Thanks for the detailed feedback. Yes, there actually is a standard place to list subsidiaries - "exhibit 21a". However, in pretty much all cases in this crowdsourcing game, the reason they're in the game is because we couldn't scrape them, and that's usually because there's no exhibit 21a.
Your other points (varieties, I think, of "it's more complicated than the form allows me to report") illustrate a real problem with crowdsourcing like this. Do we add a button for "It explicitly states there are no subsidiaries", and another for "It doesn't mention subsidiaries"? How about a separate confidence level for jurisdiction as well as name? Eventually we end up with a forest of buttons and options, and we quickly end up with a game that only people who are detailed and dedicated are going to be able to consider playing. It stops being a game, and just becomes a "relationship editing tool" (which we are developing at the same time, but trying to work out how we induct people into using it -- volunteers / ideas welcome!) There should be a place for crowdsourcing for lower quality, higher quantity data extraction, but the only way I can think of removing these grey areas / cognitive blocks would be to decompose the task into much smaller parts (maybe "draw a box around any sentence mentioning subsidiaries", "click boxes mentioning subsidiaries and jurisdictions" or something like that). Even then I wonder if a crowdsourcing game is even going to much good at this kind of natural language parsing. And it would take vast amounts of work to get right for reducing returns. On the other hand, I hope we can tweak the current game so it's clear that if you're not sure, or you hit any block at all, it's OK just to continue. I think that should work OK, but haven't yet found a way to make it really clear in the UI. Seb On 3 October 2013 19:21, James McKinney <[email protected]> wrote: > I continued playing until I found a document with relevant info near the > search term "subsidiar", and it seems there is no standard place to disclose > subsidiaries. For example, the following document has a single phrase at the > start of Part 1 mentioning the company's two wholly owned subsidiaries (I > didn't check the document to see if it discloses any non-wholly-owned > subsidiaries). > > http://www.sec.gov/Archives/edgar/data/1349976/000114420413044838/v350454_10k.htm > > I've found a document where in Item 15 it lists "Active Subsidiaries of the > Registrant", which can then be searched for in the document to determine > ownership %. Another document calls it "List of Subsidiaries" in Item 15. A > strange thing is that a subsidiary StarTrak Systems, LLC was sold to ORBCOMM > Inc., yet it's also described as being a wholly-owned subsidiary - my > understanding is that the deal has not yet been closed. > > http://www.sec.gov/Archives/edgar/data/98618/000009861813000011/k10_063013.htm > > The same document lists subsidiaries under a "Principles of Consolidation" > heading, in case that heading appears in other documents. Similar headings in > further documents include "Basis of Presentation" and "Basis of Presentation > and Consolidation". It may be worthwhile to compile a list of these. > "Percentage of Ownership" seems to be a popular column heading. > > It may be worthwhile to add a field to the data collection form to express > the user's certainty as to whether all wholly-owned subsidiaries mentioned in > the document have been collected. For example, one document in the "forward > looking statements" before Part 1 explicitly says that the company has five > wholly-owned subsidiaries, which makes me certain that I've found all five: > > http://www.sec.gov/Archives/edgar/data/1425173/000101968713003726/novagen_10k-123112.htm > > Another document helpfully says "We do not have any subsidiaries." but it > seems most are not so helpful. If there is no mention of subsidiaries, does > that mean there are no subsidiaries? Should a button be added for cases where > no info can be found, or should I just click the "it's complicated" button? > In once case the list is described as "previously filed", with no other > mention of subsidiaries. > > Similarly, this document says it has subsidiaries, but doesn't name them. > Should a button be added to report these cases, or should I just click the > "it's complicated" button? > http://www.sec.gov/Archives/edgar/data/1435936/000143593613000153/f10k_20130531r5.htm > > Have you noticed cases where in one part of the document, a subsidiary is > described as wholly-owned, but is later described as not wholly-owned? e.g. > How much attention should be paid to verb tenses? > > Some documents don't give the jurisdiction of subsidiaries. What to do? I > can't submit the form without filling in the jurisdiction, but I would expect > that a partially filled form is better than no form? For example: > > http://www.sec.gov/Archives/edgar/data/1020477/000135448813004921/robk_10k.htm > http://www.sec.gov/Archives/edgar/data/825322/000121390013005438/f10k2013_mphase.htm > http://www.sec.gov/Archives/edgar/data/896429/000118811213002827/t77452_10k.htm > > The instructions should perhaps warn that some documents mention the > subsidiaries of other companies, so it's important to check that you're > reporting about the right company. Most confusingly, the following document > starts talking about companies with entirely different names as though they > are the subject of the document: > > http://www.sec.gov/Archives/edgar/data/1496741/000116169713000700/form_10-k.htm > > Lastly, I'd eliminate one click and put the "Step 1" content after the other > content on "Step 3". > > James > > On 2013-10-03, at 1:09 PM, James McKinney wrote: > >> Hi Seb, >> >> I've tried the game, but I'm getting very long (50+ page) documents that >> don't seem to have any relevant information near search terms like >> "subsidiar", "own", "wholly", etc. It may be the case that these documents >> have no subsidiary information. However, I also notice that the two >> documents I've looked at so far have identical tables of contents. Could you >> provide instructions as to which section(s) within this (at least somewhat) >> standard table of contents would contain the relevant information, to avoid >> having to carefully read 50+ pages? >> >> Examples: >> >> http://www.sec.gov/Archives/edgar/data/50292/000091431713000879/form10k-131780_ieh.htm >> http://www.sec.gov/Archives/edgar/data/874710/000114420413052878/v355226_10k.htm >> >> Thanks! >> >> James >> >> On 2013-10-03, at 7:39 AM, Seb Bacon wrote: >> >>> Hi all >>> >>> If you fancy helping turn unscrapeable info about corporate control >>> networks into open data, consider having a go at our >>> fun-for-all-the-family game! >>> >>> Background: >>> http://blog.opencorporates.com/2013/10/02/help-uncover-corporate-networks/ >>> >>> Game: >>> http://opencorporates.com/games/secfilings >>> >>> We know it's a bit rough around the edges, but our goal with this >>> iteration was to make it good enough that 100 people could >>> successfully use it to add relationships. If you can give it a go, >>> ask your partner, your children, your grandparents, that would be >>> great! >>> >>> Feedback to [email protected] please >>> >>> Thanks >>> >>> Seb >>> >>> >>> -- >>> skype: seb.bacon >>> mobile: 07790 939224 >>> land: 01531 671074 >>> >>> _______________________________________________ >>> developers-public mailing list >>> [email protected] >>> https://secure.mysociety.org/admin/lists/mailman/listinfo/developers-public >>> >>> Unsubscribe: >>> https://secure.mysociety.org/admin/lists/mailman/options/developers-public/james%40opennorth.ca >> > -- skype: seb.bacon mobile: 07790 939224 land: 01531 671074 _______________________________________________ developers-public mailing list [email protected] https://secure.mysociety.org/admin/lists/mailman/listinfo/developers-public Unsubscribe: https://secure.mysociety.org/admin/lists/mailman/options/developers-public/archive%40mail-archive.com
