Meta-question about FLEx : blank-slate verses strait-jacket

Fri Oct 14 01:40:03 UTC 2011

I’ll pitch in by noting two things about FLEx I’ve encountered while preparing a dictionary of Chitimacha, one positive, one negative. There is much more I could say in either direction, but these serve as good examples. First the positive: FLEx makes it very easy to format lexical entries in a dictionary to show precisely the parts of an entry you want, in the order you want them, with the punctuation you want, etc. This has made a big difference for me, because I’m working for two different audiences – the Chitimacha tribe, which wants a dictionary as bare-bones as possible, and linguists, who want data sources, cognates in related languages, morpheme breakdowns, original transcriptions if the word comes from an archival source, phonetic transcription, etc. With FLEx, I can easily export to a PDF and send it off to the Chitimacha tribe, and then immediately check a couple boxes, produce a second PDF, and send it off to a linguist working on areal features of languages in the region. Then, I can quickly export to LexiquePro, and from there produce a Word version for editing when it comes time to create a final manuscript. This dexterity in formatting and interoperability has been really helpful in meeting the demands of two very different communities.

An example of a negative is the list of morpheme types in the database. The list is strictly defined and uneditable. In the case of Chitimacha, this is particularly unfortunate, because the notion of root I’d like to use doesn’t appear to align with the way roots are designed to behave in FLEx. This wouldn’t be a big deal, except that the parser refuses to recognize roots when I’m analyzing texts if they aren’t categorized according to FLEx’s definition. Moreover (and I suppose this is a separate point), if you believe that ‘part of speech’ isn’t stored in the lexicon, but rather is inferred from the syntactic or pragmatic context (as linguists like David Gil, Kees Hengeveld, and Jan Rijkhoff seem to suggest), FLEx clearly isn’t adequate, since POS labels are stored in the lexical entry. In Chitimacha, many roots appear to be category-neutral, freely taking nominal, verbal, or adjectival suffixes. One solution to this problem is to create three entries (or more) for every one of these potentially category-neutral words. For native English-speaking learners of Chitimacha, perhaps this is the way to go, since they’re accustomed to the traditional parts of speech anyway. But as far as I’m concerned, this is descriptively inadequate, and thus not an accurate representation of the language. Creating a ‘neutral’ category in the part-of-speech list doesn’t help either, since I still want to know what part of speech a word when it’s instantiated in a specific context.

And I’ll add a third (or fourth): Overall I find FLEx very helpful, and I’ll continue to use it, but the latest version is still just excruciatingly slow at times, even though I’m running it with an Intel i7 processor. In addition to closing every other program, I find myself rebooting my machine about every 20-30 minutes. This is the single biggest detriment to my work, hands down.

From: clairebowern at gmail.com [mailto:clairebowern at gmail.com] On Behalf Of Claire Bowern
Sent: Thursday, October 13, 2011 7:30 PM
To: John Hatton
Cc: r-n-l-d at unimelb.edu.au
Subject: Re: Meta-question about FLEx : blank-slate verses strait-jacket

Having spent most of my field time in dongas of various states of termite-infestation, I have to agree, living in an empty flash house is so much harder - you can't just cook some damper on the coals and boil a billy when the power goes out.  And it makes the recordings really echoic...

I do hope everyone had a chance to read Beth’s reply, which addressed the inaccuracies of the original post.  I would be sorry if folks passed on such ideas as that FLEx isn't open source (you're confusing it with Toolbox), 

well, if something goes wrong with toolbox I can do stuff to the text files in a text editor if need be - the underlying data structures are highly recoverable. 

that it isn't for non-linguists (confusing with WeSay?), that it's impossible to collaborate, or that it’s still large and slow (it was largely re-written for speed and size over 2010).  What is clearly true is that web presence could be improved in order to make the facts more clear.

I think web presence is less the issue than user reports, unfortunately. I don't think I said it was for non-linguists, I implied (and I think your post confirms this) is that it's designed for someone with minimal linguistics background. I'm claiming that "making things easier" is actually just as likely (if not more likely) to lead to misanalyses and confusion. I have been told by FLEx developers that they see it as an ideal tool for field methods classes because they see it as good for someone with little training. 

But with those misunderstandings cleared up, I hope we can get on to the valid concerns and thoughtful questions in Claire's post. There are still some things in FLEx which will make it unattractive to many (no Mac version, requires clean consistent data on import, pre-Himmelman orientation, limits to interoperability, intentionally not a blank-slate like Toolbox).  I’d be happy to talk about any of these.  For now, I’d like to write a bit on this trade-off —between something which is more of blank slate and something which embodies some basic concepts in order to help with automation.

I’m not sure what Claire meant by saying that FLEx was for “non-linguists”, but maybe I agree after all.   Software design best-practice calls for identifying a core "persona". Others, below and above that persona can still use the program, but they aren't the target the designer holds in mind.  For FLEx, the target audience is not PhDs or PhD candidates, who appear to make up the bulk of folks on this list. FLEx’s core target persona is required to do a broad but often necessarily shallow amount of linguistic description as part of his job. 

oh, so females aren't part of the target audience? Good to know I guess....

We want to give him tools which guide him along, even at the cost of flexibility.  He benefits from a more guided system than would a Linguistics PhD candidate intensely researching one aspect of a language.  As the trailer-park analogies have shown, that researcher will trade a lot of pain for the flexibility he/she needs.  Hence the continued popularity of SIL's Toolbox among western academics, while so many outside that group have eagerly switched to FLEx.  And some western academics too, even on this list (they’re too smart to come out of that closet here J ). 

oh please. My original question was a genuine information question. Clearly some people like FLEx and are willing to put up with far more than I am in order to use it. I am interested in that. I am also interested in how FLEx fits with wider currents in language documentation and linguistic technology. This is no witch-hunt. 

I think comes down to this -- so many of the people on this list (academics and non-academics - I'm not interested in that distinction here) are making what may well be the *last and only record* of the languages they are working with. That's a responsibility, and it's important that we all do the best we can to make that happen. Now, if better language records are being produced with FLEx then I'm certainly not going to criticise that. But I would like to see some evidence that this is happening. Are materials being produced faster? Are they being shared more widely? Is less data being lost through computer crashes or disc malfunctions? Are we getting more accurate analyses? Are we increasing participation of native speakers in documenting their own languages? Are we archiving more?

Claire

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://listserv.linguistlist.org/pipermail/resource-network-linguistic-diversity/attachments/20111013/eed104bd/attachment.htm>