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..

need help with understanding how chat bots relate to information retrieval
 
 

Hello I just joined this site today, i am here to find out how chat bots relate to information retrieval and how information retrieval is used in chat bots.
This is the scenario that i have been given for a final year unit (IR&NLP;), in the nlp part of the coursework i researched chat bots and their use in both social as well as business life. Now i need to relate chat bots to IR and i have no idea on how to do it, i found a very vague pdf saying that chat bot IR is used in education but its not quite clear so i am lost at the moment.

Please help

 

 
  [ # 1 ]

Hi,

I’m currently using my bot (subbot.org) for simple information retrieval. If I’m studying physics, for example, I might enter an acceleration equation:

> an equation for acceleration is v_f = v_0 + a*t

> an acceleration equation is x = v_0*t + (1/2)(a * t^2)

Then later if I’m doing a problem, I might want to retrieve that information. But before I can ask it in a natural way I have to tell the bot:

> an equation for acceleration = an acceleration equation

> the acceleration equations = an equation for acceleration

> botname: if input =~ /^list (.*)/ then response = self.send(“logicagent: what is #{$1}?”); score += 2 end

Then I can ask the bot, later when I’m doing some problem:

> list the acceleration equations

And it will tell me:

the equations for acceleration is an acceleration equation, v_f = v_0 + a*t, an equation for acceleration, x = v_0*t + (1/2)(a * t^2)

Then later I might want to use a different syntax, so I have to do:

> the acceleration equations = the equations for acceleration

> list the equations for acceleration
the acceleration equations are the equations for acceleration, an acceleration equation, v_f = v_0 + a*t, an equation for acceleration, x = v_0*t + (1/2)(a * t^2)

—-

This is very basic IR, but allows me to add synonymous lookup expressions at runtime. I hope (soon) to automate some of the “A = B” expressions I’m currently entering explicitly, and to exclude these synonyms from the response listings. What I want to head towards is a system where I could tell the bot: “Learn about the equations of motion” and it would go out on its own time and search the web, reading wikipedia articles etc., then I could quiz it like: “I know that the final velocity is 0, the initial velocity is X m/s and the acceleration due to gravity is -9.8 m/s; how do I find distance?” and it would reply “First use v_f = v_0 + a*t to find the time then substitute that for t in x = v_0*t + (1/2)(a * t^2)”. And when it can do that it is a smaller jump to simply giving it the word problem and having it figure out how to do it itself smile

 

 
  [ # 2 ]

Hi Hetal! Welcome to the forums!

Information retrieval: it’s a little bit on your definition. In my understanding, chatbots are nowadays VERY often used (check the directory page), for accessing information about companies, its products and employees. Professional chatbot developers tend to refer to this kind of information retrieval as ‘knowledge management’.

Developers add examples on our directory pages themselves and sometimes they tick the box ‘knowledgement’. Therefore, you might want to checkout these examples:

http://www.chatbots.org/features/knowledge_management/

Hope this helps (just a little bit).

 

 
  [ # 3 ]

Thank you for the quick answers, really appreciate it. Your comments have been slightly helpful in giving me a better understanding.

@Robert, could you tell me which chat bot you are using from your website, i went on there but there were quite a few links and i didn’t know exactly which one you were referring to.

@Erwin, thank you for the welcome i have followed the link you provided and it had been very helpful in letting me see how various businesses would use chat bots for IR (knowledge management)
Would you say that the same pattern is followed in the social side of it as well

thanks

 

 
  [ # 4 ]
Hetal Patel - Nov 8, 2010:

Would you say that the same pattern is followed in the social side of it as well

not sure what you mean :-s. Can you explain your questions once more?

 

 
  [ # 5 ]
Hetal Patel - Nov 8, 2010:

@Robert, could you tell me which chat bot you are using from your website, i went on there but there were quite a few links and i didn’t know exactly which one you were referring to.

The agent I used in the example dialogs was the logicagent.

The full bot is controlbot.zip, which contains 24 agents.

 

 
  [ # 6 ]
Erwin Van Lun - Nov 8, 2010:
Hetal Patel - Nov 8, 2010:

Would you say that the same pattern is followed in the social side of it as well

not sure what you mean :-s. Can you explain your questions once more?

I’m sorry for the really late reply to this, in my eyes you can use chatbots in a social environment like an elderly carehome or a childrens nursery where chatbot induced learning can occur.
In the elderly carehome it can be used for the plain purpose of human iteraction with no other objectives. The chatbot will have a set of likes and dislikes that a person has and will continuously talk along those lines, as far as i can see there is nothing stopping a voip system being attached to chatbots which can make the whole experience of interacting with a chatbot reach new levels.

In the nursery the chatbot can be used to iterate basic language learning i.e. abc’s or numbers maybe even spellings etc

I hope that explains my question

 

 
  [ # 7 ]

Thanks Hetal! Much clearer in this way!

Yes, I definitely agree with you. That’s actually the reason why I started this whole website. I really really really believe in future applications in all parts of society. I wrote on the future of brand agents in 2005:

http://www.erwinvanlun.com/ww/trends/brand_dialogue/

and the progress in NLP is very slow.

But examples like this:
http://www.erwinvanlun.com/ww/full/robot_doll_lookalike/
http://www.erwinvanlun.com/ww/full/demented_people_play_robot_paro/

makes very clear that elderly like to have something to ‘communicate’ with.

The next step will be that toys like this respond to more queues from the environment, and in evolution, they’ll be more intelligent. But it will take a long time before they really start to teach us languages. However, eventually they’ll be smarter than us and speak all 5000-10000 languages in the world.

Erwin
p.s. don.t forget to add a photo to your bio!

 

 
  [ # 8 ]

Thanks for the information on the social aspects of chabots, I had a couple of questions to do with the robot paro.

Firstly - would you say that the reactions that it is given are hard set or would they be changeable?

Secondly i found this other link

http://www.erwinvanlun.com/ww/full/robot_dog_voice_commands/

It was thanks to the link you gave on the robot doll lookalike that i found this. Would you say that the chatbot shown in the video is something that could be used to teach? I mean i know that it responds to voice commands therefore allowing the children to see when and where to use the words whoever are there any other further applications that could be made.

 

 
  [ # 9 ]

They are all hard set. Reason is probably that children are not willing to train their artificial dog commands. Just like a voice system in your car, you need to pronounce very carefully.

However, I certainly expect in a few years from now we have toys that record what they hear, upload it to a central database and the collection of toys spread around, are learning from each other. That will be really be powerfull.

And that’s a little bit what’s happening with ‘ordinary’ chatbots on the internet nowadays.

 

 
  [ # 10 ]

Dear Hetal Patel.

Here is what I consider (IR) Information retrieval for beginners:

[ol]
[li]keyword matches: I AM YOUR FRIEND.[/li]
[li]wildcard matches:  I AM YOUR *[/li]
[li]invert words (pronouns, verbs, etc.):  YOU ARE MY FRIEND.[/li]
[li]default responses (no match): PLEASE GO ON. [/li]
[/ol]

There are common advance techniques,
such as recursion.  I may pop in later for
further discussion.

 

 
  [ # 11 ]

Hetal Patel asked, “Firstly - would you say that the reactions that it is given are hard set or would they be changeable?

Strictly, for the sake of conversation, I would say machine reactions may start as a hard set and then become changeable as human intervention improves the hard set by studying the conversation logs.  Advanced machines may be self-changeable, to further limit human intervention.  This is my opinion, and I am perfectly open to the views of others which may differ, who, for example, may consider a random generator to concat substrings changeable.

I am trying to make my opinion easy to understand, so that it may appeal to beginners.  The rewarding part of the elegant beginner stage is that implementing human simulation, even during this initial stage, presents a challenge.

 

 
  [ # 12 ]
8PLA • NET - Nov 19, 2010:

Hetal Patel asked, “Firstly - would you say that the reactions that it is given are hard set or would they be changeable?

Strictly, for the sake of conversation, I would say machine reactions may start as a hard set and then become changeable as human intervention improves the hard set by studying the conversation logs.  Advanced machines may be self-changeable, to further limit human intervention.  This is my opinion, and I am perfectly open to the views of others which may differ, who, for example, may consider a random generator to concat substrings changeable.

I am trying to make my opinion easy to understand, so that it may appeal to beginners.  The rewarding part of the elegant beginner stage is that implementing human simulation, even during this initial stage, presents a challenge.

Thank you for your reply on the query, I don’t know if you had looked at Erwin’s post however what you have written is fairly similar to his.
However i do appreciate your input and would like it if you gave your contributions to posts that i will be posting here in the near future

 

 
  login or register to react