Re: cTAKES questions [EXTERNAL]
Hi John, Good questions. Unfortunately, I can't really say what is going on as it seems that a lot of the information is in your images - 1000 words and all that. Unfortunately, attachments and inserted images will not go through the dev@ email system. Please copy/paste some plain text in this thread and we will try to help you. The first "NOCODE" item might come from a table name mismatch in the database, e.g. "ICD-9" vs. "ICD_9", but that is a shot in the dark. The second issue that you report is more concerning. You are correct in that it is unexpected and most likely not a great thing to have happening. Just in case it makes things easier, you can use another method for getting cuis. For instance, add the SemanticTableFileWriter to the end of your pipeline. It will write one file per note and accepts standard fileWriter parameter "SubDirectory", plus values for parameter "TableType": BSV, CSV, HTML, TAB. Sean From: JOHN R CASKEY Sent: Tuesday, April 11, 2023 11:45 PM To: dev@ctakes.apache.org Subject: cTAKES questions [EXTERNAL] * External Email - Caution * Hello, I have a minor bug to report, and a question that may be a part of a major bug. If I create a custom dictionary with multiple vocabularies and then run cTAKES using this custom dictionary, cTAKES will sometimes replace the vocabulary name with the name of the custom dictionary. An example is shown in the attached image1.png that was run on the MIMIC dataset. I noticed that if I looked up the CUI C1548802 in the UMLS Metathesaurus Browser that had the incorrect vocabulary name inserted, it had ‘NOCODE’ for the code. This only seemed to occur with CUIs from the MTH vocabulary. Is this something that can be fixed within cTAKES? The question and maybe major bug was we ran the same dataset (50 MIMIC notes) twice: once on the custom dictionary with multiple vocabularies described in the attached image1.png, and then using a custom dictionary that only included the snomed vocabulary. Next, we filtered the output from the multiple vocabulary dictionary to only include CUIs that were reported by snomed. The two outputs from cTAKES should have produced the same CUIs, but as can be seen in the attached Venn Diagrams, some of the CUIs reported by cTAKES running the snomed-only dictionary were not reported by cTAKES running the multiple vocabulary dictionary. Do you know why the two outputs would be different? We’re running user installation of cTAKES 4.0.0.1 via ./bin/runPiperFile.sh -p path/to/piperfile -l path/to/custom_dict.xml -i inputDir --xmiOut outputDir And then extracting the CUIs from the output XMI files. Please let me know if I should report this as an issue on the new GitHub repository instead of via email. Thanks! John Caskey
cTAKES running slower with each run
I'm running ctakes both as an API and as a console app. Each time I hit ctakes, the run time per document is getting incrementally slower by a few thousands of a millisecond per xml element (to normalize for different document sizes) than the previous document. Compound this over 1000 documents in 20 minutes and the runs are going from 0.06 milliseconds per xml element to 1.5 milliseconds per xml element. It's a very consistent 0.002 millisecond increase in the rate for each subsequent document I throw at cTAKES. Is there any caching or garbage collection or something I should be on the lookout to adjust or fix? Thanks ~Alex [cid:image001.jpg@01D96D43.69FD3AA0] Alex Milinovich Director of Research - Data Science Analytics | Quantitative Health Sciences 9500 Euclid Ave. - JJN3 | Cleveland, OH 44195 | m: (216) 245-7655 Please consider the environment before printing this e-mail Cleveland Clinic is currently ranked as the No. 2 hospital in the country by U.S. News & World Report (2017-2018). Visit us online at http://www.clevelandclinic.org for a complete listing of our services, staff and locations. Confidentiality Note: This message is intended for use only by the individual or entity to which it is addressed and may contain information that is privileged, confidential, and exempt from disclosure under applicable law. If the reader of this message is not the intended recipient or the employee or agent responsible for delivering the message to the intended recipient, you are hereby notified that any dissemination, distribution or copying of this communication is strictly prohibited. If you have received this communication in error, please contact the sender immediately and destroy the material in its entirety, whether electronic or hard copy. Thank you.
Re: cTAKES running slower with each run
There are many ways to package ctakes, and admittedly ours is unlike the console app and we have our own multithreaded API, but we regularly do millions of documents at a time and haven't seen this issue. The core application with our fairly standard pipeline is up for a month at a time with no degradation Are you using any unusual or deprecated annotators? It could be that one of the less used ones doesn't separate initialization properly from it's processing method and is caching something that it shouldn't.. Are you seeing a concomitant growth in memory footprint? Try running it under jvisualvm it may give you a clue. Peter On Wed, Apr 12, 2023 at 7:41 PM Milinovich, Alex wrote: > I’m running ctakes both as an API and as a console app. Each time I hit > ctakes, the run time per document is getting incrementally slower by a few > thousands of a millisecond per xml element (to normalize for different > document sizes) than the previous document. Compound this over 1000 > documents in 20 minutes and the runs are going from 0.06 milliseconds per > xml element to 1.5 milliseconds per xml element. It’s a very consistent > 0.002 millisecond increase in the rate for each subsequent document I throw > at cTAKES. > > > > Is there any caching or garbage collection or something I should be on the > lookout to adjust or fix? > > > > Thanks > > > > ~Alex > > > > > *Alex Milinovich* > > Director of Research – Data Science Analytics | Quantitative Health > Sciences > 9500 Euclid Ave. – JJN3 | Cleveland, OH 44195 | m: (216) 245-7655 > > > > > > Please consider the environment before printing this e-mail > Cleveland Clinic is currently ranked as one of the nation’s top hospitals > by *U.S. News & World Report* (2022-2023). Visit us online at > http://www.clevelandclinic.org for a complete listing of our services, > staff and locations. Confidentiality Note: This message is intended for use > only by the individual or entity to which it is addressed and may contain > information that is privileged, confidential, and exempt from disclosure > under applicable law. If the reader of this message is not the intended > recipient or the employee or agent responsible for delivering the message > to the intended recipient, you are hereby notified that any dissemination, > distribution or copying of this communication is strictly prohibited. If > you have received this communication in error, please contact the sender > immediately and destroy the material in its entirety, whether electronic or > hard copy. Thank you. >
