc_time_unit","pattern":[ {"ENT_TYPE": "cc"} , {"LOWER": "for"} , {"ENT_TYPE": "time_unit"} ]}

Output came :

But Output come as per logic ,
0 | pain in lower right side of chest | 12 | 19 | chief complaint
0 | three days | 20 | 22 | time_unit

How to reproduce the behaviour

Above Side Explain All Details. Still if any need , Please update to me.

Your Environment

• Operating System: Windows 11
• Python Version Used: 3.7
• spaCy Version Used: 3.6
• Environment Information:

[adrianeboyd]

Using token patterns, each {} pattern matches exactly one token by default, so for this pattern you end up with the matched token "chest":

{"ENT_TYPE": "cc"}

To match one or more tokens, you can use:

{"ENT_TYPE": "cc", "OP": "+"}

Docs on operators: https://spacy.io/usage/rule-based-matching#quantifiers

If you might ever have two adjacent entity spans (e.g. something like "following a week of [intense headaches] [chest pain] for three days") and you only want to match one of the spans, you'd need to also include the ENT_IOB features so you can match only one entity based on the IOB tags:

{"ENT_IOB": "B", "ENT_TYPE": "cc"}, {"ENT_IOB": "I", "ENT_TYPE": "cc", "OP": "?"}

Alternatively, you could merge all entities into a single token before running patterns if you only need the phrase (and not the tokenization) in your results: https://spacy.io/api/pipeline-functions#merge_entities. Then {"ENT_TYPE": "cc"} would match "pain in lower right side of chest".

[kamlesh0606]

@adrianeboyd Thanks For given Way.

Apply nlp.add_pipe("merge_entities" , after="ner") Pipe . And Merge All token of CC Ent.

Input Text : "A 50-year-old male patient lodges chief complaint pain in lower right side of chest , vomiting , high cholesterol for three days associated with burning in epigastrium for until chest pain is revealed. history of high blood pressure and high cholesterol 100 mg. "

Apply below pattern,
{"label":"1","pattern":[{"ENT_TYPE": "cc" } ,{ "LOWER" : "for" } , {"ENT_TYPE": "time_unit" } ]}

{"label":"1","pattern":[{"ENT_TYPE": "cc" } , {"ENT_TYPE": "cc" } ,{ "LOWER" : "for" } , {"ENT_TYPE": "time_unit" } ]}

And did get outcome.

Can Join 3 CC Ent join with Time_unit Ent ?

also tried 3 time pattern added then get below outcome,
{"label":"1","pattern":[{"ENT_TYPE": "cc" } ,{ "LOWER" : "for" } , {"ENT_TYPE": "time_unit" } ]}
{"label":"1.2","pattern":[{"ENT_TYPE": "cc" } , {"TEXT" : "," , "OP" : "?"} , {"ENT_TYPE": "cc" } ,{ "LOWER" : "for" } , {"ENT_TYPE": "time_unit" } ]}
{"label":"1.3","pattern":[{"ENT_TYPE": "cc" } , {"TEXT" : "," , "OP" : "?"} , {"ENT_TYPE": "cc" } , {"TEXT" : "," , "OP" : "?"} , {"ENT_TYPE": "cc" } ,{ "LOWER" : "for" } , {"ENT_TYPE": "time_unit" } ]}

But When get 5-6 CC Token then add 6-7 time patten. Same way other place comes Same points. Please provide good option if you have.

2nd points,
Start with "chief complaint" And End with ["." , "," ,"End of Statement"]. Please provide pattern of it if possible.

[kamlesh0606]

@adrianeboyd Thanks for given support.

Have you support groups for patterns parts ? if yes, I will take it and more discussion there.

One More Time Thanks For giving Correct Way for new Development.

[adrianeboyd]

There are some older issues related to this, but more for labeled groups than groups that can repeat with operators:

• spaCy Token Matcher does not support group capture #4642
• Sub-pattern labeling for pattern matcher #3275

The matcher alignments provide a kind of workaround for labeled groups, but it's still not exactly groups.

[kamlesh0606]

@adrianeboyd , Thanks for given update.

Can get Pipe By Pipe Doc Object ? Or can get process ID in side Pipe ? if yes then will manage Pipe Doc Output Obj using process ID .

I get one way Like , "user_data" use of Doc Object. it it right way ?

[adrianeboyd]

Sorry, I'm not sure exactly what you're asking, and it sounds like it might be a separate question from this thread.

Could you open a new discussion with a code example that shows what you're trying to do?

[kamlesh0606]

ok

@adrianeboyd , #12986 new Points added.