GroshLink

Faith, Learning and Practice in Natural Language Processing (Peter Venable 10/01) 

Note: written for the Graduate Christian Fellowship's (GCF) vocation project (VP).  Now that the GCF site has been taken down, over time I'll post the pieces for your blessing.  Thank-you to all the alum who participated in this work.  Alumni friends, I would love to have to hear your current thoughts on your vocation.  Maybe we can have that conversation at the alumni reunion on April 21.  We'll have a 9am breakfast at the Seigfrieds, a 1pm reception most probably at the Adamson Wing, and an evening opportunity yet to be confirmed.  Email me for more details.  Current GCF activities can be found at http://www.u-connectpgh.org  

I work in the subfield of Computer Science called Natural Language Processing, which deals with human languages such as English, French, or Mandarin, whether written or spoken. While I could argue that my work is science, I really see it as more technological: pushing the limits of what computer technology can accomplish. Can we get a computer to understand English? How about translating a document, or a conversation, between English and Chinese? In this paper I'll examine how my faith in the truine God motivates and informs my work.

It seems to me that my faith comes in more at the level of motivation that of content. I'm interested in building technologies that enable people of different languages to communicate because I believe God values every nation, tribe, and tongue, and so should we, seeking to understand them, communicate with them, sharing God's love and truth. I'm interested in building technologies that enable computers to respond to human language for everyday tasks, because I want to make computers easier to use and less obtrusive. Seeking to adapt the computer to the way we prefer to communicate, the speech God gave us, rather than having us adapt to the computer, shows that we value people and appreciate the way God made us.

On the other hand, I have some doubts about using computers to aid human communication. If we design a computer that can translate between Chinese and English, we can travel to China and not learn any Chinese. Even if the translator is pretty good, much better than any currently available, it can never be as good as a living person who understands both languages. Even given a perfect automatic translator, a traveler who doesn't learn any of the local language will probably not understand the culture either, since language and culture are so closely related. Though the computer might allow superficial communication in many lands where there was none before, it will never be good enough to remove the barriers between different languages and cultures. These must instead be crossed, and that crossing takes humility and hard work.

These doubts come in response to an unspoken assumption common in technological research: that a technical solution is the right kind of solution. This is related to another assumption which may not be compatible with Christian faith: that human language, like so many other problems, can be solved, because the human brain, after all, is just a very complex computer made of neurons. If you hold the reductionistic faith that the scientifically observable world is all that matters, it seems almost inevitable that computers will one day surpass human intelligence in every way. While everyday work has a very practical short- to medium-term outlook, and such philosophical isses are rarely discussed, I sometimes struggle with this question. While I have faith that God created the universe, including humans, I don't know how much of our ability to think, feel, create, love, and so on is physical enough that it could be understood by a brain scientist. Though I don't think it's that likely, perhaps one day computers will be better than humans at translating foreign languages. If that ever happens, we'll have to consider whether they are our equals or just machines, or somewhere in between.

In practice, machine translation methods today simplify language immensely. You could say they get 80% of the accuracy (in results) of a human translator, with only 1% of the understanding. The computer can look up words and phrases in a bilingual dictionary, change the word order for the grammar of the target language, and make the result sound similar to other sentences in the target language so it sounds natural, but it never understands the meaning of the sentence in the way people do.

That this simplified approach works at all is only possible because of the logical structure of human language. A little study of how language works (linguistics) shows a fascinating, well-developed system, a marvel for which we can thank and praise the creator. In general, any scientific endeavor is based on the faith that the universe makes sense. That faith is based on faith in the creator, who is subtle but not malicious (to quote Einstein). For those who have no faith in God, I'm not sure where they get their faith in the comprehensibility of the universe, but I guess they just assume the world makes sense because it's impossible to function without that assumption.

My desire to please God is a reason (hopefully the primary reason) for working hard, treating colleagues well, and asking for God's blessing and guidance. I don't think these are much different whether you study Natural Language Processing, Chemistry, or Music.