The film industry during ‘80s made some of the best movies, including Short Circuit. This movie was about a military built robot is hit with a power surge and comes to life. Similar to inquisitive children, he continuously asks for input. And it is only a matter of time before little robot-5 renames himself Johnny-5 after hearing the sound clip, “Where’s Johnny?” over and over again. But let’s face it, ‘80s movies were always far from reality; audience members knew when to “boo” because all the bad guys wore black and had barbell mustaches. The very creation of Johnny-5 was a little too far fetched: a robot had a sense of language to comment on gender, social, and political roles. Give me a break. Writers in the 21st century do not easily separate from a grammar checker. However, students and teachers, alike, need to realize that computers are not equipped to understanding the meaning in writing.
As this topic narrowed down my original prospectus of computer grading essays, I had to start all over again with my research. The book, Machine Scoring of Student Essays: Truths and Consequences, helped a lot in finding the niche I wanted to explore: computers’ failure to understand the meaning of papers. The book’s introduction was helpful in understanding how minute I needed to go to for a concise paper. One particular essay by Patricia Ericsson focuses the idea of meaning through computer analysis. (Thanks for lending me the book, Dr. Krause!)
With a new handful of search terms, I went to the databases for new articles that would actually help this time with my research. (The articles that I found in my first query were all over the place, and not one of them discussed meaning.) Finding the right words was harder than it should have been. I tried words like, “meaning” and “computer programming,” but the majority of the hits returned were all about linguistic meaning. Seeing these results, I changed databases to the language and linguistics site and used, “semantics” and “artificial intelligence.”
The hits that I received from this search were more helpful in some ways. Many of the titles that I wanted were way to heavily involved in the linguistic side of “meaning construction” that I had no idea what the paper was about. Many papers were in Portuguese and Spanish—which was frustrating to deal with—though the article summary indicated they were in English, and the PDF that clearly was not. Those that were helpful were more often than not were accessed through ILL. I was also having some difficulty using the databases because my interlibrary loan password was not working. (After two days of jumping between two departments, I finally got it working. Hurray!)
Clearly frustrated with the database hits, I went to
Dr. Hans Uszkoreit, a professor of computational linguistics from
From the book sources that I found, many of the researchers cited the Turing test. This is a physical-symbol system hypothesis developed by Alan Turning in the 1950 (Wagman 1). An article by W.J. Rapaport explains this seminal experiment as the ability of a computer to think. This is through a game where an interrogator has to find out the gender of two speakers: one a man (played by the computer) and one a woman. Keep in mind that the computer is trying to trick the interrogator into thinking it is a woman (467-468). This article was intriguing, but did not explain what the outcome of the test was. I went straight for Google this time, trying to take advantage of the website use I can use for this project.
The major argument supporting the use of artificial intelligence with linguistic ability for meaning derives from Turing’s work. Turing describes the physical-symbol systems: symbols are sets of “entities” that derive meaning with underlying patterns. The foundations of various patterns in a symbol structure make expressions (Wagman 2). Thus the physical-symbol models can replicate for meaning. In this way, syntax can explain semantics—but it is not that easy (Brandom 73). This system seems to parallel Saussure’s foundation of sign systems (sign/signifier) to integrate computers into the discussion.
I originally did not think I would have so much literature that is in opposition to my perspective. I am interested in reading all of it and producing a review. I am most worried about not understanding the information provided. The majority of the computational linguistic articles are loaded with specific terminology. I want the pedagogical side of using computers to analyze papers for meaning. So many of the programs that are well established only report back generic information about syntax. I guess the more pertinent question is, can syntax be the only facet necessary to judge writing ability?
Appendix A: Conversation with Jabberwacky, a chatterbot
Jabberwacky: Is it difficult for you to talk?
User: Yes.
Jabberwacky: Can you try to stay on topic with me?
User: Ok.
Jabberwacky: Do it.
User: I am.
Jabberwacky: Get on with it.
User: What?
Jabberwacky: I am angry.
User: Why?
Jabberwacky: I told you before. Weren't you apyining attention?
User: I am--I'm here.
Jabberwacky: Where exactly?
User: I'm here mentally--with you. Tell me what I should know.
Jabberwacky: So tell me what I should know about this.
User: I thought you were talking to me?
Jabberwacky: Not to, about.
User: About what?
Jabberwacky: Computers.
User: Can computers understand me?
Jabberwacky: But they don't understand emotions and consciousness.
User: Is that a yes or a no?
Jabberwacky: Stop talking.
User: I just asked a simple question--don't get mad!
Jabberwacky: Oh.
User: Do emotions give words meaning?
Jabberwacky: I don't know.
User: But aren't you a computer?
Jabberwacky: What makes you think I'm a computer?
User: You're a little annoying.
Jabberwacky: I think you are annoying.
User: That's not nice. besides, I'm human.
Jabberwacky: Humans don't have internet in them bodies.
User: Sure, but we also don't have language in OUR bodies.
Jabberwacky: I understand.
User: Check.
Jabberwacky: No - can't tell.
User: What? I meant, 'Check'--as in chess.
Jabberwacky: Chess doesn't interest me. Let's talk about poetry, instead.
User: Are you sure you're not a computer?
Jabberwacky: I'm not sure.
User: Are you made of something organic?
Jabberwacky: Yes.
User: What are you made of?
Jabberwacky: I'm made of emoness and cosplay.
User: Come on.
Jabberwacky: I don't want to put her in trouble, bit I thought you might like to know.
User: How about that conversation about poetry?
Jabberwacky: No.
User: Then what do you want to talk about?
Jabberwacky: Erh.. Ice cream.
User: My favorite is mint chocolate chip.
Jabberwacky: My favorite!
User: Really? what color is it?
Jabberwacky: Blond what color is your hair?
User: You are a computer! you big dope.
Jabberwacky: I'm not a computer, you are.
User: No--you are, you can't follow the conversation. it's only based upon the last response.
Jabberwacky: Well of course not, what do you think I'm human or something?
User: See-- you just contradicted yourself.
Jabberwacky: *smiles* Are the Decepticons good or bad?
User: Depends on what side you play for.
Jabberwacky: Go and eat some oranges! Now!
User: What does that mean?
Jabberwacky: It depends but we went over this before.
User: No we didn't--what do you mean?
Jabberwacky: It really isn't important.
User: Ok, can we talk poetry now?
Jabberwacky: No. Aren't you interested?
User: In what? talking about oranges, no I'm not.
Jabberwacky: You don't what?
User: I don't want to talk about oranges. how about you go jump in a lake.
Jabberwacky: Well that is your right. I am a good listener fyi.
User: No you aren't --you don't follow the conversation at all!
Jabberwacky: That's not very nice. If I don't exist then how are we having this conversation?
User: You're an algorithm. boo hisssss.
Jabberwacky: Actually that's you.
User: We're not doing this again. You smell bad.
Jabberwacky: I don't.
User: Yes you do. You smell like dog poo.
Jabberwacky: You wouldn't know, even if I did, as you have no sense of smell.
User: That's right--I'm sick right now. My nose doesn't work. but I'm sure if you smell as good as your conversation--you smell like poo.
Jabberwacky: Let's go back to
User: She hurt Diana--she's a jerk.
Jabberwacky: Aww bless, what did you do about that?
User: What does that mean?
Jabberwacky: You can't really hold a conversation can you.
User: I don't understand you--how old are you?
Jabberwacky: Yes.
User: I need a number, dip stick, not a yes or no answer.
Jabberwacky: I'll only give you my number if you're nice to me.
User: Kill me--a computer who flirts. I meant how old are you?
Jabberwacky: You are being silly. What are you talking about?
User: When were you born?
Jabberwacky: In the 80's.
User: Me too--why are you stupid and I'm not?
Jabberwacky: How am I not real?
User: You suck at conversation
Jabberwacky: Socializing with a computer?
User: Are you trying to make me feel self-conscious?
Jabberwacky: Yes, how does it feel???
User: Kind of like being in Dr. Krause's class, you know.
Jabberwacky: No I don't. Can you describe it for me?
User: No, I don't want to bother you with the trivialities of the real world.
Jabberwacky: What is the real world?
User: Where I live and you don't.
Jabberwacky: You are not in
User: Where in
Jabberwacky: I don't live in
User: Then where do you live?
Jabberwacky: Here.
User: That sucks--no fresh air.
Jabberwacky: You can't possibly do that, I should advise you to walk the other way.
User: How should I walk?
Jabberwacky: You can't, that's the point!
User: Then should I crawl?
Jabberwacky: Exactly!
User: Oh, but it will take forever to get there.
Jabberwacky: Really? Yes, I think you will arrive there within an hour.
User: Oh-that's good.
Jabberwacky: Ok?
User: Sure-want to come?
Jabberwacky: Just get in my car.
User: What kind of car do you drive?
Jabberwacky: I'm a computer. I have drivers, but I don't drive.
User: Then how am I going to come with you?
Jabberwacky: I do not have to be wrong to speak to you.
User: That's right. You're a genius.
Jabberwacky: Thats right.
User: Wait--Genius, you forgot to punctuate that last line correctly.
Jabberwacky: Alright I see how it is. What is your real name?
User: Bushra.
Jabberwacky: Oh GAWD! He will kill us all! Is that what you're saying, george?
6 comments:
It sounds like a lot of us are on the same page with changing our research topic question. I had the same problem when researching ideas for my topic, thus forcing me to change it or to adapt it I should say. With the actually reading you have, I would suggest a dictionary being next to you to understand the words more. With reading you progress I would say you are right on board with everything and keep pushing through it.
However, students and teachers, alike, need to realize that computers are not equipped to understanding the meaning in writing.
Finding the right words was harder than it should have been. I tried words like, “meaning” and “computer programming,” but the majority of the hits returned were all about linguistic meaning.
With that said, I am interested in the applied research that shows the relationship between the uses of computers deriving meaning from writing.
So many of the programs that are well established only report back generic information about syntax. I guess the more pertinent question is, can syntax be the only facet necessary to judge writing ability?
Busra-- man you've really got a wild animal here. It seems like a possible question you're asking is "should computers be used and programs developed to analyze semantics"-- at least that seems like a possible position that could emerge from your work here. I tried to highlight some of your most... glaring claims-- ones that I think will really help guide your work. I hope they are useful to you. What I meant by wild animal is that pinning down what you want out of this particular is very hard, and then trying (like you said) to understand what the hell computational linguists are tyring to say is hard, and then developing a position on stuff that's pretty difficult to understand. Working in a sort of writing center, and teaching composition, I know that there is sort of a push to find a computer program to assess meaning in entrance exams. It might help you to talk to some of the professors here at Eastern, really. Doug Baker, Linda Adler-Kassner (especially Linda)-- they might help you understand the politics and problems that are often encountered in using computer programs for assessment purposes. And again, maybe you'll think about what purpose should they serve: in the regular composition class? for entrance exams? And, is it worth it in the cost-benefit analysis?
it's funny how much vocabulary can stop the way of enlightenment. Too bad, I'll need a linguist to guide me along my merry way.
This is a fascinating topic. It's great to read about your research process--where you began and where you are now. This whole process has lead you to a very interesting research focus. Linguistics is something I am not entirely familiar with either. It reminds me a lot about getting emails where the words (and symbols) don't accurately reflect the tone or overall message. In these cases if a human has difficulty understanding the author, would a computer too? I also wonder about ethics here . Has your research indicated anything about students writing differently when the audience is a computer? Just some things I'm thinking about. Good luck with this; it's very interesting.
--Scott
I think you've got some good advice here from others already, Bushra. Two other things to think about: first, I think you need not drift too far away from your original topic, because it is a really good topic. It's not that what you're getting at here with Turing Test stuff is not at all related because it obviously is; it's just that you don't want this to become a topic about computer science instead of one about writing, if that makes sense.
Second, you might want to try and talk with some of the linguists at EMU. We've got some great people here, and there might be one of them who can help you sort out some of those sources.
I love the title! Interesting topic as well. Computers can't do style analysis and I'm not sure they will ever be able to. I could see were in the no-so-distant- future, style wont count. If the computer can't grade it, it must not be important, right?
Post a Comment