It was a few months ago that I read about the inauguration of the first internet cafe for the blind in Islamabad. I had wanted to write about it earlier but somehow skipped my mind, until the recent discussion on Islamabad allowed me to jog my memory a bit. I do want to share the news, even if a bit belatedly, because it is so uplifting and empowering.
According to a report in The Daily Times, the country’s first-ever Internet café for the visually impaired was inaugurated with the help of generous funding from the World Bank and the Pakistan Foundation Fighting Blindness (PFFB).
It appears that the software technology that enables such a user experience for the blind was introduced to Pakistan by an ATP regular, Aqil Sajjad. The software used by the internet cafe is called JAWS, and was first introduced by Aqil to Pakistan in 1999. Aqil has since then moved to the US for graduate studies in Physics, but the software is now more widely available to the blind in Pakistan, and making a dramatic difference in their lives.
These software packages are quite expensive but are really remarkable in what they are able to achieve. With their assistance, blind are able to operate the computers without any sighted assistance. It requires no special hardware and can be installed on a standard Windows machine, essentially making Windows accessible to a blind person by providing speech output as a substitute for the information displayed on the monitor for sighted people.
It is awe-inspiring how a software solution has made it possible for the digital world to finally become accessible to the blind. For once, there is a chance that the technological gap between the blind and the sighted may close. I am told that it is not just the simple use of computers that has now become possible with these softwares, but with a few accessory technologies, many other important intellectual tasks have also become much easier for the blind.
For example, the blind were previously dependent on sighted people for access to reading material. However, thanks to the internet and the availability of such software, that is no longer the case. Now one can even find entire coursework from some of the major universities (such as MIT) online, including class lecture notes, homework assignments and examinations.
In addition to this, by using a simple scanner and OCR software, they can now even read books on their own. Scanning books does take a bit of extra time, and the recognition is not 100% perfect (there are some errors), but having this option for reading a book without any sighted assistance is nothing short of a blessing. It might be worthy to mention here the project launched by Google to place a significant number of library books online.
While JAWS and Windows Eyes are good for general use, such as operating Windows, word processing, e-mail, internet surfing, instant messaging etc, doing Math (and in Aqil’s case, also Physics) with them is tricky. For that purpose, special software is now available and carries high recommendations from users.
One such software for Mathematics use is called WinTriangle. According to a Harvard paper from a set of conference proceedings on this topic, a new conversion tool, LaTeX2Tri, makes LaTeX files of physics and mathematics accessible to blind or visually impaired users of WinTriangle. Through a variety of pathways, common file types, such as TeX, Word, and PDF, may be converted to Triangle, the working language of many blind or visually impaired students and researchers. Textbooks, arXiv preprints, class notes, and problem sets are now readily accessible to WinTriangle users, completing the loop of mathematical communication between the blind and sighted communities.
I am inspired, not just by the progress that technology has made to make resources available to more people than before, but also by the committment and personal motivation showed by the blind people in Pakistan. They managed to reach out and find a solution for their needs and are now making it more widely available. However, a majority of our country does not speak, read or write in English, and I am still unaware of such software being available in Urdu. Unless it already exists out there, this presents a great opportunity for Pakistani software industry to contribute to the society.
For those interested, JAWS is published by Freedom Scientific and is available online for download. The price of Jaws is about $1,095 for the professional version and $895 for the standard version. Unregistered versions require a reboot of the computer every 40 minutes. However, a demo version of the program is available free of cost. A similar program called Windows Eyes can be downloaded here. The Mathematics software WItriangle is available here.
keep up the good work.
Aqil,
I did a little background research and here is what I came up with.
A Text-to-Speech Synthesizer can be broken down into two broad categories i.e. Natural Language Processing (NLP) and Speech Synthesis (SS). NLP deals with parsing the text and building up a ‘language’ database. The models used here are the same ones used in an OCR and grammar correction software i.e. hidden markov models. The ‘phonetic’ layer of the language is then overlayed onto this ‘language’ representation. This part faces the problem that you mentioned regarding urdu not having any vowels.
The search also revealed the research group making headway in this area. Its called the Center for Research in Urdu Language Processing (CRULP) and is at the National University of Computer and Emerging Sciences (NUCES) in Lahore.
It looks like they had made some real progress in the area. I’m pasting the links below. These guys definitely deserve recognition.
Homepage:
http://www.crulp.org/
Urdu Localization Project: http://www.crulp.org/ulp/
Relevant research paper:
http://www.crulp.org/ulp/download/LTS%20for%20Urdu %20TTS.PDF
Naveed:
I am not a programmer, so may even be wrong about this, but I think the issues in generating speech are different from those in OCR.
For speech, we use rules for pronounciation, whereas for scanning, we need rules for recognizing the characters without worrying about how they are pronounced.
If I am not mistaken, the major problem in urdu text to speech is due to the fact that it does not have any vowells. For example, consider the Urdu word ‘kub.’ It’s written as kaaf bay. Now, without zer, zabar or pesh, you can not know by default how to pronounce it. So you not only have to create some rules for the alphabets, but also need a large dictionary.
Interestingly enough, if I write ‘kub’ in roman urdu, jaws is able to pronounce it properly. That is because of the vowell u in the middle.
I read about a Pakistani research team that had considerable success with an OCR for documents in urdu. They showcased their product in an IT fair (it was either PASHA or KRL I forget). I would assume that the underlying heuristics or models for picking out OCR data from documents in urdu would be quite similar to those needed to generate speech, since you would need to build some sort of language semantics and grammar to recognize the words in the first place. Obviously generating speech and tone would impose other challenges but somebody please correct me if I am wrong regarding the above assumption.
If that is the case then I’m sure theres a case for writing up a project grant to fund such a project. I have no experience in the field of speech generation but I do know of one faculty member in LUMS, Lahore working in vaguely the same area. If we can manage to come up with a working proposal then we can start it up as a student project.
Any takers?
Thanks for this post Bilal.
However, I must clarify that my mention as the person who introduced the software to Pakistan is not accurate. I was just one of the first people to use it in Islamabad (simply because I was fortunate enough to have the means to buy it), and naturally, people needing info about screen reading software were often directed to me for their queries. So my name just got mentioned rather undeservingly in that daily times report.
Anyway, the latest on the internet cafe is that it’s up and running. They are developing some training manuals for Jaws with instructions in Urdu, which is really helpful for those who may not be too comfortable with English. The response has been encouraging, and they are even considering the possibility of holding some training workshops in other cities once in a while.
Urdu software: yes, this is indeed a huge problem; what I understand with my discussion with some cs people is that it’s going to take a while before we have an urdu speech program. The problem has to do with the way Urdu is written, especially the fact that it does not have vowels. But yes, I do hope that enough resources and effort is put into it.
Eidee man, the open source community has not had much input in jaws and windows eyes. These are purely commercial programs.
WinTriangle has been made an open source program, but not much has come of it. So it remains the rather crude program that it has been for the last 3-4 years or so, with a lot of room for improvement.