| | |
|---|
Welcome and agenda | @Former user (Deleted) | Please note that the SLPG will be meeting in London on Sunday 5th April (9am to 12:30pm) - see schedule |
Concrete values | @Former user (Deleted) | SCG, ECL, STS, ETL - Ready for publication - on hold until after MAG meeting in April confirming requirement for Boolean datatype. |
URIs | @Peter Williams & @Former user (Deleted) | Please review updates to the URI specification, and raise any comments in the next 2 weeks. Draft URI standard for review - URI Standard 2.6 URIs for Language Syntaxes 2.7 URIs for Language Instances 2.8 URIs for Modelling Resources 3.1 Resolving SNOMED CT URIs
|
Expression Constraint Language | @Former user (Deleted) | NEXT STEP FOR ECL: {{ term = [ termSearchType : ] "String", language = <langCode> }} Example - {{ term = "heart att", language = es }} Question - Do we want to reconsider including optional parameters for 'type', 'dialect' and 'acceptability' typeId = 900000000000013009 ; type = <synonym | fsn> dialectId = 900000000000508004 ; dialect = <en-GB | en-AU | en-Patient | de-CardioSpecialist> acceptabilityId = 900000000000549004 ; acceptability = <acceptable | preferred >
Term Search Type Wild Card Match (collation) - e.g.
Word Prefix Any Order - e.g.
Default (word prefix any order) - e.g.
{{ term = "hear att" }} {{ term = "*heart*“ }}
Potential Examples << 64572001 |Disease| {{ term = “heart”}}
<< 64572001 |Disease| {{ term = “heart”, language = "en"}} << 64572001 |Disease| {{ term = “heart”, language = "en"}} AND << 64572001 |Disease| {{ term = “hjärta”, language = "sv"}} << 64572001 |Disease| {{ term = “heart”, language = "en"}} {{ term = “hjärta”, language = "sv"}}
<< 64572001 |Disease| {{ term = “heart”, language = "en"}} OR << 64572001 |Disease| {{ term = “hjärta", language = "sv"}}
<< 64572001 |Disease| {{ (term = “heart”, language = "en") OR (term = “hjärta", language = "sv")}} (<< 64572001 |Disease|: |Associated morphology| = *) {{ term = “heart”, language = "en", }} {{ term = “hjärta", language = "sv"}} (<< 64572001 |Disease| {{ term = “*cardio*” }}) MINUS (<< 64572001 |Disease| {{ term != “*heart*” }}) Recommendation to be made on (based on investigation of grammar): << 64572001 |Disease| {{ term = “heart”, language = "en"}} AND {{ term = “hjärta”, language = "sv"}}
<< 64572001 |Disease| ( {{ term = “heart”, language = "en"}} OR {{ term = “hjärta”, language = "sv"}} )
<< 64572001 |Disease| ( {{ term = “heart”, language = "en"}} MINUS {{ term = “hjärta”, language = "sv"}} )
Use Cases Intentionally define a reference set for chronic disease. Starting point was ECL with modelling; This misses concepts modelled using the pattern you would expect. So important in building out that reference set. Authors quality assuring names of concepts Checking translations, retranslating. Queries for a concept that has one word in Swedish, another word in English AU use case would have at most 3 or 4 words in match Consistency of implementation in different terminology services Authoring use cases currently supported by description templates A set of the "*ectomy"s and "*itis"s
Questions Do we include 'typeId' - e.g. << 64572001 |Disease| {{ D.term = “*heart*”, typeId = 900000000000013009 |Synonym| }} Do we include 'type' - e.g. << 64572001 |Disease| {{ D.term = “*heart*”, D.type = synonym }} Do we include 'languageCode' - e.g. << 64572001 |Disease| {{ D.term = “*heart*”, D.type = synonym, D.languageCode = “en” }} Do we include 'caseSignificanceId' - e.g. << 64572001 |Disease| {{ D.term = “*Heart*”, D.caseSignificanceId = 900000000000017005 |case sensitive|}}
Do we include 'caseSignificance' - e.g. << 64572001 |Disease| {{ D.term = “*Heart*”, D.caseSignificance = sensitive }} Do we include 'language' and 'version' - e.g. << 64572001 |Disease| {{ term = “*heart*” }} VERSION = http://…, LANGUAGE = (999001881000000108|Gastro LRS|, |GB English|) Do we include syntactic sugar - e.g. << 64572001 |Disease| {{ preferredTerm = “*heart*”, languageRefSet = en-gb}} << 64572001 |Disease| {{ fullySpecifiedTerm = “*heart*”, languageRefSet=en-gb}}
<< 64572001 |Disease| {{ acceptableTerm = “*heart*”, languageRefSet = en-gb}}
<< 64572001 |Disease| {{ preferredTerm = “*heart*”}} FROM version = X, language = Y NO
Do we use/require the "D" at the start of "term"?
Packaging - How do we package this extension to ECL
|
Querying Refset Attributes | @Former user (Deleted) | Proposed syntax to support querying and return of alternative refset attributes (To be included in the SNOMED Query Language) |
Returning Attributes | @michael lawley | Proposal (by Michael) for discussion For example, I can write: << 404684003|Clinical finding| : 363698007|Finding site| = <<66019005|Limb structure| << 404684003|Clinical finding| . 363698007|Finding site| But I can't get all the attribute names that are used by << 404684003|Clinical finding| |
Reverse Member Of | @michael lawley | Proposal for discussion What refsets is a given concept (e.g. 421235005 |Structure of femur|) a member of? |
Expression Templates | @Peter Williams | Examples: [[+id]]: [[1..*] @my_group sameValue(morphology)] { |Finding site| = [[ +id (<<123037004 |Body structure (body structure)| MINUS << $site[! SELF ] ) @site ]] , |Associated morphology| = [[ +id @my_morphology ]]} Note that QI Project is coming from a radically different use case. Instead of filling template slots, we're looking at existing content and asking "exactly how does this concept fail to comply to this template?" For discussion: Is it correct to say either one of the cardinality blocks is redundant? What are the implications of 1..1 on either side? This is less obvious for the self grouped case. Road Forward for SIGenerate the parser from the ABNF and implement in the Template Service User Interface to a) allow users to specify template at runtime b) tabular (auto-completion) lookup → STL Template Service to allow multiple templates to be specified for alignment check (aligns to none-off) Output must clearly indicate exactly what feature of concept caused misalignment, and what condition was not met.
Additional note: QI project is no longer working in subhierarchies. Every 'set' of concepts is selected via ECL. In fact most reports should now move to this way of working since a subhierarchy is the trivial case. For a given template, we additionally specify the "domain" to which it should be applied via ECL. This is much more specific than using the focus concept which is usually the PPP eg Disease. FYI @Michael Chu |
Description Templates | @Kai Kewley | |
Query Language
- Summary from previous meetings | @Former user (Deleted) | FUTURE WORK Examples: version and dialect Notes Allow nested where, version, language Scope of variables is inner query
|
| Examples: where Notes Allow nested variable definitions, but recommend that people don't due to readability Scope of variables is the inner query No recursion e.g X WHERE X = 1234 MINUS X ie can't use a variable in its own definition ie X is only known on the left of the corresponding WHERE, and not on the right of the WHERE
|
Keywords for Term-based searching: D.term D.term = "*heart*" D.term = wild:"*heart*" D.term = regex:".*heart.*" D.term = match:"hear att" D.term = (sv) wild: "*heart*"
D.languageCode D.languageCode = "en" D.languageCode = "es"
D.caseSignificanceId D.caseSignificanceId = 900000000000448009 |entire term case insensitive| D.caseSignificanceId = 900000000000017005 |entire term case sensitive| D.caseSignificanceId = 900000000000020002 |only initial character case insensitive|
D.caseSignificance D.caseSignificance = "insensitive" D.caseSignificance = "sensitive" D.caseSignificance = "initialCharInsensitive"
D.typeId D.typeId = 900000000000003001 |fully specified name| D.typeId = 900000000000013009 |synonym| D.typeId = 900000000000550004 |definition|
D.type D.acceptabilityId D.acceptability
Additional Syntactic Sugar FSN synonym synonymOrFSN textDefinition Unacceptable Terms
|
Language preferences using multiple language reference sets LRSs that use the same Language tend to use 'Addition' - i.e. child LRS only includes additional acceptable terms, but can override the preferred term E.g. Regional LRS that adds local dialect to a National LRS E.g. Specialty-specific LRS E.g. Irish LRS that adds local preferences to the en-GB LRS
LRSs that define a translation to a different language tend to use 'Replacement' - i.e. child LRS replaces set of acceptable and preferred terms for any associated concept
|
Confirm next meeting date/time | @Former user (Deleted) | Next meeting is scheduled for Wednesday 11th March 2020 at 20:00 UTC. |
A few thoughts on item 4 – 'search'...
I’m sure it’s achievable (and desirable) to introduce a ‘minimal’ search filter to the ECL whilst deferring the ability to specify various parameters. However in order to maximise the chance of a search filter producing the ‘intended’ results wherever it is used, it looks as though a number of (processing) assumptions need to be declared rather than left implicit for each implementation. Eventually many of these default ‘assumptions’ could themselves become configurable.
To start, and in a slightly different spirit, looking at the search string candidate syntax itself:
I presume the existing stringValue = 1*(anyNonEscapedChar / escapedChar) would be used to specify a search string.
The current definition of escapedChar would allow quotation marks to be used in a search and thus distinguish between "ghosts" and ghosts:
270072003 | Patient registration-"ghosts" (finding)
247871001 | Fear of ghosts (finding)
However escapedChar would also need to be extended to include an escape for the asterisk in order to test for terms containing *, e.g.:
738783002 | Human leukocyte antigen A*31:01 positive (finding)
Regarding a default ‘description substrate’: this seems to include some inevitable assumptions. At least one language reference set (LRS) is ‘essential’ [1], [2] in any SNOMED CT build, and thus in any default ECL substrate (“the set of active components from the snapshot release (in distribution normal form) of the SNOMED CT versioned edition currently loaded into the given tool” [3]). However if the ability to specify the actual LRS (or set of LRSs) is deferred, then only specifying the languageCode could lead to inconsistent results in implementations that use the same Descriptions table but filter it by different LRSs. Clearly this is better than nothing, but the potential for variation needs to be articulated.
Tokenization.
I know we touched on this on the call - I feel it’s significant. I am unaware of any truly ‘standard’ tokenizer for indexing descriptions – if somebody knows of a suitable one then this specification should cite it as the assumed position. Anne and Fadi mention a set of candidate separating characters in the Search and data entry guide [4], specifically item 4 here [5], and I suspect that we all use something like this.
I know I’m personally guilty of tokenising differently on different occasions. My RF1-based implementation uses " ""'.,:;!?()[]{}<>^-+/#~*&%$£", more recently I see that I used ".,/=-+><^?@~{}[]'\"*():;#" in an unrelated project, and my basic RF2 snapshot build uses Sqlite’s FTS3/4 simple tokenizer [6]. All have behaved ‘well enough’ but I suspect they are sufficiently different for the occasional variation in search/matching behaviour to arise.
Building on Daniel’s comment on the call that many non-English languages that use compound words perform poorly with a ‘word prefix’ search convention… There is a glimpse of this also in English terms with their idiosyncratic use of hyphens or compounding, and consequently some implementations treat hyphens as special case token separators, concatenating rather than splitting tokens to try and ‘normalise’ these variations, thus indexing these two concepts to appear ‘the same’ for search purposes:
85722001 | Pelvi-ureteric contraction (finding)
236711008 | Calculus in pelviureteric junction (disorder)
Likewise, some implementations treat periods/full stops differently if they seem to be separating the letters of an acronym, and for the following examples would give both a key of LE rather than the second one L and E:
167901007 | Synovial fluid: LE cells |
37954002 | L.E. cell preparation |
I'm not saying that either variant would or should be specified in a 'default' tokenizer, just that some implementations may already use them and may need to be aware that they are not 'assumed' in the creation of the search string.
Normalisation.
Here I confess I get simultaneously out of my depth, English-centric and ‘my search use case’-centric. For me, search normalisation has predominantly about making the token and candidate text as much like basic ASCII characters (and thus my keyboard) set to minimise false negatives in search (esp. diacritic removal & case-independence).
However I recognise that non-English languages will have different requirements, and indeed there are times when I want to retain diacritic or case sensitivity.
What I feel I'm bumping up against is a tension between my use case and published language-sensitive normalisation options to present as a 'default'. W3C’s string matching WG Note [7] and its referenced Unicode standards comes close, but I’m scratching my head to understand if any of the ‘appropriate normalization step’ choices at [8] or referenced from the Unicode collation document [9] correspond to my (I hope not unreasonable) desire for a search based on ‘match:“SJOGR SYN” language=en’ to include 762303003 | Pediatric onset Sjögren syndrome (disorder) in its return set, even though all ‘en’ terms associated with this concept include the accented Sjögren form. Maybe it is unreasonable and I (and other 'en' users) will instead have to specify all diacritic-sensitive search variants - and therefore need to know them in advance!
Clearly there are plenty of other things to consider regarding ‘search’, but the points above seemed like a reasonable set to kick off a discussion.
Kind regards
Ed
[1] https://snomed.atlassian.net/wiki/display/DOCEXTPG/4.3.2.4.1+Language+Reference+Set
[2] https://snomed.atlassian.net/wiki/display/DOCTSG/4.2+Essential+Reference+Sets
[3] https://snomed.atlassian.net/wiki/display/DOCECL/3.2+Expression+Constraint+and+Query+Requirements
[4] https://snomed.atlassian.net/wiki/display/DOCSEARCH/Search+and+Data+Entry+Guide
[5] https://snomed.atlassian.net/wiki/display/DOCSEARCH/Appendix+B+-+Future+Additions+to+this+Guide
[6] https://www.sqlite.org/fts3.html#tokenizer
[7] http://www.w3.org/TR/charmod-norm/
[8] http://www.w3.org/TR/charmod-norm/#performNorm
[9] http://www.unicode.org/reports/tr10/#Step_1