Tajik on Google Maps!

I wrote a while back about Google Maps occasionally labelling objects (not located in Kyrgyzstan) in Kyrgyz. I have not seen Kyrgyz on maps for a long time... but today I saw a label... in Tajik! The Wuhan International Airport was shown to me, for unknown reasons, as Фурудгоҳи байналмилалии вуҳон! Perhaps something is going on that I am not aware of...

P.S. P.S. More, now in a mixed (Cyrillic + Arabic) alphabet

The Freat Soviet Eccyclopedia

More wonders of OCR:

Bibliographic information
Title: Freat Soviet Eccyclopedia
Published: 1980

https://books.google.com/books?id=vnOFYI3g-N4C

Google Maps Sprachbund

A while back, as an experiment, I set my interface language for Google products to Macedonian. In reality, only a small portion of all the Google label, legend, and message texts may be available in that language, so I also had to pick a "backup" languages, for which I choose Bulgarian (I think by default, as it is the closest related language).

So you would imagine that the text I would see e.g. in Google Maps would be a mix of Macedonian and Bulgarian, plus the original language of the region being mapped? Wrong! There are actually four languages I often see. And the fourth is... Kyrgyz!

In the example above, the standard Google Maps messages are in Macedonian (e.g. "Пребарувајте во близина", i.e. "Search in the vicinity"), the country name is also in Macedonian ("Соединети Американски Држави" = the USA), while the object name is a translation/transcription of the "West Side Avenue Light Rail Station" into a curious mix of languages:

• "Запад" must be the Bulgarian or Macedonian translation for "West" (the proper adjective form would actually be "Западна", but we cannot expect that!);
• "Сайд Авеню Лайт Рейл" is the Bulgarian transcription for "Side West Light Rail". (We know that it's Bulgarian and not Macedonian because of the letters й and ю; Macedonian would use ј and ју (or just и), respectively. In principle it could be Russian too);
• and finally "станциясы" is the Kyrgyz for "the station". (The word would be spelled the same in Kazakh, but for many other place labels I saw words that were definitely Kyrgyz, e.g. "аэропорту" rather than "аэропорты" for "the aeroport"; Kyrgyz has something called "labial harmony", which Kazakh doesn't).

Printing Google / Sogou maps

Have you had this experience? You've centered and zoomed the map of an interesting area on Google Maps, or a similar site (e.g., maps.sogou.com) and wanted to print it. Some map sites actually have a "print" button, or you can just use the "print screen" / "screen capture" button of your operating system's window manager to save the image into a PNG file to print it later (directly, or after including it and the caption text into an HTML / PDF / LaTeX file). But, whatever you do, chances are that the printed image will be almost un-readable: Google, Sogou, and other map sites optimize the image for on-screen viewing, and printing is just an afterthought.

There is, however, an easy way to improve the legibility of the map image. The idea is, an image is hard to read because there is not much contrast to it: both the brightest pixels (the background) and the darkest pixels (text and lines) are merely shades of grey (or, generally, some intermediate RGB colors), rather than good old printable black and white.

So the solution for improving the printable quality is, basically, as follows: figure what color you want to convert into black, what color you want to convert into white, and linearly transform all colors in between.

There are of course plenty of image manipulation tools on the market. The one I happen to use is ImageMagick, and sure enough, it has a command for it.

To find the colors to be made into black and white, you probably can zoom in on your screenshot image in some image editing program, and take a look at what the RGB values for the darkest pixels (parts of text, or lines showing roads etc) are, and what the RGB values for the brightest pixels (the not-quite-white background). In my case, I did it with cropping out a small section of the image with "convert -crop", and then querying the range of colors with "identify -verbose". Having chosen the colors this way, all you need to do is a single command:

convert original-map.png -level-colors 'rgb(144,144,10),rgb(241,238,232)' printable-map.png

Enjoy!

Before	After

And what did they write about railways in 1624?

When using Google Books yesterday, I was offered to fill in a survey. Besides asking whether I am aware of various new ways to sell e-books, they asked for a general feedback. Below, with minor edits, is what I wrote.

---

Google Books is a great tool of course, but there are quite a few things that can be improved: 1) Sometimes the font is too small - and zooming in the browser does not help (i.e., it's still illegible). 2) Old books are occasionally scanned in careless ways, especially when they are "centerfold" type pages that no one bothered to unfold prop erly. (Can't find a good example right now, but this one comes close, complete with the image of the scanner operators' fingers: Regni Sinensis la Tartaris Tyrannicle evastati depopulatique concinna Enarratio 3) A fair number of books are mis-dated, and the cool new NGram viewer has made it painfully obvious. Just ask, for example, "Who wrote about railways or railroads 300 years ago?" 4) It seems that the mechanism for Google Books "importing" reviews from the "usual places" elsewhere (i.e., Amazon.com, I reckon...) does not always work. E.g., the review existing this Roel Sterckx's book at Amazon does not show at the books's page at GB 5) How do you insert hyperlinks into reviews anyway? Even at Amazon you can insert a link to another Amazon product (via its ASIN); at the very least, Google Books should allow one to insert a link to another book at GB.

Google knows how to divide text into words

... in Chinese, that is. This is no small feat, because Chinese text, when written in the usual way (in Chinese characters) does not reflect in any way the division of text into words (with the exception of some very special cases, such as when transcribing foreign personal names into Chinese). When Chinese speakers need to write a sentence in Pinyin (Latin transcription), they often end up writing every syllable as a separate word, or, more rarely, run all words together. (The photo above shows both possibilities). Most automatic Chinese-characters-to-Pinyin converters also separate the transcription of all characters with spaces. Google Translate, however, appears to have a pretty good idea how to put spaces between words in Pinyin. Getting to Pinyin, though, is a bit tricky. To do it, one can enter a Chinese phrase, ask Google Translate to "translate" it into another version of Chinese (e.g., simplified to traditional), and click on the "Read phonetically" link below, which will give you the Pinyin transcription of the phrase. E.g., for "有可能朱棣立神道碑加工期间，发现龟趺脖子下裂缝而弃之" ("Perhaps, during Zhu Di's installation of the Sacred Way Stele, cracks were discovered under the neck of the stone tortoise [serving as as the pedestal] and it was abandoned") you get "Yǒu kěnéng zhūdì lì shéndàobēi jiāgōng qíjiān, fāxiàn guī fū bózi xià lièfèng ér qì zhī". Which I think is pretty good for a machine, although of course the name Zhu Di should be capitalized and written with a space, and I would probably write "guīfū" ("tortoise-shaped pedestal") as a single word.