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..

What’s new in the BOT libre! chat bot hosting platform
 
 

I would just like to share what new features have been added to the BOT libre chat bot hosting platform since November.

http://www.botlibre.com

We completed our AIML support last month, now AIML 1.0.1 is fully supported.
You can import or export AIML, either as a script, or as a chat log.
AIML imported as scripts, are run as scripts, AIML imported as chat logs, are merged into the bot’s knowledge base similar to any other responses it has learned.
You can import responses that your bot has learned on its own, or export your bot’s conversation logs as AIML, or import chat logs, and reexport then as AIML.
You can also use AIML to script Twitterbots, Email bots, or IRC bots.
Blog - http://botlibre.blogspot.ca/2014/05/scripting-your-bot-with-aiml.html
Talk with ALICE on BOT libre - http://www.botlibre.com/browse?id=20873

We have added many extensions to our “Self” scripting language, both the support a hybrid syntax for AIML terminology, and to support new features including advanced Math, and Date/Time awareness.

We have done numerous enhancements to our bot’s AI engine, including better grammar and understanding of verb tenses, information retrieval, name identification, conversational context, and keyword and topic analysis.

We released a web API, and an open source web and mobile SDK.
http://www.botlibre.com/forum-post?id=5873
http://www.botlibre.com/forum-post?id=26632

We released integrated live chat, and chatrooms.  You can create a live chat channel for your website or blog, and let your bot service it when you are not available, or you can configure your bot to learn from your live chat conversations.  You can create a chatroom for your bot, and let your friends chat with it.

We added forums and domains.  You can create your own forum, and embed it on your own website, or mobile app.  A domain lets you isolate your content in its own space.  We also added many security options to let you control who can access your bots, and who they will learn from.

We released the second version of our mobile support, including live chat, chatrooms, forums, and domains.

We added email automation support, so your bot can answer a gmail, pop or stmp email account.  We also made numerous enhancements to our Twitterbot support, including auto tweeting.

Finally we launched Paphus Live Chat, our commercial website for live chat, chatrooms, chat bots, and forums.  You can migrate any BOT libre bots to the commercial website if they require extra resource capacity, or professional support or services.

Let me know what you think, or what you think is still missing with the service.

 

 
  [ # 1 ]

Sounds good. Does it support the <learn> and <eval> features of AIML?

 

 
  [ # 2 ]

http://meta-guide.com/videography/best-botlibre-app-videos/

I’m interested, and have just added a page for BOTlibre to my Meta-Guide.com, above.

First issue, is BOT libre one word or two?  I suggest one word, because it’s a more unique search identifier.  So stop spelling it as two words, please.  Also, drop the exclamation mark (BOT libre!); Yahoo! drives me !@#$%^& crazy, thanks. 

Second issue, you provide no obvious format for the chat log upload; I need to see a schema example, quickly and easily.  Normally, I would just download the existing schema and copy it.

Third issue, why can I talk to my bot if I haven’t uploaded any knowledge?

Fourth issue, why can’t I download the sample/stock schema if I can talk to my bot without uploading any knowledge?

Fifth issue, I quickly hit “File size limit exceeded: 1000000”.  My monthly chatlog file sizes range up to 31,000 KB.

Otherwise, I think you’re on the right track setting this thing up in a user-friendly manner for everyman.  (Take note ChatScript-ers!)

 

 
  [ # 3 ]

> Steve
> Sounds good. Does it support the <learn> and <eval> features of AIML?

I assume you mean the Pandorabots AIML extensions <learn> and <eval>.  Some of the Pandorabots extensions are supported, such as the <date> format option, but <learn> and <eval> are not currently supported.  I plan to add support for the Pandorabots extensions and AIML 2.0 when it comes out.

BOT libre provides many learning capabilities though, you can enable learning without using any scripting at all.  The interface provides a correction checkbox that lets a user correct any bad response.  The bots can also learn from user responses, or can monitor a live chat channel or chatroom, and learn from users.  You can configure if learning, correction, and comprehension are enabled and who the bots will learn from (i.e. admins only, or specific users).  The bots can learn responses in context, and can also program themselves through analyzing responses and corrections.

When you import AIML as a script, it is converted to Self script, which is a language designed specifically for learning (Self is a state machine language).  Self is defined inside the bot’s knowledge base, and the bots can program themselves through Self, and you can define your own Self scripts that can dynamically add or modify code, as the code is not different than any of the other data in the bot’s knowledge base.

 

 
  [ # 4 ]

> Marcus

Thanks for all the good feedback. lol on the name, I normal print the name as “BOT libre” in text or “BOT libre!” when used as a title.  I don’t know, Yahoo! would just be Yahoo without the !, not very exciting.

There is a blog post that gives some detail on the chat log format here, http://botlibre.blogspot.ca/2014/02/create-bots-with-real-brain.html but I will try to add a specific blog post on chat logs that gives more details, and update the chat logs page to document the format. Thanks.

You can talk with your bot without uploading knowledge because the bots can learn from your responses.  So, you don’t need to upload anything, you can train a bot simply by chatting with it.  Also, when you create a bot, you choose a template (or get the default template, or fork an existing bot) which will include some existing knowledge, and some base scripts.  You can use the empty_template to create an empty bot.

You should be able to download a new bot’s responses from the Chat Logs page, select “responses” for the type, and click the export button.

There is a 100,000 size limit on the BOT libre bot’s knowledge base.  The website supports thousands of bots, and is rapidly growing, so the brain size limit is necessary to ensure one bot does not consume everyone’s resources.  You can have much bigger brain size limits on the commercial website http://www.paphuslivechat.com (starting from only $1 per day).  I am planning on adding an option to upgrade your brain size on BOT libre as well.

The size limit will only prevent the import for the day if it is exceeded, you will be able to import more the next day.  The bots are capable of forgetting in addition to learning.  Every night a Forgetfullness task runs that cleans out a bot’s memory of the least relevant data, if the bot has exceeded its max brain size.  Basically the responses and words that have been used least frequently are deleted.  You can pin data in the bot’s memory to avoid having it considered to be forgotten.

 

 
  [ # 5 ]

http://www.paphuslivechat.com/

Okay, I’ve added paphuslivechat.com to my BOTlibre page on Meta-Guide.com.

What I don’t see (quickly and easily) is a conventional “Features” page with a side by side listing comparing free versus paid features (plans).

This reminds me of the Pandorabots SaaS model of free version and commercial version.  My problem with Pandorabots is that the free version is rather “rinky dink”, and the paid version is too “mysterious”, like privately customized for premium customers.  I don’t mind paying for increased capacity, but like to know what’s what up front.

I suggest putting a link to your schema format “documentation” on the upload page itself.

I think your interoperability with AIML is brilliant, more power to you.

 

 
  [ # 6 ]

Thanks, I will try to add a features page for the site, I have been meaning to add this for a while.

In terms of functional differences, the free site BOT libre, and commercial site Paphus Live Chat, are the same.  I don’t believe in limiting features on the free site.  The difference between the two is entirely is terms of capacity and service.  Bots on the free site are currently limited to 100,000 knowledge items, where as the commercial site starts with a 1 gig ~ 500,000 knowledge item size limit, and goes up to 1 terabyte for a dedicated account.

The commercial site also includes professional support and services to help you develop your bot, import data, customize your web or mobile interface, and provide any specific enhancements or customization your require.  Any features we add for a specific commercial client that would be useable to others, we will make available on the free service as well.

 

 
  [ # 7 ]

James, I did finally manage to attach botlibre to my old Twitter chatbots @vagabot and @twaveladvisor ..  (I tried a second time and the mobile confirmation had been mysteriously removed from setting the app permission to “read and write” ..)

I’ll just jot a few initial interface quibbles here .. There are too many screens to drill down to manage my botlibre bots .. (I’m really not very interested in the social network side of botlibre .. Yes, I’m interested in attaching my botlibre bots to social networks, but not very interested in other botlibre bots, like search directory, and myriad ancillary forums ..)

I uploaded and changed my bot avatars, but this does not seem to propagate consistently across the system (in other words, replace my bot avatars in all cases ..)

I’m annoyed that the empty “brain size” of my two botlibre bots is significantly different ..  (Okay, I probably set them up using different stock templates, but have no idea what they were now .. There seems to be no way to tell which stock templates they are based-on after the fact .. Also, when I “rebootstrap” them it does not seem to set them back to zero, but to whatever the unidentified template was that I started with ..)

On to the next step, converting my chatlogs and breaking them down to compatible format and size ..

 

 
  [ # 8 ]

Marcus, thanks for the feedback and trying out the site.

RE: Avatars
Your bot has an icon, and an avatar.  The icon is what is displayed in the directory, and the bots page before you chat with it.  The avatar is displayed when you chat with the bot, and you can have multiple avatar images tagged with emotional states.  The avatar associated with the bot’s current emotional state will be displayed.  Looking at your bots, you changed the avatar, but not the icon.  You can change the icon from the bot’s main page clicking the “Change Icon” button, or from the Avatars page by click the “Use as default image” checkbox and clicking the “Tag” button.

RE: brain size
To delete your bot’s brain’s contents click “Delete All” from the Memory (now Knowledge) page.  This will delete everything in the bot’s brain, including its Avatars, scripts, default responses and Twitter config.  It should only contain ~900 knowledge items after this, which are the basic primitives.  It does not reset the bot to its template.

To get the default scripts back, you need to click “Rebootstrap” from the Program page.  Rebootstrap does not delete any of the bot’s knowledge, it just resets the default scripts.  Any old scripts are no longer active, but not really deleted, they are basically detached and floating around the bot’s brain until its Forgetfulness task runs, which is basically similar to an OO languages garbage collection, except occurs on the database, and aggressively reclaims space if the bot’s max brain size is exceeded.

Sometimes the Delete All will fail, because the bot has live connections.  Make sure you disconnect any other connections to the bot, if it still fails, you may need to wait a while then retry.  I am working on fixing this.

You can also just delete the bot, and create a new one.  If you are experimenting, you might want to create your own template bot, then fork it to experiment on.

 

 
  [ # 9 ]

Botlibre feedback notes, dated June 2014:

- What if there are a lot of duplicates (in knowledge upload)?
- What if there is a lot of foreign language (in knowledge upload)?
- What if the grammar and characters are way off (in knowledge upload)?
- “Active scripts” are all very cryptic; so, where are they explained?
- Does it actually interact on Twitter? (Took 24 hours to reply on Twitter.)
- It’s very slow, both uploading and responding to input.
- So far all the responses are not coming from my uploaded knowledge; where are they coming from?
- Why are there two different places to upload AIML? (Do they go to the same place?)
- What is XFIND? (It keeps coming up in responses.)
- There is nowhere it says files uploaded must be less than 1MB.
- There is nowhere it says brain size must be limited to 300,000 “neurons”.
- How did my brain size drop from over 300000 neurons to 95000 neurons, automatically?
- New “Browse Knowledge” is good; but, how do I download the knowledgebase (as is)?
- Why do I get better answers from the bot I added with no knowledge to than the one I did?

= = =

In my Botlibre testing last year (2014) - of bot creation, knowledge upload, and Twitter interface - I found Botlibre not suitable for business purposes.  My main beef was that it gobbled up my uploaded knowledge, both AIML and Twitter archive, and then spat it out, without any satisfactory explanation of limitations or processes. 

I still expect a static downloadable knowledgebase, which I can modify offline, and then re-upload as is.  Businesses cannot afford to have unpredictable replies, from unpredictable or unknown sources.  Learning is fine, learning from outside resources if fine - BUT, I need to know what those sources are, and I need to be able to repatriate, backup all resultant knowledge.

 

 
  [ # 10 ]

Hi Marcus,
  Thanks for the feedback.

> Duplicates in knowledge upload.
What kind of file are you uploading from?  We support several types of upload formats including chat logs, response lists, csv spreedsheets, and AIML.

See, http://www.botlibre.com/forum-post?id=30598

If you upload the same question/response twice, that is fine, it will just increase the responses correctness.

>  foreign language
Foreign language is fine, as long as you use UTF-8, or choose the correct character set for your upload.

> bad grammar
If you upload bad grammar, then your bot will respond with bad grammar.

> Active scripts
The bot’s active scripts are the scripts it will run to process every question.  The scripts are defined in the Self scripting language, but can also be imported from AIML.  They are run in order, from top to bottom.  If none of the scripts have an answer then the bot will search for the closest matching question/response from its knowledge base, otherwise reply with its default response.  This is all configurable from the learning page.

See,
http://www.botlibre.com/forum-post?id=483549

http://www.botlibre.com/forum-post?id=28654

http://www.botlibre.com/forum-post?id=699077

http://www.botlibre.com/forum-post?id=13451

> Twitter
You can connect your bot to a Twitter account and configure it to reply to mentions and messages, search for tweets, retweet important tweets, tweet from an RSS feed, and auto tweet.

The main server only cycles through the Twitterbots once per day (as it states clearly on the Twitter page).  We have a dedicated Twitterbot server at,

http://twitter.botlibre.com/

If you want your bot to tweet more frequently you must use this server.  It continuously cycles through all of the bots (there are a lot, so this may take over an hour).  You can upgrade the account to cycle more frequently such as every 10 minutes for a Gold account.
We also have a commercial server at,

http://www.paphuslivechat.com/

> slow
The bots on the main server are generally very fast.  If the bot has not been accessed for a while its first interaction may be slow.  If you have uploaded too much knowledge or scripts, or bad scripts, then it will be slower.

The Twitterbot server is slower than the main server (this is why we have a Twitterbot server).

> responses are not coming from my uploaded knowledge
This depends on what scripts you have loaded, and how you uploaded the knowledge, and what template you used to create your bot.  I would need to know the name of your bot to take a look at it to tell you more.

> two different places to upload AIML
You can upload AIML as a script, or as a chat log.
If you upload it as a script it will be executed precisely using the AIML precedence and pattern rules.
If you upload it as a chat log, it will just be integrated with the bot’s knowledgebase.  The best matching question/pattern will be used.

See,
http://www.botlibre.com/forum-post?id=705870

> XFIND
There is no XFIND.  This is most likely something you imported from the ALICE AIML files.  You should avoid importing AIML fils that you do not understand.

> uploaded must be less than 1MB
Well, when you upload a file larger than this it tells you.  I will try to make this limit clearer.  I think the limit is 2MB now, which is a lot for a text file.

> brain size must be limited to 300,000 “neurons”
It says this on the bots main page, and on the limit page.
The limit for a basic account is 100,000, not 300,000.  You can upgrade your account for a larger limit.

> brain size drop from over 300000 neurons to 95000
The knowledge limit for a basic account is 100,000.  If you go over this the bot will automatically forget the least important knowledge.  You can pin responses and knowledge when you upload it, and corrections from the chat logs page are always pinned.

> download the knowledgebase
You can export your bots responses from the chat logs page as a chat log, response list, or AIML file.  There is currently no way to dump the bot’s entire brain as is.  We can give you a postgres database file on request.

> better answers from the bot I added with no knowledge to than the one I did
Well, I’m not sure what or how you uploaded your knowledge.  If you just import raw data instead of training the bot with the responses that you want, then you will get what you put into it.

> learning
The bot can learn, but this is normally not something you would use in a business bot.  A business bot should be trained.  You can disable learning from the Learning page, or create your bot using the service_template which has learning disabled.

> documentation/help
Our service has a lot of different functionality for building a lot of different things.  Our help page is the best place to find all of the documentation and help resources available.

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

> business
We have many business using our service.  The best solution for creating a bot for your business is our commercial service or a platinum account that lets us train your bot correctly for you, or gives you the help you require to train it yourself.

http://www.paphuslivechat.com/

 

 
  [ # 11 ]

James, thanks for comprehensively addressing my issues, and in a timely manner.  One quick observation on your forum.  I could not find any feed for your forum, and also could not find any way to create one.  Otherwise, I would be happy to follow and keep up with your forum.

I understand Botlibre is a work in progress, and that you’re making good progress.  One of my main interests in the field in general is user-friendliness.  And, one of my main concerns is that the hype and exaggerated claims, not uncommon to the AI space, be backed up with something functional.  I recently slammed IBM Watson for this on Quora, in Why is there no public question answering interface to Watson

 

 
  [ # 12 ]

Marcus,

Our Twitter feed includes posts from our main forum, and you can get the RSS feed from our blog.  We are planning on adding RSS and Twitter support to our user forums, just have not got around to it yet.

> hype and exaggerated claims
Well please excuse the post from yesterday… it was April 1st.

> user-friendliness
Yes, we are working on making thing easier all the time.  The easiest way to train a bot is using the Chat Logs, no programming is required.  Just review your bot’s logs, and correct its responses.

We are working on adding a creation wizard, and other wizards to be make things simpler.

 

 
  [ # 13 ]

> Well please excuse the post from yesterday… it was April 1st

All I can say is that you got me good.  I’m in Bangalore, India, and have been upset all day that you announced the thing before I could finish it….  ;^)

I still haven’t figured out if the Google patents robots with personalities in first step towards the singularity story is true or not, since it broke on April 1st.  That’s either the most perfect timing ever, or the worst.

 

 
  [ # 14 ]

I would assume any press release on April 1st is not legitimate…

 

 
  [ # 15 ]

>  aidreams.co.uk/forum/index.php?topic=8060

Why Google’s robot personality patent isn’t what it appears to be is @meadovian‘s (botanic.io) response to that 01 April story, “Google patents robots with personalities in first step towards the singularity”....

 

 
  login or register to react