Opencog noobie questions

[neurop...@gmail.com]

Hi, I am new to Opencog and hopefully building a chatbot with it soon, I was going to use OpenNARS but after some reading I am convinced that PLN > NARS

So here are my questions:

1 I am thinking of writing rules that convert input text to a understandable format for Opencog, I am aware that this has already been done with the NLP pipeline

However from what I've been told rules that for example convert "that" into what is meant with "that" aren't implemented yet.

Example: Tim: My computer keeps crashing 

Tim: I hate that 

This rule would turn this into : Tim said that Tim hates that his computer keeps crashing (In atomese)

This is just one example of such a rule.

2 I should've asked this before but are such rules even helpful? And is atomese even the language in which such things would be inputted to the system? 

I am extrapolating from OpenNARS in which input like this is done in Narsese but I am not sure how that works in Opencog yet...

3 What does MOSES do exactly? And is it needed for a chabot? (I know what MOSES does with the evolution of mini programs, I am just not quite sure on what those mini programs do..)

That's all of the questions I could think of right now, not sure if I missed one (hopefully not)

The chatbot I am building will use an up board which will run Opencog, that should be okay right?

http://up-shop.org/up-boards/62-up-board-4b-32-gb-emmc-memory.html

Thanks!

[neurop...@gmail.com]

Removed last post forgot a few things,

In Opencog are events and inference conclusions in the atomspace timestamped? (Have the time and date attached on which it happened)

Diagram, second one I made hopefully better than the OpenNARS one (Still improving my diagram skills)

[Linas Vepstas]

Three short comments:

1) PLN is a set of rules that run on top of the rule engine. Those rules, when triggered, alter the "truth values" on atoms in certain ways. You (or someone) should be able to create a different set of formulas for truth values, that emulate the NARS formulas. You could then run PLN deductions side by side with NARS deductions, and see which you like better (or even create yet other kinds of formulas)

2) we have "relex2logic" rules, which convert natural language to logic expressions. These are ... time-consuming to write, and by nature are incomplete and buggy -- because natural language is a fluid, changing thing. Hand-writing rules is a good exercise, but madness lies in that direction.

3) the current diagram of how things connect is kind-of-ish reviewed here: https://github.com/opencog/opencog/blob/master/opencog/eva/architecture/embodiment.pdf

--linas

--
You received this message because you are subscribed to the Google Groups "opencog" group.
To unsubscribe from this group and stop receiving emails from it, send an email to opencog+unsubscribe@googlegroups.com.
To post to this group, send email to ope...@googlegroups.com.
Visit this group at https://groups.google.com/group/opencog.
To view this discussion on the web visit https://groups.google.com/d/msgid/opencog/d4fa76b4-80d2-4ea6-96af-e072237b03fd%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

[neurop...@gmail.com]

Ah thanks for letting me know about those rules, one more thing about this tho. Is there already a rule which let's opencog(chatbot) know that any statement made by a person or other information source it is talking to is what they think instead of basically storing it in the atomspace without the source (You hopefully get what I mean) Sorry if I am bad at describing what I mean

Yeah, I've seen that diagram :) I posted one about how I am planning to implement this chatbot however it didn't post correctly, here it is (Hopefully it works this time)

Thanks

Three short comments:

1) PLN is a set of rules that run on top of the rule engine. Those rules, when triggered, alter the "truth values" on atoms in certain ways. You (or someone) should be able to create a different set of formulas for truth values, that emulate the NARS formulas. You could then run PLN deductions side by side with NARS deductions, and see which you like better (or even create yet other kinds of formulas)

2) we have "relex2logic" rules, which convert natural language to logic expressions. These are ... time-consuming to write, and by nature are incomplete and buggy -- because natural language is a fluid, changing thing. Hand-writing rules is a good exercise, but madness lies in that direction.

3) the current diagram of how things connect is kind-of-ish reviewed here: https://github.com/opencog/opencog/blob/master/opencog/eva/architecture/embodiment.pdf

--linas

On Thu, Jul 20, 2017 at 6:14 AM, <neurop...@gmail.com> wrote:

Hi, I am new to Opencog and hopefully building a chatbot with it soon, I was going to use OpenNARS but after some reading I am convinced that PLN > NARS

So here are my questions:

1 I am thinking of writing rules that convert input text to a understandable format for Opencog, I am aware that this has already been done with the NLP pipeline

However from what I've been told rules that for example convert "that" into what is meant with "that" aren't implemented yet.

Example: Tim: My computer keeps crashing 

Tim: I hate that 

This rule would turn this into : Tim said that Tim hates that his computer keeps crashing (In atomese)

This is just one example of such a rule.

2 I should've asked this before but are such rules even helpful? And is atomese even the language in which such things would be inputted to the system? 

I am extrapolating from OpenNARS in which input like this is done in Narsese but I am not sure how that works in Opencog yet...

3 What does MOSES do exactly? And is it needed for a chabot? (I know what MOSES does with the evolution of mini programs, I am just not quite sure on what those mini programs do..)

That's all of the questions I could think of right now, not sure if I missed one (hopefully not)

The chatbot I am building will use an up board which will run Opencog, that should be okay right?

http://up-shop.org/up-boards/62-up-board-4b-32-gb-emmc-memory.html

Thanks!

--
You received this message because you are subscribed to the Google Groups "opencog" group.

To unsubscribe from this group and stop receiving emails from it, send an email to opencog+u...@googlegroups.com.

[neurop...@gmail.com]

It didn't show did it, hopefully this works.. This is bad because this really is an unimportant diagram in comparison to my other questions however now it seems like I am making it look important (Which it isn't) because It didn't post correctly. 

[Nil Geisweiller]

On 07/21/2017 04:58 PM, neurop...@gmail.com wrote:
> Ah thanks for letting me know about those rules, one more thing about
> this tho. Is there already a rule which let's opencog(chatbot) know that
> any statement made by a person or other information source it is talking
> to is what they think instead of basically storing it in the atomspace
> without the source (You hopefully get what I mean) Sorry if I am bad at
> describing what I mean

I don't know the specifics of that but I assume that it stores it as
someone's belief rather than its own. Then PLN can in principle modifies
the system's belief, assuming for instance that the person holding the
believe is known to be trusted. I don't think it's been tried though.

Nil

>
>
> Yeah, I've seen that diagram :) I posted one about how I am planning to
> implement this chatbot however it didn't post correctly, here it is
> (Hopefully it works this time)
>
>
> Thanks
>

> <https://lh3.googleusercontent.com/-qlOvGV3itDY/WXIIFZrb1SI/AAAAAAAAAA4/bfLrDHeX7IIutEf0GXgpucWg2sgmR0m7ACLcBGAs/s1600/PIUP.png>

>
>
> Three short comments:
>
> 1) PLN is a set of rules that run on top of the rule engine. Those
> rules, when triggered, alter the "truth values" on atoms in certain
> ways. You (or someone) should be able to create a different set of
> formulas for truth values, that emulate the NARS formulas. You could
> then run PLN deductions side by side with NARS deductions, and see
> which you like better (or even create yet other kinds of formulas)
>
> 2) we have "relex2logic" rules, which convert natural language to
> logic expressions. These are ... time-consuming to write, and by
> nature are incomplete and buggy -- because natural language is a
> fluid, changing thing. Hand-writing rules is a good exercise, but
> madness lies in that direction.
>
> 3) the current diagram of how things connect is kind-of-ish reviewed
> here:
> https://github.com/opencog/opencog/blob/master/opencog/eva/architecture/embodiment.pdf
> <https://github.com/opencog/opencog/blob/master/opencog/eva/architecture/embodiment.pdf>
>
> --linas
>

> On Thu, Jul 20, 2017 at 6:14 AM, <neurop...@gmail.com <javascript:>>

> it, send an email to opencog+u...@googlegroups.com <javascript:>.

> To post to this group, send email to ope...@googlegroups.com

> <javascript:>.

> Visit this group at https://groups.google.com/group/opencog

> <https://groups.google.com/group/opencog>.

> To view this discussion on the web visit
> https://groups.google.com/d/msgid/opencog/d4fa76b4-80d2-4ea6-96af-e072237b03fd%40googlegroups.com

> <https://groups.google.com/d/msgid/opencog/d4fa76b4-80d2-4ea6-96af-e072237b03fd%40googlegroups.com?utm_medium=email&utm_source=footer>.

> For more options, visit https://groups.google.com/d/optout

> <https://groups.google.com/d/optout>.

>
>
> --
> You received this message because you are subscribed to the Google
> Groups "opencog" group.
> To unsubscribe from this group and stop receiving emails from it, send
> an email to opencog+u...@googlegroups.com

> <mailto:opencog+u...@googlegroups.com>.

> To post to this group, send email to ope...@googlegroups.com

> <mailto:ope...@googlegroups.com>.

> Visit this group at https://groups.google.com/group/opencog.
> To view this discussion on the web visit

> https://groups.google.com/d/msgid/opencog/b8288b31-9201-42e2-b979-a101541639c2%40googlegroups.com
> <https://groups.google.com/d/msgid/opencog/b8288b31-9201-42e2-b979-a101541639c2%40googlegroups.com?utm_medium=email&utm_source=footer>.

[Linas Vepstas]

On Fri, Jul 21, 2017 at 1:07 PM, 'Nil Geisweiller' via opencog <ope...@googlegroups.com> wrote:

On 07/21/2017 04:58 PM, neurop...@gmail.com wrote:

Ah thanks for letting me know about those rules, one more thing about
this tho. Is there already a rule which let's opencog(chatbot) know that
any statement made by a person or other information source it is talking
to is what they think instead of basically storing it in the atomspace
without the source (You hopefully get what I mean) Sorry if I am bad at
describing what I mean

I don't know the specifics of that but I assume that it stores it as someone's belief rather than its own. Then PLN can in principle modifies the system's belief, assuming for instance that the person holding the believe is known to be trusted. I don't think it's been tried though.

If you mean "do we model other people's beliefs?" that answer would have to be "I wish". We still have multiple severe fundamental problems with creating and manipulating models of the world: It's still hard to wrap up into one big ball the fact that blob 42 in the visual field is a human face and its name is billy-joe-bob and the last thing bliiy said was foobar, which implies that billy believes that foobar. This is well beyond what we can do from first principles.

But, many corporations, e.g. advertisers e.g. cambridge analytica most infamously has developed a psychological model of hundreds of millions of Americans, and what they believe. Their model is crude, full of mistakes, and not based on first principles, but it is a model which they claim was used to elect Trump.  You (or we) could hack up something like this too, but "hack up" is the keyword here: we'd rather try to be more organic, more first-principles-driven.

--linas

[Ed Pell]

Linas, sorry for the stumbling about question, where can I find the ""relex2logic" rules, which convert natural language to logic expressions"? Thanks.

[Ed Pell]

opencog/src/opencog/nlp/relex2logic

ah I have it

[Ed Pell]

Is there an example of text input and rule output from the relex2logic ? Thanks.

[Nil Geisweiller]

On 07/31/2017 03:51 AM, Ed Pell wrote:
> Is there an example of text input and rule output from the relex2logic ?
> Thanks.

This half-complete example has some

https://github.com/opencog/opencog/blob/master/examples/pln/chicken-feet-or-pizza/chicken-feet-or-pizza.scm

Nil

>
> On Sunday, July 30, 2017 at 8:37:37 PM UTC-4, Ed Pell wrote:
>
> opencog/src/opencog/nlp/relex2logic
>
> ah I have it
>
> On Sunday, July 30, 2017 at 8:32:56 PM UTC-4, Ed Pell wrote:
>
>
> Linas, sorry for the stumbling about question, where can I find
> the ""relex2logic" rules, which convert natural language to
> logic expressions"? Thanks.
>

> --
> You received this message because you are subscribed to the Google
> Groups "opencog" group.
> To unsubscribe from this group and stop receiving emails from it, send

> an email to opencog+u...@googlegroups.com
> <mailto:opencog+u...@googlegroups.com>.

> To post to this group, send email to ope...@googlegroups.com

> <mailto:ope...@googlegroups.com>.

> Visit this group at https://groups.google.com/group/opencog.
> To view this discussion on the web visit

> https://groups.google.com/d/msgid/opencog/cc764cc1-c3cf-4c86-a6c3-f0a347ab7911%40googlegroups.com
> <https://groups.google.com/d/msgid/opencog/cc764cc1-c3cf-4c86-a6c3-f0a347ab7911%40googlegroups.com?utm_medium=email&utm_source=footer>.