Screen-Reading and Speech Recognition Programs

Screen-reading programs produce audio versions of electronic text and speech recognition programs change what is spoken into an electronic document. Blind and visually impaired foreign language learners can use these types of software programs to access and produce many critical need languages documents. It just requires the right preparation and combination of software programs or equipment.

Scanning Print to Produce Electronic Text

If one has acquired an electronic text version of the foreign language material from a teacher or publisher, then using screen-reading software can help to access an audio version of the text. However, if one does not have an electronic copy of the print textbook or handout, then scan the physical pages with Optical Character Recognition (OCR) software programs, such as ABBYY FineReader or Nuance OmniPage.

Before scanning, turn on the language options of the OCR program. Nuance OmniPage supports 119 languages, but only Russian and Turkish of the critical need languages. ABBYY FineReader 9.0 has capabilities of recognizing letters, script and characters in 184 languages, including the critical need languages, and can have dictionary capabilities in 38 languages. It automatically detects the language of a document, and allows users to select any combination of languages to recognize multilingual documents.

Some students use products such as Nuance’s Open Book to scan and read materials from class on their own. The two scanning OCR software programs mentioned above are used with Open Book, and utilize the RealSpeak Solo voice mentioned below – all of which provide recognition of various critical need languages if the student selects the right combination.

A Sampling of Common Screen-Reading Programs’ Foreign Language Options

Kurzweil 1000/3000 has been used for blind students in Dutch, French, Spanish, German, Italian, Portuguese, and Swedish, but not critical need languages.

Window-Eyes screen reading program offers five common foreign languages, but has a list online of distributors in other countries that have adapted it for their local languages. While none of the critical need languages are currently included in this list, this may expand in the future. 

JAWS has been translated into 17 languages and used in 50 countries. If a student has Windows NT 4.0 or later, then the unicode version of JAWS should be installed when upgrading to 6.10. Unicode allows JAWS to provide seamless support for languages with non-Latin alphabets, complex scripts, or special reading orders. With the purchase of an appropriate SAPI-compatible speech synthesizer, JAWS users can now access information in a number of other languages, including Arabic, Japanese, Russian, and more. Eloquence is the default synthesizer installed with JAWS and only covers the most common foreign languages. An alternative plug-in with JAWS is RealSpeak Solo, which has more options for critical need languages.

Since many beginning level foreign language textbooks have both English and the foreign language on the same page, JAWS can handle this according the JAWS 6.0 help file:

“As you navigate or read text in a document containing multiple languages, and if you are using Eloquence, JAWS automatically switches languages appropriately. If you are not using Eloquence, the synthesizer in use still alerts you to the language change by speaking the language name in the Message Voice before the text in that language. Say Line, Say Sentence and Say Paragraph all honor language detection and switching. We recommend you have all Word language modules that you intend to use installed before using this feature.”

As with all screen-readers, JAWS does not translate foreign languages into English, it will only announce the text with the correct phonemes for the language it is written in. A selection of high quality, SAPI 5, voice samples are maintained by several developers.

For a student using RealSpeak Solo, it supports 30 languages and dialects including Chinese, Hindi, Japanese, Korean, Russian and Turkish. Although it doesn’t have Arabic or Persian/Farsi, their website explains Nuance is able to develop custom languages and voices on demand. Text Aloud offers many languages including Arabic and Turkish. NeoSpeech offers Chinese, Japanese and Korean voices.

If a student needs a good electronic dictionary, and uses JAWS, Jerry Neufeld has a scripted version 4.3.4 of Ultralingua, a specialized foreign language dictionary and grammar program for which multiple language modules are available (though mainly in Romance languages). The scripts only work with this version of the program, and if the student wants the JAWS scripts that make it usable, write to Jerry at: jerry.neufeld@rogers.com.

Many critical language speakers overseas use JAWS. For example, people in Hong Kong can get a JAWS Chinese version from this website (under Product & Services - Accessibility). Sometimes for cost-effectiveness or performance, they will also use screen reading programs developed in their own countries.

Screen-Reading Programs Developed Overseas

In Egypt, Sakhr and Harf software products have been developed, and distributed in North America, including the Ibsar Reading Machine which gives Arabic- and English-using blind and low vision users customized voice command of their computers, audio of their keystroke, use of the Internet, and access to scanned text. This company also has an OCR program that handles the Arabic written language.

The Ibsar Reading Machine removes the step of needing to add in the diacritical (short vowel) marks that are not typically represented in written Arabic. It automates how to resolve the ambiguity of the words and selects the proper pronunciation of the input undiacritized text. Based on Sakhr Text-To-Speech engine, it involves a three-step process to achieve this: the Linguistic Module that converts the input text into a phonetic transcription, the Phonetic Module that calculates speech parameters, and the Acoustic Module that uses these parameters to generate synthetic speech signals. Once complete, it is fully compatible with Microsoft Speech APIs version SAPI 5.0 to create a natural sounding speech output.

In India, the National Association for the Blind (NAB), Delhi, and a Lucknow-based software development company built a Hindi screen reader that can handle electronic text in Hindi at a fraction of the cost of JAWS.

The Japanese BRLTTY screen reader added new functions such as Japanese voice synthesis, using the software voice synthesis library developed by Create System Development Co., Ltd. It also supplies detailed read functions using voice during Japanese Kanji composition input method, enabling Kanji homonyms to be inputted. It is currently being developed to handle both English and Japanese. The beta-version screen reader and its source code are openly available online.

There are several Chinese screen readers used by native Chinese blind people, but the student may need to be more advanced in understanding or instructing it. If a textbook for a Chinese course includes information also in English, then the Chinese Text Reader may work.

Speech Recognition Programs for Foreign Languages 

If a student uses the keyboard to type in the foreign language, go to the discussion under “Converting Foreign Language Braille into Print” that follows for how to install foreign language fonts and keyboards. If not, then speech recognition programs can do the task.

For example, Via Voice automatic speech recognition and text-to-speech functions can function in Korean, Chinese and Japanese.

J-Say Pro is an interface product by Next Generation Technologies that enables using Dragon NaturallySpeaking with JAWS. Dragon NaturallySpeaking comes in a few common foreign language versions, with Japanese being the only critical need language. These versions are available in foreign countries, primarily the European market. Nuance, makers of this software, have an international marketing department listed online.

The Japanese version includes acoustic models and vocabularies developed specifically for Japanese speakers. The vocabularies include regional words, such as location names and proper nouns. Although it automatically selects whether to use Hiragana, Katakana or Kanji, the user can purposely select from the three writing scripts by using specially designed voice commands.

Once the students train the speech recognition system with their own voice and pronunciation, it will accurately record it. Speech-to-text engines work in that they listen to individual phonemes instead of words. The program then seeks the highest probability of what words fit into a common phrase that sounds similar to what the student spoke. In a number of cases, a student could incorrectly pronounce a word and still have it come out correct. However, if this doesn’t work, the student must know enough of the language’s spelling to correct the program and train it.

If the program is purchased from a foreign country then the foreign language versions may be more accurate than the American English version, shortening the training curve significantly. The user will also get correct accent marks or punctuation, in their proper places without having to specifically name the grammatical mark to be inserted.