AI Zone Admin Forum Add your forum

NEWS: Chatbots.org survey on 3000 US and UK consumers shows it is time for chatbot integration in customer service!read more..

Thesis
 
 

We would like to develop an online stress management and analysis tool that woul provide level of stress with advice that would help a user that is undergoing stress. Our group is still conversing for the proper algorithm for this project as we would like it to be as credible as possible. Any suggestions? We would like to hear from you as soon as possible. thanks.

 

 
  [ # 1 ]

It will be an online stress management and analysis chatterbot. (forgot to include this)

 

 
  [ # 2 ]

there are many research projects going on in the health area. We’ll list a few project in our coming up synonym pages, on virtual persons and embodied agents, to be published next week. Keep an eye on this thread, because the terms mentioned will be linked automatically as soon as we publish this synonym pages.

 

 
  [ # 3 ]

I suspect you could use various current “sentiment analysis” tools to gauge stress..  I strongly urge trying to develop voice-in capability along the lines of MIT WAMI http://wami.csail.mit.edu ..  Further, you should design it from the bottom-up to be deployable to mobile devices..

- Marcus Endicott
  http://twitter.com/mendicott

 

 
  [ # 4 ]

okay, thanks! I guess we’ll just look up on that. We’re looking forward that this site would give us big help while our thesis is on the process.

 

 
  [ # 5 ]
Marcus Endicott - Jan 5, 2011:

I suspect you could use various current “sentiment analysis” tools to gauge stress..  I strongly urge trying to develop voice-in capability along the lines of MIT WAMI http://wami.csail.mit.edu ..  Further, you should design it from the bottom-up to be deployable to mobile devices..

- Marcus Endicott
  http://twitter.com/mendicott


>_< We’ve thought on that also but for now, we’re just focusing on developing a chat bot and develop it via only online. For now, all we need is the right algorithm or maybe algorithms. Like, for example, our chat bot would only accept American English and that is one of our problems as How can the chat bot know if the user entered American English…
maybe after this, we will look on developing that.. but anyway, thank you for your advice.

 

 
  [ # 6 ]

By algorithm do you mean:

What software should we use to create a chatbot?

Or

What rules should we use to detect stress in an American English response?

 

 
  [ # 7 ]
Merlin - Jan 8, 2011:

By algorithm do you mean:

What software should we use to create a chatbot?

Or

What rules should we use to detect stress in an American English response?

The rules should we use to detect stress in an American English response. That is for now one of our problems.

Hey, Maybe the software you are saying can help us to create a chat bot, hmm. well, is there already a software? I’m sorry, coz I am not aware.

 

 
  [ # 8 ]
Jjjy Yjjj - Jan 9, 2011:

The rules should we use to detect stress in an American English response. That is for now one of our problems.

I think you need to look at psychology journals rather than chatbot tech for answers to that question. Are there studies about phrases English speakers often use when stressed? Types of answers stressed people normally give to certain questions? Or how about how fast they respond to a question unrelated to their stress? Certain aspects of conversation that they focus on when stressed?

I don’t know what a good metric would be, but I’m sure there are a plethora of studies that a lit review will turn up. Your university should have access to relevant journals.

Jjjy Yjjj - Jan 9, 2011:

Hey, Maybe the software you are saying can help us to create a chat bot, hmm. well, is there already a software? I’m sorry, coz I am not aware.

Once you know what type of responses you are looking for, you can build an Eliza-style bot that will ask appropriate questions tailored towards testing for those responses. I think an AIML-based bot will serve your purpose nicely, but it depends on what you decide to test for.

 

 
  [ # 9 ]

For basic bot building you can use:

http://pandorabots.com/botmaster/en/home
http://www.personalityforge.com/index.php

I agree with CR that psychology journals may be the right place to look.

You might be interested in the ANEW (Affective Norms for English Words.) studies. These researched how emotion ties to some words. Maybe you could do the same thing to identify “stress” inducing words.

http://dionysus.psych.wisc.edu/methods/Stim/ANEW/ANEW.pdf

 

 
  [ # 10 ]
C R Hunt - Jan 9, 2011:
Jjjy Yjjj - Jan 9, 2011:

The rules should we use to detect stress in an American English response. That is for now one of our problems.

I think you need to look at psychology journals rather than chatbot tech for answers to that question. Are there studies about phrases English speakers often use when stressed? Types of answers stressed people normally give to certain questions? Or how about how fast they respond to a question unrelated to their stress? Certain aspects of conversation that they focus on when stressed?

I don’t know what a good metric would be, but I’m sure there are a plethora of studies that a lit review will turn up. Your university should have access to relevant journals.

Jjjy Yjjj - Jan 9, 2011:

Hey, Maybe the software you are saying can help us to create a chat bot, hmm. well, is there already a software? I’m sorry, coz I am not aware.

Once you know what type of responses you are looking for, you can build an Eliza-style bot that will ask appropriate questions tailored towards testing for those responses. I think an AIML-based bot will serve your purpose nicely, but it depends on what you decide to test for.

I think I should rephrase that question. What we need is how can the bot detect if the input of the user is American English.. not detect stress in an American English response. I’m sorry about that.. Thank you for your responses, we’ll keep this post updated because our thesis is still on the process of research.

 

 
  [ # 11 ]
Merlin - Jan 9, 2011:

For basic bot building you can use:

http://pandorabots.com/botmaster/en/home
http://www.personalityforge.com/index.php

I agree with CR that psychology journals may be the right place to look.

You might be interested in the ANEW (Affective Norms for English Words.) studies. These researched how emotion ties to some words. Maybe you could do the same thing to identify “stress” inducing words.

http://dionysus.psych.wisc.edu/methods/Stim/ANEW/ANEW.pdf

Thanks! smile We’ll look up on that

 

 
  [ # 12 ]
Jjjy Yjjj - Jan 9, 2011:

I think I should rephrase that question. What we need is how can the bot detect if the input of the user is American English.. not detect stress in an American English response. I’m sorry about that.. Thank you for your responses, we’ll keep this post updated because our thesis is still on the process of research.

Well, if you simply search for specific strings, you will turn up matches if the reply is indeed American English. If what you’re interested in is a parser or dictionary, then the tools you need are dependent on what programming language you want to use.

I’m perpetually (it seems) plugging NLTK for python. I guess I will in this case too. smile The Natural Language Toolkit (NLTK) includes parsers, stemmers, and a dictionary of English words. WordNet is the comprehensive dictioanry/thesaurus that NLTK uses. You can find WordNet implementations in many programming languages. Just google around.

 

 
  [ # 13 ]
C R Hunt - Jan 9, 2011:
Jjjy Yjjj - Jan 9, 2011:

I think I should rephrase that question. What we need is how can the bot detect if the input of the user is American English.. not detect stress in an American English response. I’m sorry about that.. Thank you for your responses, we’ll keep this post updated because our thesis is still on the process of research.

Well, if you simply search for specific strings, you will turn up matches if the reply is indeed American English. If what you’re interested in is a parser or dictionary, then the tools you need are dependent on what programming language you want to use.

I’m perpetually (it seems) plugging NLTK for python. I guess I will in this case too. smile The Natural Language Toolkit (NLTK) includes parsers, stemmers, and a dictionary of English words. WordNet is the comprehensive dictioanry/thesaurus that NLTK uses. You can find WordNet implementations in many programming languages. Just google around.

We’ve thought of string search algorithm. What can you say about that?

 

 
  [ # 14 ]
Jjjy Yjjj - Jan 9, 2011:

We’ve thought of string search algorithm. What can you say about that?

Do you know how to program? Any programming language has regular expression functions that will search text for specific strings or patterns. You can simply have the bot give a response that you specify whenever it successfully matches an input string to a specific reg ex pattern.

If you don’t know how to program, AIML is a markup language (like html is) that is easier to work with than most programming languages. (Regular expressions in general can take time to master.) Pandorabots lets users write AIML that the bot uses to form answers. So if you write AIML code that searches for a specific phrase, the bot will respond with whatever you tell it to if that specific phrase is found in the user’s input.

For more on AIML, check out: http://www.alicebot.org/aiml.html

Even easier to use is MyBot: http://www.a-i.com/ MyBot works by the user directly interacting with the bot. (There are two types of bots, MyBot and iHal, available on the website. I’m not really sure what the difference is between them, but I’m sure you can look into it and figure it out.)

The way you interact with MyBot, as I understand it, is the following: The user enters in text. When the bot responds with an answer the user doesn’t like, the user responds with “wrong” followed by whatever the bot should have said. This can take longer to train and reacts poorly with input that hasn’t been encountered before, but it doesn’t involve any coding. Transcripts of conversations are available to the bot owner, which is I suppose what you’ll have to use to do stress analysis.

 

 
  [ # 15 ]

The easiest way to detect a language would be through Google’s language detect API.

Sample:
http://www.google.com/uds/samples/language/detect.html

 

 1 2 > 
1 of 2
 
  login or register to react
‹‹ My Chatbot      Project River Updates ››