- From: Cutler, Roger (RogerCutler) <RogerCutler@chevron.com>
- Date: Thu, 20 Apr 2006 12:36:10 -0500
- To: eneumann@teranode.com, public-semweb-lifesci@w3.org
It seems to me that Davide's case study can pretty easily be made into an ROI story titled something like "Computer Aided Literature Searches". I'm obviously not the one to do this for real, since I don't know enough about the subject matter to talk about it convincingly, but let me take a crack at some verbiage that may be all wrong in detail: A case study from L&C (Language and Computing) illustrates the cost savings and new capabilities that can result from combining NLP (Natural Language Processing) with Semantic Web reasoning capabilities. It seems that some entirely unexpected side effects occurred during clinical trials of an anti-diabetic drug, and this small company took on the task of trying to track down the mechanism of the drug interaction. They ran 8800 journal articles through their NLP process, deriving thousands of relations like "enzyme X may activate protein Y" out of this literature and combining the resulting ontology with information from relational databases and other public sources. They analysed the results with a Semantic Web inference engine using queries like "what paths connect 'PKC Beta inhibitor' and 'Diabetes'", resulting in a plausible explanation as to the previously unknown mechanism causing the effect. Taking the specifics of this case study as representing a typical example of what might be expected from computer aided literature searches, the above results were obtained in a "couple of days". Let's assume that a few people were working on it and that the cost was about 10 man-days. The alternative would be to examine all the journal articles by hand and make database queries as appropriate to supplement that information. Taking this at face value, perhaps an expert person could "process" ten journal articles per hour, which for 8800 journal articles translates to 110 man-days. On the face of it, the computer aided process has ten times less cost. There is also the possibility that the computerized process could bring to light relationships that a person would miss, and if the number of journal articles were much greater perhaps this "possibility" would become highly probable -- but it is very difficult to quantify the value of this potential benefit so we will pass over it in this analysis. Guessing that the total cost for the services of an expert in a major industry is about $1000/day, we have a cost comparison for this operation of $10K vs $110K. Supposing that a large company might have fifty such issues to research per year, we find a $5.5M expense being reduced to $0.5M, a cost savings of $5M for a single company. Replicating this to, say, ten companies involved in this industry yields a cost savings from this one type of application of $50 million. Note that in this example I have tried to include the following elements: 1 - A specific example in which Semantic Web technology is used to solve a specific problem, and a plausible technical indication of how that specific example works. 2 - A cost analysis for this example of the SW technique and ALSO a cost analysis of the likely alternative if SW is not available. 3 - An analysis of how this might be replicated and scaled out to provide total cost savings. Displaying the pieces of this analysis in detail allows a skeptical reader to substitute his or her own guesses for each atomic part of the analysis and also perhaps to make some sort of quantitive estimate of the reliability of the guesswork. -----Original Message----- From: public-semweb-lifesci-request@w3.org [mailto:public-semweb-lifesci-request@w3.org] On Behalf Of Eric Neumann Sent: Wednesday, April 19, 2006 10:54 PM To: public-semweb-lifesci@w3.org Subject: wiki page for semantic web ROI I have added a wiki page to begin the discussion (as well as collect different anecdotes and examples) into the various kinds of return of investment (ROI) that could be realized by applying semantic web technologies: http://esw.w3.org/topic/HCLSIG/ROI Any additional examples people would like to include and discuss should be added to the list of examples on this page. Although areas of potential benefits are identified and presented in economic terms, estimates of how much the semantic web could improve these situations do not yet exist. It is hoped that one goal of this discussion will help put into context the potential value of some of the current HCLS tasks. Working demonstrations will allow us to begin estimating how much of an impact (e.g., reducing development and data design time and costs) these approaches will have in the various HCLS areas. Eric
Received on Thursday, 20 April 2006 17:36:57 UTC