GNU-devel ELPA - guess-language

(require 'guess-language)

;; Optionally:
(setq guess-language-languages '(en de))
(setq guess-language-min-paragraph-length 35)

guess-language-languages defines the candidate languages that should be considered. It is recommended to only include languages that are actually used because this improves performance. Languages are identified using ISO 639-1 codes (see table below).

guess-language-min-paragraph-length specifies the minimal length that a paragraph needs to have before guess-language-mode starts guessing. Based on some informal tests texts shorter than 30 characters are not enough to give good results. However, above 40 characters the algorithm performs well. Of course, these numbers depend on the target language (some are easier to detect than others) and on the (number of) candidate languages that are considered. (Open the files in the directory testdata and do M-x guess-language-mark-lines to see for yourself.)

For each language, there is a default Ispell dictionary that guess-language-mode tries to use. However, for some languages there are several dictionaries available and guess-language can’t know which one you’d like to use. For example, there are several different dictionaries for German and for English. If the dictionary that guess-language-mode uses by default is not present, you will get an error message like the following:

Error in post-command-hook (flyspell-post-command-hook): (error "Undefined dictionary: en")

In this case, use the variable guess-language-langcodes to tell guess-language-mode which dictionary should be used instead. For example, use the following definition if you want to use British English and Swiss German:

(setq guess-language-langcodes
  '((en . ("en_GB" "English"))
    (de . ("de_CH" "German"))))

The key of each entry in this alist is an ISO 639-1 language code. The first element of the value is the name of the dictionary that should be used (i.e., what you would enter when setting the language via M-x ispell-change-dictionary). The second element is the name of the language setting that should be used with typo-mode (if present). If a language is not supported by typo-mode or if you are not using typo-mode, enter nil.

For a list of all dictionaries available for spell-checking, use the following:

(mapcar 'car ispell-dictionary-alist)

Languages that are currently supported by guess-language-mode:

While changing the spell-checker’s dictionary is the main purpose of guess-language, there are other things that a user might want to do when a new language is detected, for instance, a user might want to change the input method. Things like that can be easily achieved by adding custom functions to the hook guess-language-after-detection-functions. Functions on this hook take three arguments:

Template:

(defun my-custom-function (lang beginning end)
  (do-something))

(add-hook 'guess-language-after-detection-functions #'my-custom-function)

The code snipped below illustrates how guess-language can be configured to automatically change the voice used by the text-to-speech engine Festival (install festival.el for this to work):

(defun guess-language-switch-festival-function (lang beginning end)
  "Switch the voice used by festival.

LANG is the ISO 639-1 code of the language (as a
symbol).  BEGINNING and END are the endpoints of the region in
which LANG was detected but these are ignored."
  (when (and (featurep 'festival)
	     (festivalp))
    (pcase lang
      ('en (festival-voice-english-female))
      ('de (festival-voice-german-female)))))

(add-hook 'guess-language-after-detection-functions #'guess-language-switch-festival-function)

The pcase needs to be modified to use the voiced that are installed on your system. Refer to the documentation of Festival for details.