So, how would WeSay run on this relatively slow (900 Mhz), relatively low-wattage machine?  Getting Windows installed and running, the Eee felt very sluggish.  Because of delays, I found it easy to make errors when using a web browser.  Very slowly, I grabbed .net 3.5 sp1, then WeSay build 1451.  Very slowly, I ran the installer.  By this time, I wasn’t expecting to have a happy WeSay experience.

To my surprise, WeSay runs GREAT on this box! Changing records, finding words, and bringing up new tasks were all pretty snappy.

Note: I only did a quick walk-through using the sample data, so please don’t go out and purchase a hundred Eee’s to run WeSay based on this blog post.

• See which words are missing from the dictionary, and add them along with a definition.
• Jump into the entry screen for a word in WeSay to do more advanced editing.
• Point to a word and see a list of similar words they might choose instead (thesaurus lookup).

A linguist working with the group will want to:

• Click on an unfamiliar word and see the full dictionary article for it.

If the language has any affixation, both users will need to be able to:

• see a list of entries, ordered by how similar their spellings are to the word be investigated.
• find words based on their inflected/derived forms, not just by the citation form in the dictionary.
• add variants to the word so that it is clear that this form is covered by the dictionary, and make it easier to lookup next time.

The first round of this work is now available for other application developers to use (the italicized bullet items above will come in some future version).

To help developers add these features to their programs, I’ve built a little sample application so they can see what’s possible and how to do it. Here’s a little crummy video showing it:

[kml_flashembed movie="http://www.wesay.org/downloads/movies/dsDemo/dsDemo.swf" height="464" width="512" scale="noborder"/]

A few technical details for developers

Currently, I’ve implemented support for .net applications to make use of these services. But support via any language, via xml-rpc, should be easy to add when needed. All .net applications need to do is get our Palaso library and use the DictionaryAccessor class. You currently need to tell it where on the user’s machine to find WeSay, and where the dictionary is that you’ll be accessing.

Here’s some code to show what it takes add this ability to a .net application:

Getting some HTML of matching entries to show in a WebBrowser control

DictionaryAccessor dictionary = new DictionaryAccessor("c:docsnoosupunoosupu.lift", "c:program fileswesaywesapp.exe");

string[] forms;

string[] ids;

dictionary.GetMatchingEntries(writingSystemIdForWords, "foobar",

       FindMethods.Exact, out ids, out forms);

string html = dictionary.GetHtmlForEntries(ids);

Adding a new word

dictionary.AddEntry(writingSystemIdForWords, wordBox.Text,

                    writingSystemIdForDefinitions,definitionBox.Text,

                    writingSystemIdForWords, exampleBox.Text);

Ok, so you get the idea that this will be a very easy service to add to your .net program.

A plug here for .net 3’s WCF (Windows Communication Framework), which made implementing this a very nice experience.

Update:  This has now been re-written to use cross language, cross platform “XML-RPC”messaging.  So programs written in non-.net languages can now participate.

To see how this works, let’s add some of the domains that would apply to a Papua New Guinean bilum. 

To look for a domain, we click in a box and start typing. First, we start typing "crafts" and see a domain matching that word:

Next, since bilums are used to transport firewood, we type that in.  A promising domain appears, fuel.  But is that right? We don’t want to say that bilums are something you burn.  Happily, when we point to the word Fuel, WeSay displays a description confirming that domain is also for things you use in collecting fuel.

Finally, since at least one of my kids has slept in a bilum, we should find a domain for that.  Here we’ll pretend we know the domain number, and just start typing that until we see Bed:

I had fun finding domains for a few words. I hope you do too.

In this screen shot we see the three ways you can now specify sorting:

Sort like another language

If the text sorts just like some major language, just select that language in the list and you’re done.

Custom Simple

Many languages based on Latin characters introduce a small number of "special characters" used to represent sounds not covered by A-Z, like a barred i. In these situations, you can specify the rules just like you do in many existing apps, like Toolbox and Lexique Pro. When you choose "custom simple", the rules box is filled with rules needed to sort English. You can enter vernacular works in the "Test Sort" area:

We want the barred-i to sort just after i, so we add it to the rules and click the button:

Normally, these secondary distinctions are enough.  But for some languages, tertiary distinctions are needed. We get these in the simple rules by using parentheses. Consider this list of words:

Now, imagine we want the upper-case words to sort together.  We need to add in another level of distinction, so that case can trump the accents.  We do this by adding parentheses around all case pairs, and putting the two sets of e’s on the same line:

Eric has written up the details on our wiki.

Custom ICU rules

For languages that need them, WeSay also supports ICU tailorings, which look like this:

& C < č <<< Č < ć <<< Ć   –for Serbian (Latin) or Croatian

Like many features of WeSay, this simple-to-advanced collating actually lives in our "Palaso Library", which is of course open-source and can be included in other programs.  Thus we foresee a day soon when the setup you do in one program (e.g. WeSay) will be trivially usable in other language-development tools.

Happy sorting!

For example, while we have always put a lot of effort into keeping WeSay clean and easy, its companion WeSay Configuration Tool has always been about being just-functional-enough-to-get-by.

So I spent two days remodeling it, trying to get it clean and easier to navigate. Here’s the before:

(click to see full size)

and after:

Yes, lacking any artistic ability myself, I’ve ripped off a bunch of open-source Tango icons, tweaked them with the indispensable InkScape, and otherwise copied the latest Skype’s options dialog. There are still quite a number of rough edges in there, but this was a good step.

When you choose a picture, it is copied to the “pictures” sub-folder of the WeSay folder.  So everything stays together.

Some more things to do someday:

1) reduce the picture size, if we can determine that this is a good thing to do. This would make backing up a lot less expensive.

2) Show the pictures in the article preview (top of the window) and add it to the html output.

Here’s a preview of the new relations field type. Here, it is being used to add a “base form” field (i.e. the inverse of a subentry).  Notice that you pick the base form by beginning to type it. You get the same fuzzy lookup as with the Find box.  Also notice that, to help you pick the correct one, WeSay shows the gloss or definition of each word as a yellow tool-tip.

If the word you want isn’t in the dictionary yet, you can click the button with the plus sign to add it. 

• Finding the WeSay Files on Disk
• Using a zip program
• Using an email program

Instead, he just clicks on this big email button:

This pops open your email client, where all you have to do is click “Send”. Here’s what it looks like on my machine, using MS Outlook:

Of course, many of our users will not have email, even occasionally. But for those that do, this may come in handy.

WeSay Configuration pt1

A quick look at making a new project.

WeSay Configuration pt 2

A walk-through of setting up a WeSay project using the WeSay Config tool.

WeSay Dictionary Tab

A detailed walk through the elements of the Dictionary Tab.

WeSay Wordlist Task

A walk-through of the task designed to elicit vernacular words based on a list of foreign words.

And still to come:

Semantic Domain Gathering Task

Missing Field Task

All about WeSay files

Overview of the WeSay project

What’s the idea? Who is the intended audience? How does WeSay fit in with more traditional linguistic applications?

—–

From the heading off the inevitable technical questions department: I produced these using Camtasia Studio (of which I’m not a big fan) + GoldWave to eliminate the hiss, outputting to Flash using JPEG encoding at 20% quality, 32k mp3 audio.