Hmm, now I have a problem. I am signed in but the admin section keeps telling me I am not connected. I don’t see how to select a bot :/

In fact none of the admin tabs are working.

Attachments here do not seem to be working, so here’s a screenshot.

http://i799.photobucket.com/albums/yy275/Rogad/misc/hmmm_zps6dfb9aab.jpg

Are you sure it’s “Santa+bot”, rather than “Santa bot”? It may be that the API is not processing the URL encoding.

Yep that’s it Dave, thanks   I removed the ‘+’ and it works.

Still can’t access my bot though :/

It’s possibly a settings thing for your chatbot, then, I would think. I haven’t delved into things at all, so I could just be blowing smoke.

Make sure your bot is public, you cannot access private bots from rest without connecting and passing the user/key.

The website error seems like your connection timed out.  Try reconnecting and going back to browse and selecting your bot.

I will look into Yahoo pipes and an Http GET API.

I logged in again, but my bot Laura is not there. Poor Laura…

I see Laura there, and she seems fine,

http://www.botlibre.com/BrowseServlet?browse=Laura

I tired multiple different browsers, and don’t see any errors in the logs.  What browser are you using?

I see, I am misinterpreting your directions then.  I browsed the ‘Personal Bots’ assuming that meant my bots…

I do see her from that link, thanks !

I’m on Chrome.

RE: Roger/Dave HTTP GET API, Yahoo Pipes

Hi Roger, I looked into adding a forms style GET API, and it was pretty easy to add.  I am in the process of documenting the API, but the basics is here,

HTTP FORM GET API

form-chat

Description - Receives a chat message and returns the chat bot’s reply as an XML document.

URI

Parameters - application, instance, user, , token, conversation, message, emote, correction, offensive, disconnect, includeQuestion.

application - OPTIONAL: The application ID.  If not passed, the application will be anonymous.

instance - REQUIRED: The ID of the bot to chat with.  The bot’s name can also be used, but the ID is better as it is guaranteed to be unique.

user - OPTIONAL: The ID of the user who is sending the message.  The user must be registered with BOT libre. If not passed the user will be anonymous.  The user is required if the bot is private.  The user/password are only required on the first message.

password - OPTIONAL: The password of the user who is sending the message.  A token can also be used.

token - OPTIONAL: The token of the user who is sending the message.  A token can be obtained through check-user, and is valid until reset.

conversation - OPTIONAL: The conversation ID for the current conversation.  This must not be passed on the first message, but will be returned by the response, and should be used for all subsequent messages in the conversation.

message - OPTIONAL: The chat message to send to the bot. The message can be omitted if you wish the bot to start the conversation (if the bot has a greeting set). The message must be encoded in the URI.

emote - OPTIONAL: A emotion to tag the message with.  This is one of LOVE, LIKE, DISLIKE, HATE,
RAGE, ANGER, CALM, SERENE, ECSTATIC, HAPPY, SAD, CRYING, PANIC, AFRAID, CONFIDENT, COURAGEOUS, SURPRISE, BORED, LAUGHTER, SERIOUS.

correction - OPTIONAL: A boolean that defines the chat message is a correction to the bot’s last answer.

offensive - OPTIONAL: A boolean that defines the bot’s last answer as offensive.  The message will be flagged for the bot’s administrator to review.

disconnect - OPTIONAL: A boolean that defines the end of the conversation.

includeQuestion - OPTIONAL: A boolean that indicates the question should be included in the response.

Example

Example Result

That’s great, thanks !  I will have a play in the next few days hopefully.

Thanks James, got your example working in Yahoo! Pipes right away.  Next question, do you have any conversion utility for AIML into whatever chatbot language you use?  I haven’t even taken a closer look yet, just looking for quickstart shortcuts before diving in.  Is the Botlibre knowledgebase even uploadable/downloadable??  I like to be able to hit things from all sides, automatic, semi-automatic, and manual.  I especially like to use “power tools” offline, particularly when there is no “backend API” into the knowledgebase (as opposed to the “frontend API” for the dialog system).

On the theoretical side, a SaaS chatbot system is basically a cloud hosted interpreter for a “chatbot” language, or dialog system.  Ideally, this should be as modular as possible.  So far, all the SaaS chatbot systems only offer modular frontends, and not modular backends.  There are lots and lots of both database and knowledgebase formats and systems.  SaaS chatbot systems need to have more modular backends for accessing, converting and connecting data.  There are now lots and lots of business solutions for converting and connecting data in the cloud, for example SalesForce at al.  I want SaaS dialog systems that can access this kind of widely available data.  Unfortunately, the cloud based turnkey business services like SalesForce are not yet up to doing experiment AI on their own.

... for example SalesForce et al.  I want SaaS dialog systems that can access this kind of widely available data.  Unfortunately, the cloud based turnkey business services like SalesForce are not yet up to doing experimental AI on their own.

Hehe, what can I say, I’m online from a beach cafe on the Indian Ocean..

> http://www.botlibre.com/import.jsp

James, I’m looking at “Import Data” for my private bot now.  I don’t see anything about what kind of data is accepted, or how it’s handled internally….  I have lots of AIML files.  I have lots of chatlogs in CSV.  I have millions of tweets downloaded in CSV, like one-sided chatlogs; I have both questions and answers, just not keyed to each other yet.

I don’t see any “Export” capability for the knowledgebase….  Normally, I would download a “forked” knowledgebase, copy the format, and build a new one offline for re-upload.  I’m unlikely to surrender large amounts of proprietary data, especially without some way to “repatriate” it.  Can Botlibre chatlogs be exported, for instance monthly in CSV?

Perhaps you could explain a little about how uploaded data is handled internally?  In any case, it would be nice to have some kind of generalized AIML conversion tool (not just for Botlibre), which could calculate and convert any AIML file into a “chatlog” format, containing both all the potential questions and their responses….  ;^)

RE: Marcus,

Thanks for all the questions.  I’ll do my best to answer them.

You can import chat logs from the Import tab under Admin,

http://www.botlibre.com/import.jsp

The chat log format is basically,

So,
<name>:<question>

The names are important as the bot will use them in the context, i.e. “Hey Marcus” becomes, “Hey <target>”.  Also 2 new lines indicate a new conversation, the bots learns things in context, so separating the conversations is important.  What format are your chat logs in, I could add support for another format.

The bot will actually relive the conversation and learn from it, so the import of a big file can take a while.  I am looking into adding an import that just adds the responses and does not try to learn.

There is no AIML import yet, but it is something I am investigating.

You can also import data from the web such as Freebase, or Wiktionary (the bot will actually look up every new word in Wiktionary, and “What is” questions in Freebase).  I am working on other web imports, such as forums, FAQs, etc.

The backend data format is a big database.  The bot actually has its own brain, so the format is complex.  You can browse it from the Memory tab,

http://www.botlibre.com/memory.jsp

I would like to have en export, that exports the data as XML, or as a Postgres .sql file, but have not implemented that yet.  I can send you a .sql file through email if you ask for it.

The bot does not require any programming, but you can program it using what I call the “Self Scripting Language” (not to be confused with the old Self language).  The bot will actually program itself as it learns (which you can disable by disabling comprehension).  You can see all of the code from from Program tab,

http://www.botlibre.com/self.jsp

The language is basically a state machine language.

I’ve spent the past two days traveling by train across South India, and thinking about this thread….

The AIML - Machine Learning connection is a significant one; because, a lot of people have spent a lot of time thinking about how to connect AIML with machine learning.

This chatlogs solution is extremely elegant; as, many people have many years of chatlogs available.

For instance, I have at least a year and a half of substantial chatlogs from Verbotsonline, in simple “input, output” format (without any speaker identification per se).

What this means to me is that if people have a few years of chatlogs - from any chatbot system - then machine learning may be applied, regardless of chatbot language.  In fact, an AIML chatbot for example could simply have machine learning attached to the chatlogs, and discard the AIML altogether after any specified “critical mass” was achieved.

I don’t know of any other SaaS chatbot that offers machine learning for uploaded chatlogs.

What I really want to do with my chatbot is to actually TALK to it via Skype (NOT Skype IM), and to be able to teach/correct it by voice interaction - not so differently than one would a child for instance.  And, a lipsync-ed avatar (talking head)  would be a plus for Skype.  This can already be achieved to a degree using the AIML-based Guile3D Denise with a program called ManyCam; however, to my knowledge no one has yet applied machine learning to Denise.