This is at the heart of the whole issue. Medical Data is hard and that leads to crappy implementations, unintuitive UI, less desire for docs to use it yada yada...
Although I disagree that one needs to represent everything "symbolically" there are several EMRs with free-text note fields and post-process the notes with NLP for meaningful secondary uses. Yes NLP is not perfect but things are getting much smarter and better.
PS: Those purist self-righteous PhD Description Logicians. Leave us Alone!!!:)
Yes you are right but Google/Amazon are read-heavy so a big hash table works perfect since there is less disk access - however FaceBook is both READ and WRITE heavy so its more challenging. From their last SIGMOD Cassandra paper, I gather they use some sort of in-memory write which is then synced with their bigtable. Very cool stuff!
I worked at a big tech company doing SemWeb, where my experience was exactly the same. Everyone was scratching their head.
Now I've moved into Healthcare IT environment, where SemWeb makes perfect sense. Its like the best tool for the job.
The essential difference is what end of the stick you are picking up. The tech folks who are trying to shoe-horn RDF/OWL onto anything n everything (e.g. search) are failing. On the other hand, Healthcare/Life science folks who have to work with heavy knowledge intensive stuff, its working like a charm.
The SemWeb story is quite similar to Amazon Kindle.. wherein the tech folks are hating it whereas real users are all over it.. So it might seem like a failure to all you tech bozos.. but the domain experts are lovin' it.
OK so I just googled this phrase and it got me this..
"You must be new here.â (Invoked frequently after a poster complains of a common Slashdot issue such as duplicate stories or perceived bias by certain editors)
This is sometimes answered with "No, I'm New Here" by a user named New Here."
I just find it remarkable that you can find almost anything and everything online these days... so ARE YOU THERE "New Here" ???
pardon my ignorance in digital electronics - I thought capacitors "store" charge for a while with some decay rate, can this be comparable to the function of "memory"?
FB already allows one to target by profile items or keywords.. It does not do any form of "hyper" "meta" clustering.. but its provides pretty good targeting based on location and age.. which itself is very powerful.
Google has done a great job in searching raw free-text data. However, healthcare data is a different beast. The sheer number of datatypes is mind-boggling -- the number of different labs, drug classes, diseases etc that can get coded in patient records runs in to millions. So over the years healthcare databases have been constructed differently - they follow an EAV (Entity Attribute Value) representation, which means that the patient databases are generally just ONE BIG TABLE! Here is the database schema used at New York Presby. Schema - all past 20 years patient data is stored in one table! oh yeah.. DB2 Baby!
Essentially all data/knowledge complexity is present in the Ontology/Terminology (such as SNOMED or LOINC) and the patient data itself instantiates from these.
Also doing NLP over medical notes is a difficult problem requiring years of tuning and domain knowledge to construct one -- which again is so specific to a given institution or region that it just does not work elsewhere.
It would be interesting to see what *real* innovations Google brings on the table.
well you are right but when you talk about the domain of healthcare or biomedicine in general, the complexity of data and processes is so high that to develop a software system you need "extra" data-structures/information models such as HL7 standards, ontologies etc. to meet the requirements of the application. So somehow I tend to think that X (health care domain-specific) language would a superset of a general purpose language that simply provide basic programming elements (say OO, loops, variables etc). Not sure about verbosity though.
In mathematical terms:
A = {basic set of programming artifacts} B = {domain-specific structures and computable knowledge elements}
yes its a huge market. it all works fine in a local institution, but the real challenge lies when you try to "generalize" it to different institutions, each with their own idiosyntric processes and data elements. Keep in mind unless you make the underlying engine some standards based (using RIM or terminology driven) or use good design software practices (Archetypes) you ll have a lot of trouble customizing it.. unless of course.. you become like existing vendors who develop the whole thing from scratch at each installation site and send a team of IT services who work there forever and keep your revenue stream running. Good Luck.
dont underestimate the power of MUMPS. My advisor was one of the developers of the language (Octo's lab at Mass General).. sometimes during our meetings he pulls up the MUMPS command prompt and writes 2 lines to do stuff that would probably take me alteast 50 lines of Java code. seriously no joke.
Someday I plan to learn it once he lets me graduate:)
Unlike the data from for-profit companies (AOL, Google etc), medical records are not necessarily "owned" by hospitals. In the US, the data is owned by the PATIENTs versus in the UK it belongs to the NHS Reference
So atleast in the US, hospitals can get into legal trouble for even disclosing anonymized dataset without consents from each and every patient (although several hospitals make patients sign a form waiving their rights to ownership)
You guys bitching about the presentation software dont understand big picture -- Presentations (that include a powerpoint-like slideshow) are a means to COMMUNICATE an idea or information in general. If used properly these can work to your advantage (remember 1picture==1000 words!).. imagine trying to explain PageRank to a non-mathematical audience without a network diagram. Further, IMO, experienced listeners (in the given domain) generally dont have problem in filtering out the "bells and whistles" from the actual CONTENT. So stop whining, go to your mothers basements and write your kernel code.
what if we discover or invent an alternative for software? say for example, "neuro-self-programmable-semantic-object" machine that understands all requirements, data, processes etc and generates the relevant code? then what happens to all the programmers?..can we all then just go back to jungles..and let the bussinesses run themselves.. i know its very hypothetical but nevertheless a possibility..
i have seen answer.com showing up frequently on top.. and AFAIK it largely mirrors wikipedia articles (and may be adds something more of its own)
so my question how and where does it get the incoming links for a high PageRank? do people point to answer.com URLs?
I would argue that Semantic Web notion puts the control in your hands where different people/communities/businesses "reuse" a common schema/ontology such as RSS, FOAF - that allows people to create their profile and put URIs to their respective friends' FOAF. So nobody *owns* your information (as in case of FaceBook, Myspace) and moreover you can control your information using public/private keys.
Slashdot Mirror
MOD PARENT UP!
This is at the heart of the whole issue. Medical Data is hard and that leads to crappy implementations, unintuitive UI, less desire for docs to use it yada yada...
Although I disagree that one needs to represent everything "symbolically" there are several EMRs with free-text note fields and post-process the notes with NLP for meaningful secondary uses. Yes NLP is not perfect but things are getting much smarter and better.
PS: Those purist self-righteous PhD Description Logicians. Leave us Alone!!! :)
Yes you are right but Google/Amazon are read-heavy so a big hash table works perfect since there is less disk access - however FaceBook is both READ and WRITE heavy so its more challenging. From their last SIGMOD Cassandra paper, I gather they use some sort of in-memory write which is then synced with their bigtable. Very cool stuff!
I worked at a big tech company doing SemWeb, where my experience was exactly the same. Everyone was scratching their head.
Now I've moved into Healthcare IT environment, where SemWeb makes perfect sense. Its like the best tool for the job.
The essential difference is what end of the stick you are picking up. The tech folks who are trying to shoe-horn RDF/OWL onto anything n everything (e.g. search) are failing. On the other hand, Healthcare/Life science folks who have to work with heavy knowledge intensive stuff, its working like a charm.
The SemWeb story is quite similar to Amazon Kindle.. wherein the tech folks are hating it whereas real users are all over it.. So it might seem like a failure to all you tech bozos.. but the domain experts are lovin' it.
It is the antithesis of the anti-net neutrality argument.
This has gotta be the quote of the year!
YES
I live in Manhattan and I dont even have a bike.. i have to take stinking subways, U insensitive clod!
Dan Gilbert did this experiment - its described his book Stumbling on Happiness. Here is the NYtimes article. In this video, he describes the theory of Choice Paralysis
The theory in short means we all think that "Breasts of my wife GREAT! All others suck! " (except Tina Fey.. and Natalie Portman and.. Penelope Cruz)
every1's out applying to M$ for the vacant position..
"You must be new here.â (Invoked frequently after a poster complains of a common Slashdot issue such as duplicate stories or perceived bias by certain editors) This is sometimes answered with "No, I'm New Here" by a user named New Here."
I just find it remarkable that you can find almost anything and everything online these days... so ARE YOU THERE "New Here" ???where t[1.5billion] = 1.26, p 0.005, CI=96.5%, (never forget your confidence intervals)
pardon my ignorance in digital electronics - I thought capacitors "store" charge for a while with some decay rate, can this be comparable to the function of "memory"?
FB already allows one to target by profile items or keywords.. It does not do any form of "hyper" "meta" clustering.. but its provides pretty good targeting based on location and age.. which itself is very powerful.
director of Google Research, "a large corpus of data can be much more valuable than an efficient algorithm" - Unreliable reference
Google has done a great job in searching raw free-text data. However, healthcare data is a different beast. The sheer number of datatypes is mind-boggling -- the number of different labs, drug classes, diseases etc that can get coded in patient records runs in to millions. So over the years healthcare databases have been constructed differently - they follow an EAV (Entity Attribute Value) representation, which means that the patient databases are generally just ONE BIG TABLE! Here is the database schema used at New York Presby. Schema - all past 20 years patient data is stored in one table! oh yeah.. DB2 Baby!
Essentially all data/knowledge complexity is present in the Ontology/Terminology (such as SNOMED or LOINC) and the patient data itself instantiates from these.
Also doing NLP over medical notes is a difficult problem requiring years of tuning and domain knowledge to construct one -- which again is so specific to a given institution or region that it just does not work elsewhere.
It would be interesting to see what *real* innovations Google brings on the table.I think all of these content-generating sites have some form of revenue-sharing..
HuPages: http://hubpages.com/
Squidoo: http://www.squidoo.com/
Gather: http://www.gather.com/
well you are right but when you talk about the domain of healthcare or biomedicine in general, the complexity of data and processes is so high that to develop a software system you need "extra" data-structures/information models such as HL7 standards, ontologies etc. to meet the requirements of the application. So somehow I tend to think that X (health care domain-specific) language would a superset of a general purpose language that simply provide basic programming elements (say OO, loops, variables etc). Not sure about verbosity though.
In mathematical terms:
A = {basic set of programming artifacts}
B = {domain-specific structures and computable knowledge elements}
X = {A U B}
and Y = {A}
yes its a huge market. it all works fine in a local institution, but the real challenge lies when you try to "generalize" it to different institutions, each with their own idiosyntric processes and data elements. Keep in mind unless you make the underlying engine some standards based (using RIM or terminology driven) or use good design software practices (Archetypes) you ll have a lot of trouble customizing it.. unless of course.. you become like existing vendors who develop the whole thing from scratch at each installation site and send a team of IT services who work there forever and keep your revenue stream running. Good Luck.
my 2 cents
dont underestimate the power of MUMPS. My advisor was one of the developers of the language (Octo's lab at Mass General).. sometimes during our meetings he pulls up the MUMPS command prompt and writes 2 lines to do stuff that would probably take me alteast 50 lines of Java code. seriously no joke. Someday I plan to learn it once he lets me graduate :)
become a real JEDI you will.. when u use Lotus Notes for 6 months.
So atleast in the US, hospitals can get into legal trouble for even disclosing anonymized dataset without consents from each and every patient (although several hospitals make patients sign a form waiving their rights to ownership)
You guys bitching about the presentation software dont understand big picture -- Presentations (that include a powerpoint-like slideshow) are a means to COMMUNICATE an idea or information in general. If used properly these can work to your advantage (remember 1picture==1000 words!) .. imagine trying to explain PageRank to a non-mathematical audience without a network diagram.
Further, IMO, experienced listeners (in the given domain) generally dont have problem in filtering out the "bells and whistles" from the actual CONTENT.
So stop whining, go to your mothers basements and write your kernel code.
OMGF.. ROTFL.. this is hilarious...
what if we discover or invent an alternative for software? say for example, "neuro-self-programmable-semantic-object" machine that understands all requirements, data, processes etc and generates the relevant code? then what happens to all the programmers?..can we all then just go back to jungles..and let the bussinesses run themselves.. i know its very hypothetical but nevertheless a possibility..
i have seen answer.com showing up frequently on top.. and AFAIK it largely mirrors wikipedia articles (and may be adds something more of its own) so my question how and where does it get the incoming links for a high PageRank? do people point to answer.com URLs?
I would argue that Semantic Web notion puts the control in your hands where different people/communities/businesses "reuse" a common schema/ontology such as RSS, FOAF - that allows people to create their profile and put URIs to their respective friends' FOAF. So nobody *owns* your information (as in case of FaceBook, Myspace) and moreover you can control your information using public/private keys.