- The following discussion is an archived debate. Please do not modify it. Subsequent comments should be made in a new section. The result of the discussion was
Approved.
Operator: AndrewGNF
Automatic or Manually Assisted: Automatic
Programming Language(s): pywikipedia (probably)
Function Summary: Create ~10,000 pages corresponding to gene information. See ITK for an example. Proposed trial will create ~10 gene pages for proof-of-concept and further feedback from Wikipedia:MCB
Edit period(s) (e.g. Continuous, daily, one time run): One time run, estimated to take ~17 hours (easily scheduled in non-peak hours). Quarterly updates of gene infoboxes will be proposed separately later.
Edit rate requested: 10 edits per minute
Already has a bot flag (Y/N): N
Function Details: Full specs are at User:ProteinBoxBot#ProteinBoxBot_specs
Discussion[edit]
Good task, I support bot-created voices. I also support the high edit rate, 10k pages are many and the high edit rate will help. Anyway, 1k minutes means over 40 hours, not two. Snowolf (talk) CON COI - 00:10, 7 April 2007 (UTC)[reply]
- Very interesting.
Approved for trial. Please provide a link to the relevant contributions and/or diffs when the trial is complete. Create 10 articles, and we can check them out. —METS501 (talk) 00:19, 7 April 2007 (UTC)[reply]
- I'm looking forward to seeing the result of this trial. Would it be possible to make sure that the trial includes a demonstration of what happens when the bot hits an already well developed page, such as IL-2? --Selket Talk 13:58, 19 April 2007 (UTC)[reply]
Just want to post a quick update that the first pages will probably made in the next couple of weeks. We had a bit of a lag while we identified a student to take this on (and the bot approval process was much more seamless than I'd anticipated!) AndrewGNF 14:41, 11 May 2007 (UTC)[reply]
- Glad to hear it. Take your time, there's no rush. ST47Talk 00:54, 12 May 2007 (UTC)[reply]
- ((OperatorAssistanceNeeded)) OK, maybe there's a little rush...still with us? --ST47Talk 12:24, 28 May 2007 (UTC)[reply]
- Absolutely, still here. ;) And I'm still very excited about this project. (I've given a couple talks about this in biology depts, and even though it's only a quarter of the subject matter, it's 90% of the questions/comments...) Why the delay? We're trying to do this as a collaboration with a professor and graduate student at San Diego State University, and therefore we've had to identify the student, work out communication issues, compete with classes, get up to speed on wiki bots, etc. Admittedly, longer than I'd hoped/expected, but now that it's summer, things should pick up. Just got an update earlier today on parsing our XML database content into java objects. Next step will be outputting those objects to wiki text. To be more precise than last time, I think we'll be making WP pages with our prototype bot by the end of this month. Anyway, as I mentioned before, I'm perfectly happy :withdrawing this bot request until we're ready to hit the go button. Just let me know if that would be preferable! AndrewGNF 19:17, 29 May 2007 (UTC)[reply]
- No worries, we can wait. If it'll be a few months though, then we can archive it for the time being. —METS501 (talk) 03:10, 9 June 2007 (UTC)[reply]
- Sorry for the delay in replying here, got tied up in another project for a few days. We're still on target to have our first test pages built by the end of June. I'll try to remember to post periodic updates here, and the ProteinBoxBot user page will be the main area for updates. As always, feel free to contact me directly with any questions, updates, kicks in the pants, etc... AndrewGNF 16:53, 14 June 2007 (UTC)[reply]
- ((OperatorAssistanceNeeded)) (Please note that I am not a member of the BAG as the template may express) Any further updates that you can tell us? It has been quite a while since the actual request has been open? Are you sure you don't want it archived temporarily by the BAG? E talk 04:20, 23 June 2007 (UTC)[reply]
- Sorry to be the laggards of the bunch. But there has been lots of progress behind the scenes -- we are trying to manage lots of data about genes and proteins! Nevertheless, the first manually-uploaded pages were posted this weekend here: User:JonSDSUGrad. The next step is to write the code to handle automated bot uploads and edits. Again, the target is the end of June (in the next week or so). (As I mentioned above, since the original approval was so fast we are certainly open to temporarily archiving if the BAG sees fit...) AndrewGNF 16:31, 25 June 2007 (UTC)[reply]
Quick update... we've been working though some problems using the java wiki bot framework. no bot-generated pages yet, but we're close and working hard! AndrewGNF 17:20, 10 July 2007 (UTC)[reply]
- First automated bot edits done! Not quite yet the stage at which we want to publicize for the scrutiny of the wider community (and propose for an extended trial or full bot approval), but close. User:JonSDSUGrad. AndrewGNF 23:52, 12 July 2007 (UTC)[reply]
- Sounds great, and if you need any help with JWBF, I use it almost entirely for my bots, so I can answer any questions etc. It's a great framework, just takes a bit of getting used to. Also, could you point to the contributions so far? They aren't logged under the bot account. Matt/TheFearow (Talk) (Contribs) (Bot) 02:46, 20 July 2007 (UTC)[reply]
- Sorry, out of the office for the past few days. Thanks for the offer of help, I will pass it along to User:JonSDSUGrad (but it sounds like he's gotten most of the kinks worked out). The auto-generated page tests are currently being done under User:JonSDSUGrad's user account, and most of the test pages are posted on his user page. (In retrospect, probably the point of the preliminary approval was that we'd do it with the ProteinBoxBot user account...) Anyway, I think the latest pages are here: User:JonSDSUGrad/Sandbox/Auto_Gen. Cheers, AndrewGNF 06:29, 25 July 2007 (UTC)[reply]
- ((OperatorAssistanceNeeded)) Where are you at with this? ~ Wikihermit 01:10, 10 August 2007 (UTC)[reply]
- I have been in discussion with them on the programmers talk page, see discussion there. They are almost ready for full trial. Matt/TheFearow (Talk) (Contribs) (Bot) 02:04, 10 August 2007 (UTC)[reply]
- The target is to make the edits in the main WP namespace tomorrow (Friday) or latest monday and report both here and at Wikipedia:WikiProject_Molecular_and_Cellular_Biology/Proposals. The latest debugging run is at User:JonSDSUGrad/Sandbox/Auto_Gen5, and we've put together a flow chart and manual on User:ProteinBoxBot. Sorry, I get the sense our timeline has been atypical, but I think people will agree that it's worth it. More soon... Cheers, AndrewGNF 02:55, 10 August 2007 (UTC)[reply]
I'll admit this is out of my area of expertise, but I am concerned that this bot's task runs afoul of both Wikipedia:N and Wikipedia:C.
I have only reviewed one article created by the bot, but that article utterly failed to assert the notability of the subject, so it was speedily deleted. Please be sure to consider the requirements of Wikipedia:N, including citation to reliable sources that significantly discuss the subject per Wikipedia:V.
Personally, I also think there's a serious Wikipedia:WWIN problem here. Wikipedia is not supposed to be a directory of proteins or anything else in particular. If a protein is notable, it warrants an article. Otherwise, it doesn't. I suspect a wiki created for this purpose would be a more appropriate place to do this.
Additionally, the article created by the bot was flagged as a copyvio. The bot is apparently taking its information from another website. The website in question is apparently derived from public domain sources, but it bears a copyright notice, and it does not necessarily follow that a work derived from public domain sources is itself public domain. If sufficient creativity is involved in compiling or arranging the data, it can acquire copyright protection. I haven't reviewed this situation completely at this point, so I can't determine whether this is so. I'm just raising the issue for discussion.
Finally, if the bot is simply importing data from the other website (I'm not saying that's all that's going on, but if it is), then I would also question the benefit of that, as the other website already exists, and having a duplicate here would create more administrative work for WP, as well as requiring updating as the original is modified.
Thank you for considering my concerns. -- But|seriously|folks 17:02, 13 August 2007 (UTC)[reply]
- Yikes, you raise a lot of issues so I'm not going to try to tackle them in detail all in one go. First, the proposal was discussed in quite a bit of depth by both the Bot Approval Group and the WP:MCB project. Many of these issues were previously discussed (regarding notability and WWIN) and the consensus was to go ahead with the trial. I'm fully in support of considering these issues in depth (and now that the trial is "complete", now is probably the right time to do it), but at some point we have to make the decision of whether or not to do this project. And, if the decision is to move forward, I'd like not to have to re-justify when individual users come across a page and have objections. (To reiterate, I fully support addressing these issues now, but my question is: what are the criteria we have to satisfy and who do we have to convince? Otherwise, I fear this becoming a recurring and motivation-sapping exercise...)
- On to specific points. Notability of the genes themselves, I think, is a given. These are human genes, the stuff of life! (If more is needed to convince you of this, I'll gladly write more, but for the time being, I'm considering this point agreed upon.) The notability of our bot-generated articles of human genes is a reasonable question. Our proposal is to start with the most well-described genes in the scientific literature. These are the genes that influence human disease or basic human processes (the genes here User:ProteinBoxBot/PBB_Log_Wiki_Live_Run are good examples). We'll generate stubs for these most-well-studied genes first, and go down the list (in groups of ~100) until the stubs no longer become useful (with much input from WP:MCB). The goal of this, of course, is not only to provide useful information for each one of these genes, but to encourage the biological community to help really bring the notability of the article in line with the notability of the gene itself.
- Regarding copyright, both the "other website" (in fact there are many) get this text from this NIH-run website: [1]. All content is in the public domain (I can dig up the specific declaration of that if necessary). But the point is not merely to duplicate content found elsewhere. The point is to seed stubs so that we encourage new users to submit content that is not available elsewhere. In fact, there is no *mechanism* elsewhere to collect this content, which is why this idea is so great for both WP and the biological community. (again, I'm biased.) True, we could create a dedicated wiki as others have done, but I think for this particular project, it's a win-win to have it directly in WP. With a little time, these stubs will turn into really great and noteworthy articles. (For example, check out p53, which is a gene page that came into fruition prior to our bot effort.)
- Finally, we just did the "live" trial run last night. Although discussion of these issues preempted our official announcement here, also take this as a formal indication that the trial is done and we are soliciting feedback on the pages created. Again, full log is at User:ProteinBoxBot/PBB_Log_Wiki_Live_Run, and more details on User:ProteinBoxBot. Moving over now to WP:MCB to make a similar announcement. Cheers, AndrewGNF 19:44, 13 August 2007 (UTC)[reply]
I've recreated the page and manually added a bibliography section to establish notability. Wasn't there some discussion about having the bot add references? Tim Vickers 20:23, 13 August 2007 (UTC)[reply]
- Not sure if there was, but if there wasn't, we should have. Easy thing to add... AndrewGNF 20:27, 13 August 2007 (UTC)[reply]
Note; I've added ProteinBoxBot to CSBot's list of allies so that it will not flag further creations from it— I'm assuming here that any sources used are used with permission and attributed? — Coren (talk) 21:31, 13 August 2007 (UTC)[reply]
- Super, thanks much. Yes, all sources come from the public domain (which to my knowledge do not need direct attribution; someone please correct me if this is incorrect.) Cheers, AndrewGNF 21:44, 13 August 2007 (UTC)[reply]
Can you apply filters to the PubMed database hits? I've got my personal PubMed to use the "limits" field to give a separate list of reviews. These would be a better set to add to a Wiki article. Tim Vickers 22:03, 13 August 2007 (UTC)[reply]
- Also, although they do not need attribution, we do here at Wikipedia. Just add a references section with the link, or sources section, or whatever. It helps us make sure its public domain, and provides a base for people looking for more information. Matt/TheFearow (Talk) (Contribs) (Bot) 22:04, 13 August 2007 (UTC)[reply]
(ec)Let me first repeat that I am not an expert in this area and that I just want to make sure these issues are addressed by those who are in a position to understand their application.
As for Wikipedia:N, genes are obviously notable in the "synonym of important" sense. My concern is with the Wikipedia definition of notability, which requires that a subject be significantly discussed by reliable sources. So if a particular gene has not been the subject of an article or book somewhere, it shouldn't have its own Wikipedia article.
On the copyright issues, it is correct that copyright law does not require attribution to public domain sources. However, good scholarship probably does, so the information can be verified. Wikipedia:V also requires that asserted facts be cited to reliable sources. The bot should identify the source of its information and explain that it is public domain (if so) so that other users do not flag them as copyvios.
Additionally, as I mentioned above, it is possible to compile, aggregate or organize public domain information in such a way that it acquires copyright protection. I would recommend that this bot be designed to take the information directly from the public domain source, not from the website it is using. If this cannot be done, that suggests that some creative effort was necessary to arrange the data on the website and that there is some copyright protection there. (They did include a copyright notice.)
If it is unclear whether the website has copyright protection, a potentially easy way to solve this problem would be to obtain GFDL release of the website content from its owner using the process at Wikipedia:COPYREQ.
Good luck! -- But|seriously|folks 22:11, 13 August 2007 (UTC)[reply]
- I can attest to Protein Box Bot taking its information directly from the source - It parses a large database xml query from the Entrez Gene database - it does not copy or scape any information from a website. Its very first step is to parse a very large XML text file and it only interacts with Wikipedia and the PDB (for images - also public domain) websites.
- Also, part of the large xml file I get includes a huge list of publications in the form of pubmed UIDs. I am going to add a subroutine to query the pubmed DB and extract information which can then be used in a cite journal template. The only problem is that there are a huge amount of publications listed in the data file - There are 375 for PPARG alone. They are also listed in order by UID and not by any 'importance' factor. I don't think it would be wise to just list all 375 publications, but we could pull a small hand-full out and post them easily enough. What would be an acceptable method of determining the 5 or so citations needed? We could also/instead put a citation for the Entrez gene Database as a nice 'catch-all'. JonSDSUGrad 22:31, 13 August 2007 (UTC)[reply]
- Tim had a good suggestion to filter for review articles. Great idea... As to how many, I'd say 5-10. (As an aside, should we be having these somewhat detailed discussions here, or over here? Cheers, AndrewGNF 22:41, 13 August 2007 (UTC)[reply]
Personally, I'm not worried about Wikipedia:N. They (individual proteins) may not be notable, but they are most certainly encyclopedic, and if the notability guidelines get in the way, that's certainly a good case for Wikipedia:IAR. — Coren (talk) 22:49, 13 August 2007 (UTC)[reply]
- On the subject of copyright, this page confirms that all data taken from NCBI is in the public domain. On the subject of notability, it sounds like the addition of a few references (as Tim did for PPARG) satisfies this requirement. As described above, we will add these to the bot-created pages. Cheers, AndrewGNF 22:50, 13 August 2007 (UTC)[reply]
- Here is the place for this discussion. Also, maybe a good idea is to create a subpage of the talk page such as Talk:Proteinname/References and link it from the references section and the talk page - that way you can include them all quite easily, and they can be merged into the article by others. Maybe not linking from the article, but from the talkpage definately. Matt/TheFearow (Talk) (Contribs) (Bot) 22:52, 13 August 2007 (UTC)[reply]
- Yes, we could certainly create such a link. However, I'd suggest that we don't for two reasons. First, all the references can easily be retrieved using a link like this: http://www.ncbi.nlm.nih.gov/portal/query.fcgi?p$site=entrez&db=gene&cmd=Display&dopt=gene_pubmed&from_uid=5468, where the "from_uid" parameter is a gene ID listed in the infobox. Second, I'm not sure anyone would really go through the list and incorporate content into the main text. I think more often, contributors will come with a piece of knowledge they want to contribute and find the references to support it, not the other way around. Cheers, AndrewGNF 23:42, 13 August 2007 (UTC)[reply]
- I was just meaning to have some way of listing the references - maybe just linking to that link there from a references section would do, that saves copying it over. Matt/TheFearow (Talk) (Contribs) (Bot) 23:45, 13 August 2007 (UTC)[reply]
- I would go with copying it over. Users who want the references don't want to go searching through other links for them; they're at the bottom of the page anyway and don't bother anyone who doesn't want to see them. —METS501 (talk) 23:48, 13 August 2007 (UTC)[reply]
- We can mock up pages both ways. My guess is that both methods will seem messy and cumbersome, and that choosing the 5-10 most recent review articles will be the best method of populating a references section. AndrewGNF 23:53, 13 August 2007 (UTC)[reply]
- The references section is to provide a source for specific statements made in the article. If the bot is just retrieving recent review articles, that information wouldn't belong under References. --Arcadian 00:11, 14 August 2007 (UTC)[reply]
Also should mention here, when the ProteinBoxBot finds a gene where the gene symbol, gene name, or gene aliases already exists in the WP namespace, it writes the code for the updated infobox to the log file without making changes to the page and flags these pages for manual inspection and integration. I think these then qualify as "semi-automated" edits. I've updated two examples: Apolipoprotein E and Amyloid precursor protein. Comments on this use of the ProteinBoxBot are welcome too... Cheers, AndrewGNF 23:50, 13 August 2007 (UTC)[reply]
(RE: Automagic Bot referencing). I'm a little nervous about automatically linking x-number of PubMed refs - even the review articles can have a tendency to mention a given gene only peripherally (which sort of defeats the point). I don't have any hard examples at hand, just a concern that marginally-helpful reference links may serve to devalue the usefulness of well-chosen refs. -- MarcoTolo 00:03, 14 August 2007 (UTC)[reply]
- To clarify, I'm all in favor of PBoxBot itself, just concerned about the automatic ref generation issue. -- MarcoTolo 00:05, 14 August 2007 (UTC)[reply]
- The issue is, for a bot, that's about the best that can be got. Maybe include ~10 most recent, and a link to the full list of references? That way people who want more can get them. Matt/TheFearow (Talk) (Contribs) (Bot) 00:07, 14 August 2007 (UTC)[reply]
- My concern centers around the issue of "implied selection" - i.e. the implication that references listed in an article have been screened by someone for suitability (at least at some level). I'm tempted to suggest that we simply link to the full PubMed search (defaulting to the "Review" article list, if any) rather than arbitrarily choosing the most recent 5-10. -- MarcoTolo 00:17, 14 August 2007 (UTC)[reply]
- Two quick comments. First, the "peripheral mention of a gene" problem is partially mitigated because these are links in the "gene2pubmed" file (if you are familiar with it). So this is somewhat more selective than simply doing a free-text search at Pubmed. Second, the implied selection problem is a good point. Perhaps we should add all review articles to the references section? (On Arcadian's point above... If the references section is for specific statements in the article, perhaps we could put the "more reading" references in a separate section. How about "Bibliography", as is shown in PPARG?) AndrewGNF 00:29, 14 August 2007 (UTC)[reply]
- "Further reading" and "Bibliography" would both do the trick -- some context is available at Wikipedia:Citing_sources#Further_reading.2FExternal_links. --Arcadian 00:34, 14 August 2007 (UTC)[reply]
- Ah, I'd missed the g2p screening factor - that helps somewhat. As for a separate section, I personally like the "Further reading" option over "Bibliography" since the later still implies to me that a human decision process was involved (though this is really a minor point). Frankly, either header does the trick. -- MarcoTolo 00:47, 14 August 2007 (UTC)[reply]
Great, unless there is further discussion, we'll go with a full list of review articles under "Further reading" then. And unless there are dissenters, we generally agree that this satisfies the requirement of notability? AndrewGNF 00:57, 14 August 2007 (UTC)[reply]
- I would certainly say so, since it shows in a verifiable manner that the topic of the article is discussed in multiple high-quality reliable sources. Tim Vickers 01:07, 14 August 2007 (UTC)[reply]
- Sorry, on the issue of reviews vs. full reference list, we'll generate a couple of example pages to look at before we make a final determination. Didn't mean to summarize a consensus that didn't exist yet (my bad...) AndrewGNF 01:12, 14 August 2007 (UTC)[reply]
- From what I'm reading here, these proteins have been discussed plenty of places, so Wikipedia:N shouldn't be a problem. If there are 5-10 sources that will have discussed all (or most) of the proteins, I would use those sources to show notability. And if you're drawing directly from PD data, then Wikipedia:C is also not a problem. The bot should attribute its source in each article, then list a few references that contain a significant discussion of the subject, and you'll be fine. Thanks again for thinking through these issues with me. It looks like you're doing something very interesting here! -- But|seriously|folks 02:50, 14 August 2007 (UTC)[reply]
- Is that all the issues here, or is there more to sort out before we continue? Matt/TheFearow (Talk) (Contribs) (Bot) 04:52, 14 August 2007 (UTC)[reply]
- I have posted a mock-up of potential reference solutions in a subpage of my sandbox based on PPARG. User:JonSDSUGrad/Sandbox/PPARG uses the discussion page to add a further reading section (a discussed possibility) and User:JonSDSUGrad/Sandbox/PPARG-2 places the 'Further Reading' at the bottom of the page. This is how they would look when the bot created a new page without any additional human edits. Take a look and let us know if either one is an acceptable solution. JonSDSUGrad 06:02, 14 August 2007 (UTC)[reply]
- I prefer the latter. -- But|seriously|folks 07:32, 14 August 2007 (UTC)[reply]
- <aol>Me too!</aol> — Coren (talk) 12:24, 14 August 2007 (UTC)[reply]
- I too like the second one better, but just to be clear, this mock up only displays the review articles associated with this gene. In total, there are 341 primary and research articles associated with this gene, and if we wanted to display them all (I don't), then that probably would be better in a discussion page. (For the non-biologists, review articles are summaries of knowledge on a particular topic, in contrast to primary research articles which report new findings.) AndrewGNF 14:49, 14 August 2007 (UTC)[reply]
- Version 2 please. Tim Vickers 16:06, 14 August 2007 (UTC)[reply]
- I also prefer version 2. A small note: "Further Reading" should be capitalized as "Further reading". Also, I see that 16 references are included -- could we reduce that number? --Arcadian 20:03, 14 August 2007 (UTC)[reply]
(de-indent) Version 2 is the better one, but I have a handful of questions:
- 1. Can we stick with the Lastname Initials format (i.e. Chen AR, Wilberforce DE) that PubMed follows - and Diberri uses in his template tool - for the author list?
- 2. For tidiness, the "more than five authors triggers a first three authors + et al" rule helps.
- 3. Personally, I'd like to see a bit more complete refs (doi (if available) and page numbers, especially).
- 4. Can we use the
((refist))
template? I've made a derivative mock-up here.
-- MarcoTolo 20:27, 14 August 2007 (UTC)[reply]
- yes, "Further Reading" will be changed to "Further reading."
- yes, we will look into using the Diberri tool for formatting references (which also solves the long-author-list, page numbers, and doi issues).
- yes, will use the
((refist))
template and also the two-column formatting for the "Further reading" section.
- Thanks for all these great suggestions... AndrewGNF 22:49, 14 August 2007 (UTC)[reply]
- Diberri's tool just grabs the XML file from pubmed, which is exactly what we are doing, so I'll work to have it format the same way - not a big issue there. :) I wonder if he has a specs file on all the formatting guidelines he uses in his tool. I could just have PBB submit a bunch of PMIDs to his tool, but I don't think we want PBB reliant on too many other websites and I doubt Diberri would like having that many hits coming at him during an extensive update. I've almost got the whole thing coded anyway - just a simple matter of adding in a few more data grabs and formatting practices.
- I'm doin' my best to get these additions coded in, but I would expect a couple days turn-around at least. In the meantime, please keep the suggestions coming! JonSDSUGrad 23:52, 14 August 2007 (UTC)[reply]
- I wondered the same thing about direct queries - Diberri's using dynamic DNS, which implies that he's hosting this from home.... we probably shouldn't slashdot him <grin>. In any case, thanks for all your work on the bot coding - it's coming along nicely. -- MarcoTolo 23:59, 14 August 2007 (UTC)[reply]
- Something to check is how your code deals with articles with a volume number but no issue number. I've noticed a couple of times with Annual Review articles that this can break the Diberri tool and causes the Ref to to be generated without the publication year. Tim Vickers 00:05, 15 August 2007 (UTC)[reply]
- I spent an hour one afternoon trying to figure out what caused the occasional "disappearing year" problem when using Diberri's tool - can't believe I missed that.... -- MarcoTolo 00:14, 15 August 2007 (UTC)[reply]
- Could someone give me a MPID that has that problem - I'll look at the xml and see what is going on. That way I can be sure to not have that error occur. Thanks. :) JonSDSUGrad 04:34, 15 August 2007 (UTC)[reply]
- PMID 10966474 caused the error for me before, but seems to be working OK at the moment. Perhaps this bug has been fixed? Tim Vickers 14:06, 15 August 2007 (UTC)[reply]
- I've just run a half-dozen PMIDs that have caused me problems in the past and all appear to be properly handled at this point. -- MarcoTolo 02:44, 16 August 2007 (UTC)[reply]
In my opinion, the "Function" section of the infobox is disproportionately large. Could we default it to hide, using Javascript, as in Template:Navbox generic? Or, if that's not feasible in the short term, perhaps it would be more appropriate to omit it, and include a single link to where these classifications could be found. --Arcadian 20:03, 14 August 2007 (UTC)[reply]
- Pretty cool. Yes, I think we can definitely commit to looking at this in the future. Of course, if someone wants to help figuring out the exact syntax, that would greatly help us get it into the next version of the bot. ;) Regardless, I'd be in favor of keeping the list of all gene functions -- I think it's an important part of what someone would want to see on a gene page. AndrewGNF 22:58, 14 August 2007 (UTC)[reply]
- I think this would just be a matter of adding a few more things into the Template:GNF_Protein_box - I think we can use some of the code from the Navbox and get this change to occur without modifying PBB's code. JonSDSUGrad 23:54, 14 August 2007 (UTC)[reply]
- This stuff looks great, but the test on amyloid precursor protein popped up on my watchlist and ideally the function section could really be trimmed altogether, rather than just hidden. Some of the function listings are useful, and others - for example, APP gets listed as having the functions 'binding' and 'cytoplasm' (?) - are just so nonspecific that I can't imagine a reader learning anything from the presence of those links. Surely we can stick to functions lower in the hierarchy ('ion binding', etc)? Opabinia regalis 04:43, 15 August 2007 (UTC)[reply]
- I have a modified test version of Template:GNF_Protein_box that collapses the function section. You can see the template here: Template:GNF Protein box/sandbox2 and see it in practice here: User:JonSDSUGrad/Sandbox/TEST6_Vascular_endothelial_growth_factor. Things to also note on this page are the cleaned up references section and Further reading section (still working on a few small things), also I got that pesky expression pattern title to disappear when no image is present. So take a look and let us know if the changes look good. I'll have another example that fixes up the citations in the further reading section a bit better too (such as adding in et al for more than 3 authors) in the near future. JonSDSUGrad 08:43, 15 August 2007 (UTC)[reply]
- <Additional comment> Oh and the data that goes into the function box comes from Gene Ontology terms. At the moment we are combining a few different types of terms into the Function heading, that is why things like cytoplasm show up as a function rather than a cellular component, which is what it really is (Although maybe you could loosely interpret function as existing in the cytoplasm :D ). :: So for the moment we've taken a small shortcut for simplicity - we may be able to expand the function section in the future though. I'll leave that up to Andrew. I feel that the information is important, the format just may be a little off. :) JonSDSUGrad 09:08, 15 August 2007 (UTC)[reply]
- Additional follow up to Opabinia regalis' comment on the GO list. Jon alluded to the fact that Gene Ontology is composed of three main branches: Molecular Function, Biological Process, and Cellular Component. We've combined them all in one flat section, and to err on the side of completeness, have included all terms regardless of generality or specificity. Realizing that completeness is not the end all and be all of WP, we're happy to trim the list if there is consensus on what criteria to use. Though I agree with MarcoTolo's comment above (in the context of references) that we want to avoid arbitrary "implied selections" . Also keep in mind that we've started with *very* well characterized genes, which will tend to have the longest list of annotated GO functions. The vast majority of genes will have lists that will be significantly shorter. Cheers,AndrewGNF 16:33, 15 August 2007 (UTC)[reply]
Can someone with more wikicode experience take a look at a few browser formatting inconsistencies? It looks like the two-column, smaller-text references format that MarcoTolo mocked up doesn't render correctly in IE. (I'm testing with IE6 on XP, and it shows up single column, normal-sized.) Also, in the semi-automated edit of Amyloid precursor protein, IE appears to have issues with text flowing nicely to the side of the update infobox. If anyone has any ideas on how to fix these issues, that would be great... Thanks,AndrewGNF 20:31, 15 August 2007 (UTC)[reply]
- Am I allowed to plug FireFox at this point? :) Tim Vickers 02:56, 16 August 2007 (UTC)[reply]
- (ec) Rats - I'd conveniently managed to forget that IE doesn't support the draft CSS 3 column attribute. After a few quick tests:
- Works: Firefox, Mozilla, Safari
- Small font works, no multi-column: Opera
- Neither font size nor columns work: IE 6
- I don't have a copy of IE 7 laying around, but I'm guessing it's either in the "nothing" category or the "Opera-like" group. Hmm.... let me poke around and see if anyone has come up with a reasonable solution for this problem. -- MarcoTolo 02:58, 16 August 2007 (UTC)[reply]
- Hmmm, are you using ((refbegin)) and ((refend)) tags? They are reasonably cross-browser, and the ideal template for your use. Matt/TheFearow (Talk) (Contribs) (Bot) 04:10, 16 August 2007 (UTC)[reply]
- I took the liberty of editing MarcoTolo's mock up (hope that isn't a faux pas) and still doesn't render correctly with IE6 on XP. FF looks great... (same problem on Elephant so I assume it's not an issue of how we're using the template...) AndrewGNF 05:10, 16 August 2007 (UTC)[reply]
- IE 7 on Vista also falls into the "Neither font size nor columns work" category for me. —METS501 (talk) 05:38, 16 August 2007 (UTC)[reply]
- I'm not sure why I'm surprised that IE 7 doesn't support the
column-count:x
attribute.... <grin> -- MarcoTolo 17:29, 16 August 2007 (UTC)[reply]
(un-indent) I'm probably behind the curve here, but, after several days of fiddling with various formatting, I've come to the conclusion that no clean-and-easy solution exists for this problem at this point (I'd love to hear other opinions here, but I'm not especially hopeful); the lack of IE 7 support for the draft CSS 3 column option is problematic. However, since supporting browsers which do support the columns attribute doesn't break the non-supporting browsers (i.e. IE), I suggest we stick with using ((reflist|2))
and ((refbegin|2))
/ ((refend))
templates, with folks using IE seeing single-column lists until their browsers support the new option. -- MarcoTolo 21:09, 20 August 2007 (UTC)[reply]
- Seems like a reasonable plan. I'll have the templates integrated in the next test run, which should hopefully be towards the end of this week. JonSDSUGrad 07:38, 22 August 2007 (UTC)[reply]
Can you make it so that it goes "_____ is a gene" on the first line, or somesuch, with the name boldened, to minimise tidying up? It would save some time -85.210.30.61 05:47, 19 August 2007 (UTC)[reply]
- This is obviously easy to do. Though I'm a bit concerned that the "This is a gene" statement would be too redundant with the summary that is already included. Anyway, if others would chime in, that would be great. If there's consensus, then we'll add it. For reference, here are the eight pages that have been created so far in the main WP namespace: MMP9 HIF1A PTGS2 NFKB1 TGFB1 PPARG AKT1 MAPK1. Cheers, AndrewGNF 18:36, 19 August 2007 (UTC)[reply]
It might work to say "X is a human gene." as a first line. This would be useful information, sice the spp isn't immediately apparent. Tim Vickers 18:50, 19 August 2007 (UTC)[reply]
- Two notes: (1) per Wikipedia:Lead section, we shouldn't have an outline level ("==Summary==") before the the start of the text. (2) Even if the text being added is public domain, we still need to identify the source. I'd recommend using a reference tag (details at Wikipedia:Footnotes) at the end of the paragraph. --Arcadian 19:26, 19 August 2007 (UTC)[reply]
- The sources are there, just not inline. Matt/TheFearow (Talk) (Contribs) (Bot) 21:30, 19 August 2007 (UTC)[reply]
- A link to the prose exists, but not a reference giving credit. I've provided an example of one way to address the issue at MAPK1. --Arcadian 22:07, 19 August 2007 (UTC)[reply]
- So you know, the latest versions (the sandbox test ones) already have that style implemented, just not the summary header removed. Matt/TheFearow (Talk) (Contribs) (Bot) 22:53, 19 August 2007 (UTC)[reply]
- I'm altering the way PBB adds data to the page. Instead of using HTML comments, we thought we could streamline things a bit by using templates. In the summary template (Template:PBB_Summary), we are putting an optional section header in (currently says Summary), which could be changed or omitted at any time. I had added the Summary section to allow easier integration with pre-existing pages. It could be easily changed/removed though. Also, adding XXX is a human gene to the beginning of every summary is certainly do-able, however, there is no guarantee that the statement won't be repetitious with the summary text following the statement. But maybe the risk of a little repetition is worth making sure that basic information is present. We could do something like this: AKT1 Example where the introduction is just one line and then the summary comes right below it. I'm certainly no expert in wiki formatting guidelines though so You'll have to tell me if it works or not. :) JonSDSUGrad 18:20, 20 August 2007 (UTC)[reply]
- I'd favor taking the risk of repetition, which can always be cleaned up when it occurs, and add this info as part of the entry text instead of separated be a "Summary" header. Tim Vickers 18:25, 20 August 2007 (UTC)[reply]
- One tiny note -- in the phrase "also known as AKT1 is a human gene", there should be a comma before the "is". But other than that, I think we're on the right path. --Arcadian 18:34, 20 August 2007 (UTC)[reply]
Since there is no mouse Uniprot ortholog, should that field be under "identifiers"? (If that discussion would be more appropriate at Template talk:GNF Protein box, let me know.) --Arcadian 18:40, 20 August 2007 (UTC)[reply]
- Thanks for catching that. There *should* be a corresponding mouse uniprot ID, so there's a hiccup in our pipeline somewhere. We'll get that fixed in the next update... Cheers, AndrewGNF 22:29, 20 August 2007 (UTC)[reply]
I'm looking at the updated Amyloid precursor protein and TGFB1, and I don't think the image for the expression pattern belongs on the page. I'm also concerned about the addition of 27 and 18 journal links respectively -- can we throttle that down to three or five for new pages, and zero for existing ones? --Arcadian 01:37, 11 September 2007 (UTC)[reply]
- Nothing wrong with lots of references, as long as they are all relevant. Tim Vickers 04:11, 11 September 2007 (UTC)[reply]
- Yeah, I'm open to hearing people's thoughts on the expression pattern. Personally, I feel like this information is a great source of gene annotation. Knowing where a gene is expressed tells you a lot about what that gene's function is. (A gene that is expressed in thymus is likely involved in immune function. A gene expressed in adipose tissue is likely involved in energy metabolism. Etc etc...) Anyway, anyone else want to chime in on this issue? On the issue of references I'm also open to restricting further as long as we can define some sort of systematic criteria to avoid the "implied selection" issue mentioned above. Right now, we are using all references linked by NCBI which are also review articles as indexed in Pubmed. AndrewGNF 04:31, 11 September 2007 (UTC)[reply]
- As for References (and this is just my personal opinion as a graduate student), I generally don't like to have a flood of unorganized references about a particular subject. However, if those references are organized in some fashion (such as references for the article, then a list of papers that I could continue my research with i.e. Further Reading) then I like having quite a few of them. Nothing bugs me more than to find an informational hub that doesn't have information :). Also, for those willing to supply more information to the wikipage I bet those papers would make for a pretty good start. I'm not a wiki admin, but I always thought encyclopedias were about empowering people with information relevant to a subject. If I had handpicked those articles because I thought they were important, would they still have been deleted?
- Now as a programmer of PBB, I have good news about those extraneous citations - they can be turned off. You just have to change the value on the PBB_Controls template to "no". In the case of Amyloid precursor protein it is possible to just delete that extra list down to just a few of the good ones (I'm sure there must be a couple of high relevance to the protein) and then set the control box to 'no'. The next time that PBB updates the page, it will not touch the citations box.
- I would also like to submit that PBB never edits pre-existing pages without human permission - In the case of APP, a human (me) told PBB to place all those citations on the wikipage. It is presumed then that as a human operator I could pick and choose the citations I wanted to keep on that page and delete the rest - If I did find a page or two that I decided to delete (perhaps due to relevancy - Time to let NCBI know that their database is a little off), then I would turn the citation updates off so that PBB wouldn't change my work in a future update. So before we just delete the whole lot of Further Reading citations, I would like to manually review the citations for relevancy (this would be a good test of the system to see just how good the review articles are). I'll let you know how it goes. JonSDSUGrad 18:39, 11 September 2007 (UTC)[reply]
- UPDATE: I have hence skimmed through the 27 articles in the further reading section of Amyloid precursor protein, 20 of them seem right on the mark for information about this protein. They also have a heavy relevance to Alzheimer's Disease. Of the other 7, 2 were in a foreign language, 1 I was unable to read, and the other 4 seemed more about general molecules involved in Alzheimer's Disease and not specifically about Amyloid precursor protein. I am going to remove the citations referring to the less relevant articles and put the rest of them back in. Someone might want to move the 2 Japanese papers into a Japanese version of the Amyloid precursor protein page, but that is a bit beyond me. If there are any more issues about it, we should probably move them over to the Talk:Amyloid_precursor_protein page. JonSDSUGrad 19:09, 11 September 2007 (UTC)[reply]
Hi Everyone! I am pleased to announce that we have completed a second test run of Protein Box Bot! You can find the latest log file here: User:ProteinBoxBot/PBB_Log_Wiki_Live_Run3_Char_Fix. Note that we tried a test run before that and encountered a character encoding issue (which has hence been resolved). Please take a look at the 10 pages that were updated and let us know what you think (Keep in mind that APOE and APP were merged by hand with the bot output in the first test run). Is there anything else that we could do to improve the bot's output? JonSDSUGrad 17:47, 11 September 2007 (UTC)[reply]
I'd like to officially request permission to move forward with PBB in "production mode". To summarize the results of the trial:
- 33 of the most well-cited genes were chosen.
- 8 of these had no WP namespace conflicts (based on gene symbol, description, or aliases). PBB created new pages for these 8 genes:
- The 25 remaining genes had a namespace conflict with an existing page. These could be existing gene pages, or conflicts with "homonyms". PBB made no edits to these pages
- I took 2 of these 25 genes and manually merged PBB content with the existing gene pages. These are not officially bot edits, but illustrate how we also add value to existing pages. The two example pages are Apolipoprotein E and Amyloid precursor protein. All edits like these require human control.
- Note that by adding the PBB_Controls template to these pages when merging, we also make these entries "PBB-aware". On future runs of PBB when updates of the data are available, PBB will modify the appropriate portions of these pages based on the flags set in the PBB_Controls template. Flags can be set individually for the protein_box, the Entrez gene summary, and the citation list.
Any remaining concerns before we move forward? AndrewGNF 19:07, 12 September 2007 (UTC)[reply]
- Looks good to me. Tim Vickers 19:14, 12 September 2007 (UTC)[reply]
- Looks good to me as well, you're free to proceed as requested at 10 edits per minute to create and update pages.
Approved. --ST47Talk·Desk 18:11, 14 September 2007 (UTC)[reply]
- (Note that a bot flag has been approved. When granted, it will appear in this log. --ST47Talk·Desk 18:14, 14 September 2007 (UTC))[reply]
- The above discussion is preserved as an archive of the debate. Please do not modify it. Subsequent comments should be made in a new section.