09:29:46 From Aurélie Névéol to Everyone : Hello! 09:30:02 From Angus Roberts to Everyone : Hi! 09:47:46 From Goran Nenadic to Everyone : 7 or 8 10:00:09 From Beatrice Alex to Everyone : Jaya feel free join the speaker room to check your slides. 10:04:02 From Angus Roberts to Everyone : Programme: 10:06:13 From Angus Roberts to Everyone : Twitter 10:07:09 From Angus Roberts to Everyone : Gathertown link: 10:14:56 From Paul Rayson to Everyone : I was also thinking about languages other than English, e.g. Polish and Arabic appeared yesterday 10:25:28 From Angus Roberts to Everyone : Some one was asking about making the chat available yesterday. We will make it available after the conference, on the programme page. 10:32:23 From Elizabeth Ford to Everyone : Thanks Anna-Grace, great clear presentation and easy to follow. Do you see any future possibility that these comments could be used to suggest improvements to cancer healthcare services? What additional processing/analysis would be needed to get out this kind of information from the text? 10:36:32 From Chloe to Everyone : Thank you for your talk Anna-Grace! You mentioned that some of the categories that you sourced for you study were partially derived from previous studies in cancer and other conditions. Do you think the other chronic conditions may have different categories? Are different categories linked to different types of cancers? 10:42:27 From Luke B to Everyone : Thank you for your excellent talk, Anna-Grace. I think you're doing very important work. Would be willing to share your slides? 10:45:39 From Anna-Grace Linton to Everyone : Hello Elizabeth, thank you. I think there is a lot of potential that what we can find in the comments can lead to improvement. We do hope to see patterns in the responses and the types of patients who give these response. Highlighting areas that services can focus on. I also see potential in the final phase in which we can see features that impact late cancer outcomes. The additional analysis may need too be some of the contextual information such as the treatment the patient had, stage at diagnosis etc. Also, the hope is the approach can lead to more timely action following collection. 10:48:03 From Elizabeth Ford to Everyone : That's really interesting, thanks Anna-Grace 10:48:29 From Anna-Grace Linton to Everyone : Hello Chloe, that is a good question. There are some topics that are more likely to be mentioned by some conditions more than others, but interestingly there are some that span the board. 10:48:56 From Chloe to Everyone : Thank you for answering my question! Super interesting talk :) 10:49:44 From Anna-Grace Linton to Everyone : Hello Luke, thank you. Yes I will be able to share my slides. 10:53:05 From Elizabeth Ford to Everyone : Great talk thanks Jaya! 10:53:23 From André Bittar to Everyone : @Anna-Grace: thanks for your interesting presentation! I was wondering if you’d looked at work comparing different sentiment analysis resources to determine which might be most suited to the genre of texts you are working with. Have you done your own comparative assessment? 10:53:31 From Scarlett Bae to Everyone : Great talk, thank you Jaya! 10:53:40 From Chloe to Everyone : Thanks for the talk Jaya! Would you consider using more patient-driven sources such as online forums? 10:57:19 From Paul Rayson to Everyone : Thanks for the interesting talk Jaya, a colleague in Linguistics at Lancaster has done some work on pain intensity e.g. http://www.research.lancs.ac.uk/portal/en/publications/-(99ff8545-9f2c-4936-a9a8-0cb7c3989dd5).html 10:59:04 From Jaya C to Everyone : @Chloe I haven’t considered using patient-driven sources at the minute, but it is an interesting thought and something I will look into in more detail, thanks! 10:59:39 From Micheal to Everyone : Interesting work Jaya. My question is more of the eventual output of this work. What will give you satisfaction. Is it successfully enahncing SOTA model performance using existing knowledge bases like SNOMED or being able to evaluate how well SOTA models undersatand these entities? 10:59:42 From Jaya C to Everyone : @Paul thank you so much for sharing the link! 11:02:23 From Jaya C to Everyone : @ Michael Thank you. I think it is probably a bit of both. I would like to successfully enhance performance, and also compare to existing models to see how much of an improvement actually occurs, if that makes sense. 11:12:56 From Anna-Grace Linton to Everyone : @Andre, sorry I missed your question. I have read through literature to identify some suitable tools. I have not done my own comparative study. That is a good question as I think it is something that would be beneficial to do to confirm any choice of sentiment tools used. 11:13:04 From laurensrook to Everyone : These were 3 excellent and crystal-clear presentations. My compliments to all! 11:13:17 From Chloe to Everyone : Agreed! 11:13:25 From laurensrook to Everyone : ;-) 11:16:53 From Jaya C to Everyone : Thank you @Elizabeth @Scarlett @laurensrook and @Chloe :) 11:17:36 From Elizabeth Ford to Everyone : Great presentation Daisy! Thanks 11:17:56 From Daisy Harvey to Everyone : Thank you :) 11:17:57 From Goran Nenadic to Everyone : Gathertown link: 11:19:49 From Katrina Davis to Everyone : To Daisy: A project in our department asks young people to download their social media history. That way you could do retrospective, rather than waiting until they had a relapse. Just a thought if that was the issue. Let me know if you want to know more. 11:20:01 From Katrina Davis to Everyone : p.s. what a brilliant PhD panal 11:31:08 From Julia Walsh to Everyone : Really interesting panel - thank you all. Good to finally see you talk about your project @Jaya C :) 11:32:20 From Matthias Hartung to Everyone : Hi Nicholas, I will be the presenter of the second paper in this session — anything you need me to do before the talk? 11:37:37 From Luke Slater to Everyone : i guess i will try to fix the mic 11:41:03 From Helena Ariño Rodríguez to Everyone : Why do you think there is no overlap in the lexicon derived from MIMIC vs CRIS...? 11:42:08 From laurensrook to Everyone : Would there be cultural differences in pain perceptions? 11:43:19 From Goran Nenadic to Everyone : Any thoughts on the differences between pain terms in clinical notes and literature? 11:43:21 From Lamiece Hassan to Everyone : Thanks Jaya, sounds great - it would be great if you could post your Github link in the chat 11:43:54 From Paul Rayson to Everyone : This one I think: https://github.com/jayachaturvedi/pain_lexicon 11:44:06 From Julia Walsh to Everyone : How are you differentiating between severity and temporal aspects of pain? 11:44:30 From Lamiece Hassan to Everyone : Cheers @Paul! 11:44:56 From Luke Slater to Everyone : have you considered turning the vocabulary into an OWL ontology, so it can be easily reused? 11:45:39 From Luke Slater to Everyone : sounds good! you could also consider doing it programmatically by OWLAPI 11:46:15 From Jaya C to Everyone : @Luke thanks, I will have to look into OWLAPI as this is something I’ve been struggling to do :) 11:47:05 From Jaya C to Everyone : Here’s the GitHub link (thanks for sharing too @Paul) - https://github.com/jayachaturvedi/pain_lexicon 11:58:13 From Luke B to Everyone : @ Matthias, very cool talk! Could you talk a little bit about why the GMM is necessary and why you can't go directly from the text classifier to an intensity score/group by thresholding the network output. I may have misunderstood something along the way. 12:00:54 From Paul Rayson to Everyone : @Jaya - Lama Alsudias, who presented yesterday, used Protege for her Arabic Infectious Disease Ontology, so she might be able to offer advice on that. Also please could you share your slides? Thanks! 12:01:17 From Daisy Harvey to Everyone : @Katrina thanks for your info about downloading social media history, sorry I missed your earlier message. Would be interested to know more about it if you’re happy to share your email address! 12:02:45 From Luke B to Everyone : @ Matthias, Awesome! Thank you for the explanation. I guess you must've played around with a number of clustering/density estimation algorithms? Also, would you be willing share your slides? Thank you! 12:03:46 From Jaya C to Everyone : @Paul thanks, that’s good to know! I will get in touch with Lama for sure. Happy to share my slides :) 12:06:40 From Matthias Hartung to Everyone : @Luke, not exactly sure about different algorithms we might have used — I would have to forward this one to my co-author who did the technical work and afterwards get back to you. 12:07:15 From Matthias Hartung to Everyone : Re. Sharing slides: sure, how can I do that? 12:08:03 From Goran Nenadic to Everyone : All speakers who want to share their slides - please either send a link to your slides (preferred) or slides themselves to contact@healtex.org 12:08:47 From Matthias Hartung to Everyone : @Goran Thanks for clarification! 12:10:48 From Mat Rawsthorne (he/him) to Everyone : https://bit.ly/3zq5NVt 12:13:49 From Matthias Hartung to Everyone : @Luke Coming back to your question: The reason we used GMM, is that it’s possible to decided on the number of components of mixture model, which in our case in 3. For other more recent and advanced clustering algorithms e.g., DBSCAN/HDBSCAN it’s not possible to define the number of clusters beforehand. We also experimented with K-means clustering, but found GMM to be better performing. 12:15:55 From Hang Dong to Everyone : Thanks for the very interesting talk. I wonder whether that 0.1% difference between NoNeg and the full phenotypes can actually matter in some clinical applications? 12:16:24 From Siyue Song to Everyone : @Matthias I also adopted K means and GMM into my research, found GMM work better than K means 12:17:11 From Hang Dong to Everyone : Thanks Luke! 12:17:58 From Lamiece Hassan to Everyone : gmcr google 12:18:10 From Luke B to Everyone : @Matthias Okay, cool! Thanks a lot! 12:20:46 From Fabio Rinaldi to Everyone : hi goran all fine 12:30:42 From Beatrice Alex to Everyone : Sunshine in Edinburgh too. 12:38:54 From Aurélie Névéol to Everyone : @dmitri did you also evaluate the computational needs of models (e.g. run timee) 12:55:13 From Aurélie Névéol to Everyone : @malek, I am not sure I understand how the evaluation is conducted. Do you have gold standard annotations for test sets 1 & 2? 12:56:18 From Aurélie Névéol to Everyone : Thank you. 12:56:47 From Xiaorui Jiang to Everyone : Can you give some more details about active learning. I did not get it. How was it used with BOOT? 12:57:45 From Xiaorui Jiang to Everyone : Yes. Got it. 12:57:48 From Xiaorui Jiang to Everyone : thanks 12:58:12 From Xiaorui Jiang to Everyone : Did it improve the performance of BOOT? 12:58:42 From Hang Dong to Everyone : Thanks for the talk. I have a similar question: Why BOOT is better than BOOT+Active learning? 12:58:50 From Xiaorui Jiang to Everyone : Yes! 12:59:29 From Hang Dong to Everyone : Thank you! 12:59:37 From Xiaorui Jiang to Everyone : Got it. thank you 13:08:20 From Mat Rawsthorne (he/him) to Everyone : it would be great if someone could automate the CSRI (https://www.pssru.ac.uk/csri/client-service-receipt-inventory/) as this is a major time burden for intervention researchers and the participants 13:14:25 From Noel Kennedy to Everyone : Hello! 13:14:52 From Angus Roberts to Everyone : 0900 UK time that is 13:14:57 From Noel Kennedy to Everyone : Yes 13:17:11 From Luke Slater to Everyone : UK mimic would be absolutely phenomenal 13:23:42 From Mat Rawsthorne (he/him) to Everyone : Angus what is the weather like on Soyuz? 13:23:56 From Angus Roberts to Everyone : Gathertown: 13:24:14 From André Bittar to Everyone : Upload your *very good* weather photos (good photos, not necessarily good weather)! “Not just any old cloud” ;-) Here: https://drive.google.com/drive/folders/1TmDk59yjSB-xD-9_m2PVsbUmcht1KHI7?usp=sharing 14:00:13 From Goran Nenadic to Everyone : Upload your *very good* weather photos (good photos, not necessarily good weather)! “Not just any old cloud” ;-) Here: https://drive.google.com/drive/folders/1TmDk59yjSB-xD-9_m2PVsbUmcht1KHI7?usp=sharing 14:04:39 From laurensrook to Everyone : bless you…. 14:05:10 From Angus Roberts to Everyone : I think they have muted now :) 14:08:00 From laurensrook to Everyone : Birmingham Weather is Unbeatable ;-) 14:09:07 From Andreas Karwath to Everyone : Well ….it only rains once a year 14:09:36 From laurensrook to Everyone : Once for 300 days? 14:09:43 From Andreas Karwath to Everyone : no comment 14:10:04 From laurensrook to Everyone : ;-) 14:38:46 From Paul Rayson to Everyone : Many thanks Aurélie for the interesting talk. Will you be able to share the slides please? There are many references there that I would like to follow up but didn’t have time to copy down. 14:39:17 From Aurélie Névéol to Everyone : Yes, sure! 14:40:08 From Paul Rayson to Everyone : Thanks! 14:41:00 From Goran Nenadic to Everyone : Thanks 14:41:36 From Maria Liakata to Everyone : Are there plans for sharing slides more widely, e.g. on the conference website ? I was also asked about my slides yesterday. 14:43:03 From Goran Nenadic to Everyone : @Maria and all speakers: yes, please send the slides or a link to them to contact@healtex.org 14:43:15 From Goran Nenadic to Everyone : Aurelie was referring to this paper 14:43:17 From Goran Nenadic to Everyone : https://www.sciencedirect.com/science/article/pii/S1532046418302016 14:44:34 From Goran Nenadic to Everyone : We’d need to add ethics considerations and to ensure that the journals are recommending if not requesting that the guidelines are followed. 14:47:50 From Maria Liakata to Everyone : Thanks Aurelie! 14:52:01 From Aurélie Névéol to Everyone : Thank you all! 15:01:02 From Angus Roberts to Everyone : http://healtex.org/healtac-2021/programme/ 16:04:33 From Ghada Alfattni to Everyone : @Aurélie Névéol: https://journalofethics.ama-assn.org/article/language-structure-and-reuse-electronic-health-record/2017-03 16:10:14 From Arron Lacey to Everyone : 1 minute left Stuart 16:10:34 From Goran Nenadic to Everyone : All presenters (including demos and posters): please send links to your slides or slides themselves to contact@healtex.org if you wish to share them on the web. 16:10:54 From Angus Roberts to Everyone : We will also share the chat from both days 16:13:30 From Aurélie Névéol to Everyone : @Ghada: thanks! 16:13:37 From Aurélie Névéol to Everyone : Here's the one I mentioned: https://pubmed.ncbi.nlm.nih.gov/28962565/ 16:14:51 From Aurélie Névéol to Everyone : (article providing an estimate of information found in structured vs. non stuctured content of EHRS in French) 16:14:59 From Angus Roberts to Everyone : Join us in gather.town for a drink, after the end of the conference: https://gather.town/i/kJfi6h0e 16:17:04 From Aurélie Névéol to Everyone : @Ivo Fins: sorry we were cut-off in breakout rooms. Thanks for your poster presentation. I will not be able to join gather.town but I was wondering if you had thought about comparing your work on canine obesity to human obesity, e.g. as supported by i2b2 2006 challenge data? 16:21:30 From Arron Lacey to Everyone : 1 minute left Dennis 16:31:50 From Angus Roberts to Everyone : Panel members, please turn your cameras on 16:33:23 From Ivo Fins to Everyone : @Aurélie Névéol thanks for the question, Aurélie. That would be really interesting for sure, there are some studies looking into obesity having 'One Health' perspective (pets vs. humans) and I think it is vital to have more studies where we can analyse and compare dif datasets between populations. However at the moment, we are focusing our study in the canine population, as the overall aim is to develop tools to prevent and manage obesity at an individual, personalised patient-level! I hope in the future to have the opportunity of working on the interface human vs. animal obesity, thanks for your suggestion :) 16:33:51 From Goran Nenadic to Everyone : What are possible, acceptable and sustainable (if not the best) models to collaborate with industry, in particular if they are academia, NHS, patient charity? 16:34:19 From Goran Nenadic to Everyone : [I means as partners] 16:35:40 From Arron Lacey to Everyone : Angus and the Goran 16:35:44 From Arron Lacey to Everyone : then* 16:36:47 From Goran Nenadic to Everyone : ‘the’ Goran :) 16:37:10 From Luke B to Everyone : There can be only one. 16:38:14 From Goran Nenadic to Everyone : And would industry help fund building a MIMIC III dataset in the UK? 16:48:07 From Angus Roberts to Everyone : On David’s point, how much of the difference in the US is due to the funding impetus from Obama Care? 16:52:36 From Luke B to Everyone : To the panel, industry general lags behind the SOTA in terms of models that are deployed by 2 ~ 5 years. For example, I've seen several examples of models at HealTAK2021 that are shying away from Transformers in industry because of how long inference can take, preferring instead to use CNNs or LSTMs, which decode fastert. Do you think the medical industry lags behind even further because of its inherent dependence on government funding (which is infamous for being slow)? 16:58:07 From Aurélie Névéol to Everyone : I think we have also seen in several presentations that rule-based models do fairly well in some circumstances, so perhaps a pragmatic approach makes sense too? 16:58:21 From Aurélie Névéol to Everyone : (as was just said as I was typing :-) 16:59:45 From Aurélie Névéol to Everyone : Thank you for these very insightful discussions! 16:59:49 From Luke B to Everyone : Thank you to the panel! 17:00:43 From Andreas Karwath to Everyone : https://gather.town/i/kJfi6h0e 17:03:13 From Lama Alsudias to Everyone : Thank you for running the fantastic conference in this hard time. It's really very helpful and useful for much research. 17:06:18 From Rakhee yadav to Everyone : will get attendance certificate??? 17:09:07 From laurensrook to Everyone : Thanks for a great conference. I enjoyed it tremendously! 17:09:08 From Scarlett Bae to Everyone : Thank you, Angus! 17:09:11 From Aurélie Névéol to Everyone : Thank you. Have a good week-end everyone! 17:09:12 From Paul Rayson to Everyone : Thanks to the organisers for the great event! 17:09:15 From Jaya C to Everyone : Thank yoU! 17:09:17 From Arron Lacey to Everyone : thanks Angus 17:09:17 From Hang Dong to Everyone : Thank you! 17:09:17 From Rakhee yadav to Everyone : sir please send the link here 17:09:19 From Stuart Gough to Everyone : Thanks everyone! 17:09:19 From Nicholas Cummins to Everyone : Thanks Angus, thanks everyone! 17:09:21 From PJ to Everyone : Thanks for a great set of discussions ! 17:09:21 From Laura to Everyone : Thanks everyone! 17:09:23 From Matus Falis to Everyone : Thank you! 17:09:25 From L.SLATER.1@BHAM.AC.UK to Everyone : thanks all 17:09:26 From Rakhee yadav to Everyone : thanks everyone 17:09:26 From Daniel Schofield to Everyone : Thanks all! 17:09:27 From Beth Rushton-Woods to Everyone : Thanks all 17:09:28 From Angus Roberts to Everyone : https://gather.town/i/kJfi6h0e 17:09:29 From Andreas Karwath to Everyone : https://gather.town/i/kJfi6h0e 17:09:35 From Kristof to Everyone : thanks all! 17:09:39 From Ivo Fins to Everyone : Thanks all! Great conference 17:09:48 From Julia Walsh to Everyone : Thanks Angus, Goran and everyone else - it's been great! 17:09:51 From Aicha Chorana to Everyone : Thanks to the organizers and the presenters!