Aphasia refers to the loss of language and other neurological abilities primarily caused by damage to the Broca and Wernicke areas of the brain. You can read more detailed (and reliable) information about this dysfunction here.
The purpose of this article, however, is not to summarize what we know about aphasia but rather to discuss the idea of treating aphasia by having the patient learn a new, foreign language. I will post my thoughts on this approach and will shortly thereafter create a comment section where other, more qualified participants may tear it to shreds.
Now in studying neurology, computer science, and electronics, I have repeatedly come across incredible parallels between the fields, to the extent that I have come to understand humans to be robots. As a result of this model, I interpret aphasia to be akin to a randomly corrupted database. The program may query the database, and though most of the networking and data is still there, somewhere along the route the transmission is lost and this results in one of the most frustrating programming errors: a program that appears to be fully functional and complete but somehow doesn't accomplish the task at hand.
If you observe aphasiacs with mild to moderate severity of disturbance, they operate in much the same way as the aforementioned faulty database program. They often believe they can accomplish the task, but when they repeatedly make the queries, the routes are interrupted or the data is vacant when they arrive at the destination, and this manifests itself in the pausing and false starts characteristic of their speech.
What if, however, rather than trying to fix a frustratingly corrupted old system, the patients learned one that is entirely new? Would it be easier? Quicker? I would be interested to find out. I am prone to think it would for two biological reasons: