ucsfwk
[I used to fill out(useful)reports w/: done this wk, planned for next, road-blocks/things needed from others]
I'll be updating this soon, now that I know that it is being looked at. Also: questions &more_detail_(w/day to day links & commentary). Many of these links point to *these directories*(on the local os-mac).
Not enough summary at the start of the year, then too much at once, that I have to go back over:Â
NO MORE POSTS HERE -GOTO NEW WEBSITE!!
By end of month I will switch this page over to: http://domsetup.ucsf.edu:8080/mb/wk.html (complete by March 1st) then no more posting here at all
I was happy w/the call, it is good to see the use of SAGE, and pulling in some of the sumo and other ideas in as well; More in the daily log.
Started to check cl-http install and made some progress w/it; Still waiting on SRI/Dr. Karp's libs;Â To look at Mor's latest file, you can paste this link into the open-URL field of Protege3.4b.
Gave Mor link to the protege-server that I originally had on the win-xp box &only recently got up again on the g4/lnx;Have umlsupalong w/a few other pprjsyou should be able to get to. [Also, all the links on this/these pages will be open.]
Re:Sysadmin: I still feel we've done most of what we need right now & still haven't gotten the one thing that I wanted (a network that lets rmi/etc through, which I can do in OS-X w/o problem); Further, to have us learn how to tighten down the hatches & then start paying for sysadmin, seams like another 'we waited to long, &now it isn't helping as much'. [And if I haven't said it enough; Focusing on one dom/RH machine, is a BAD idea; It will only further slow us down; (no vnc, almost no sudo, slower, not as lisp friendly as debian); We just need Dr. Karp to give us the Updated Libs, then to install them on our test linux box (rather than worrying about the suns, which has been the mission up till a few days ago). [For some tables;See: reason.pdf ] <-if not on vpn, log into Linux box & you can get it from there.
Machine-wise, I use VNC to get in&get things done. I don't want to mess around w/getting separate X windows tunneled through, right now; I don't have this on the DOM box yet; The two versions of Linux are very similar, &I think Dr. Karp said Debian 1st[you can ask him, when you remind him abou the libs]. I just don't want an(other semi)arbitrary decision, that is an impediment, that could slow things down, vs. helping to speed them up. btw:[Andrew said he would go for Debain (he uses it at home)&it is more Lisp-friendly.] pps:[The project would of benefitted from trying things on more than one OS/machine vs. only-one.]
Got ptools, should get new(er) libs from Dr.Karp soon; Still more worried about the NL parse, than the final format; Also want to see concepts flowing;
Given the NL parse; There can be ambiguity w/the mapping to concepts as well as the parse to a rule. It should be broken down in stages w/annotations as to the best guess. This is probably best done w/a network of annotation objects, that should be kept, rather than one small output, unless it is carefully looked over by a human. ~Related: SMI wants the full text and not just the ruleEd output, when they give a shot at it. [If it isn't obvious; I've asked for more of this kind of output already, if any exists.] I'd ask Ben but he hasn't responded to much more simple queries in some time. [Here is an example of a miss on a 'good parse': Hemoglobin at least 8.5 g/dL,(("Hemoglobin" "C0019046") ("at least" "8.5" "g/dl")) If I did this by hand, I wouldn't use that cui; As it is a level I wan't not a heme.]The ec.tar I emailed, has extra concepts imported to show this very thing, but I haven't had time to walk anyone through this yet..
got the bioportal-webservices example code; in the short term Niggam suggests his perl module; krr.pdf;reason.txt; pprj;
Ben is doing nothing for arches much less the Linux install; I assume that SRI has been compiled the setup w/something newer than v6, but can't tell from the public download; I'll probably have to write SRI directly soon. If you could ask Dr. Karp (more specifically, about a more modern compile, that would be very helpful:) [If this isn't the case, I'd say we should think about using unsupported code, that can't even be ported beyond one machine by ther person who has been maintaining it.]rctbank aic-export setup is from 2004(4yrs old)&uses a Lisp(2versions old)probably from the same period. The install link I was given-goes-to-this>license_page;That we need to find someone who can sign, unless we already have this??
Got hints&wsdl for bioportal onto-tagging service.;Ben&I will work on Arches now.; Got Elig_frames update from Mor this weekend.
Looking at my kkr & jan-report, I realize I need to write more about ontology-as-a-grammar*.
need a tagging service(like ontrez), so prepping to use umlsV6 till I get that;[re: I'm more worried about having something to match than how to match it.]
Upcoming bioportal telecon should be intersting/useful. I started a putting down thought for a KnRep&Reasoning comparison writeup. [Still a wk away, so finally wrote Mor asking about ontrez&pprj updates, &use of ontrez.] (maybe I should try the perl umls qry?)
KM could still use umls-lisp &/or the new v6 web-services. Note on km conversion/use. Snomed update.
It turned out I was already on some of the things Dr Yao wanted to hit; I've been looking at a semi automated way to onto-tag the eHR (~like ontrez); He is also ok w/being able to map from/to ccr, ¬ be slowed by it's convolutions; I'm also going to skip some of the many other tech-impediments, &work w/the small/clean KM code for just a bit(finally), &talk again next Thursday, w/a bit more of the(~tagged)bpd record; represented that way. [Mid day was ~admin; but the initial ocelot to KM was easy]
I'm still trying to make a push in Lisp, but had machine, then other impediments; I need to simplify more than just the ccr, and try a few more things; I think I will make a toy in KM as well (that sri calo uses). It has the ability to add descriptive elements, that would give the DL-style classification; &in fact can export to OWL. It has some other nice features worth trying.
CCR-thoughts: It isn't just the xml that makes it convoluted; it is needlessly complex; I suggest for now a more simple(paired down/altered)Representation; that the parts that are filled out in example CCR could be mapped from to this simple (in)use/(able) form. Also, it is not mixing/(playing well) w/our schema in the Elig-schema and the umls terms as it is being brought together. [To remap a simple rule,(in a few places)I either have to make ~1/2doz needless instances, or add a(~kludge)slot; So I might as well accept, this needs changing.]
When could we try even the starting functionality from the Ontrez paper p3 sec4?? [Aside: I asked Simona about the IE work, I'ld like to learn more; I looked into this again this weekend, &found systems that would tag text for bio-molecular rather than bio-med(like ontrez should do now?/someday?; I want to try it&find out.]
Even though we are promised time-builtins, and jess then swrl, that works; I'm still cautious about this. I was going to hit UMLSks rmi|socket to get the info, or our UMLS work; I'd like to have the words and related concepts as CLOS ins(matches);that I could use both lisa.sf.net vs.jess and then whatever other matching (even DL classify)in Lisp(loom/racer/whatever, if need be). I tried to look at Ben's umls code he gave me, and gkb, but once again run into Sun/our sys, problems; I really could use help getting this set up(Ben?). I'd like to get back to trying out the(heart of the)basic ideas.
Even though others have looked at my ruleEd qry from last yr, I'm not so sure how helpful it will be now; Along w/the smi ontrez ideas; they suggested also getting full-txt; [Even]Getting a (form of) full-txt to full-txt match as a filter (might miss a few things);but could still be helpful; I found out more about UMLS mainly through it's Protege-tab, but could learn more; I'd like to see what sort of word2concept linking (like the bag of words, w/concepts) that keep getting more specific; Then all match instances, incl assoc action matches, can be compared; I'm not sure what the present NL work is doing; &how it might be able to be better put to use. [Summary, I will need more than one CUI, I will need the bits of the related concept heir/graph/relationships] [maybe the match &nl work so far, should completely be separate; related to smi idea&others; more soon][We should see some of the ontrez related tech sooner than later.]
Elegibity_frames+CCR = ec.pprj which can have rules to take things like ranges to symbols; I'm also getting umls concepts from the patient-record, to fit into this pprj; This should be semi-automated.
2008: Even if Ben can't get os9 on arches, this is the lowest priority, as it is only for software support. If he can't even get in to login, then he should at least tell me what to move over; So still waiting on Ben for arches.[I found os10 that he could never get in the drawer, and installed it; still need help w/our lisp setup on arches]
Good Dec talk w/Dr.Sun; will look at getting CCR/vMR* cleaning into the grammar based onto &will get patient in w/him next time. [*look@sage'svMR] I'm going the 19th so want to meet before that.
Conference call w/Mor was really good, we got an overview/description of her protege frame pprj, which extends our grammar a bit; For DL use of swrl(rule-bridge)w/builtins; It looks like the time ability is complete enough. [We got one sun visit d10&expect the last4arches on d12;when I also talk w/Dr.Sun again].
LHts meeting, showed the generic(Ben like)constraint match w/ccrPatient data, that I mentioned to Dr.Sun last time; I'll ask him to annotate the patient info so I can enter it into a ccr record; and will take(pertainent?)eligibility rules, through rule-ed&use a version of the output, w/an even more specific/complete demo of this generic match.
PowerBackup should be in; can update sun w/vnc (&look@firewall); Working on re-establishing support, & can try the Linux-setup though I could use a little help from Ben.
f(test,age,gender)=>normal-range; need support contract # for sun ;look@ccr w/in acl addones
Focus on CCR extraction: Work w/Dr.Sun to get Problem, Medication, Procedures slots filled &on wiki;will add Procedure class, as that wasn't in the example; can pull ccr tags
Check that Graphvizcanhelp w/CTView, &the parsing/use of NormalRangedata.
Met w/Dr. Yao, setup Protege, explained how Ben's output could automatically be turned into (contstr/restrict)/classifica-tion rules; Will meet again Dec6 &send him more before that.
Quick-Stanford-notes,got:normal range data,also some CTView inspired viz(n19)to look at.
I wasn't/am unsure about getting the 30k ccr records; I found out I should contact natasha&nigam&sent a quick email, w/a ccr record nigam asked for&asking if his perl umls module was really what we would use; I'd like to contact Dr. Karp's lab about the software we both use(&how to run it on Linux);&also about an OKBC server, for interop w/Protege.
After the meeting near thanksgiving, I had to cancel my trip to IL (as I had something for the 18th scheduled); I was right in between finding a place for Dec on as well, &just sorted all that over the weekend of the 10th.
Question: Did you get to talk to the others not on the last conference call, like Natasha, about umls bioportal abilities; &did they say anything about it yet? I could write them if you 'd like.
Got to give Dr.Sun a quick overview of much of what is going on, incl a little of what I've been reading. i'll meet w/him again right after the Stanford trip; He saw Protege &liked the tree-view, but hasn't played w/it yet; I'll set it up for him next time. ... more soon ...
Little run-down, still reading* will try one of the time/etc systems soon. &will prep for thur meeting w/Yao
Good call, found out they are less ready than I thought, so there is less of an imminent constraint.
B.Dai sent NCBO current state of software report, next day Dr. Sim caBig rule-engine report, which I had around. NCBO overview actually caused more reading, not less (before Tue call); Hope to get back to test case soon though. [see: book]
In short I went from looking at umn ccr to importing your xml document and yesterday turning it into a set of classes&instances. Interspersed I'd been doing lots of looking around/reading. Esp. after the smi emails. I also wrote Dr. Sun, but no reply yet, I should probably write again. I plan to take the ccr classes, test-case document info, and the code for the first rule match, &iteratively bring them together.
sorry I gave up for awhile, as I realized no one was looking at any of it
Wk3: Hope to move from HW/etc to more meeting prep.
2nd week: plan: continue w/a sun-sysadmin to fix it &/or get a cheap lnxPC/mac &get going; Can the G3 be wiped? I'd like to put os-x on it;&if people like having on the network; we can get a machine just as nicer than one of the sun-blades all upgraded, from the dept-store for ~$900. !!??? [Mac can't be updated, arches stuck/waiting4sun]best to use rctbank,PC&/or home-machines. [hardware is at a stand-still; I'm waiting for sun &/or a new purchase.]Turns out refurbishment really isn't worth it; Got arches upgraded w/some networking. Will have a quote by Mon; Had a good meeting&got some prep for the next.
1st week [2+4-1] Lots of initial setup, but time wasted from road-blocks, info wanted from others
Lots of time looking@bugtrackers+related,& dealing with the suns & wishing we had something else.
Presently I can't get the libs for bugzilla on the suns but can get a KnBased version on the PC. Sun is getting moving; Though we could get something nicer than arches&rctbank for ~1k each.
If trailbank.servehttp.com is up, then I'm(definitely)in the office.