Introducing ALICE 2.0

What is ALICE 2.0?

ALICE 2.0 is a new AIML chatbot personality designed specifically for mobile devices such as handsets and tablets.   ALICE 2.0 includes a number of innovations including:

• Enhanced learning to remember client profile, preferences and personal information.
• Access to external web services and even other chatbots provides richer, more accurate responses.
• External services provide up-to-date knowledge of current events and popular culture.
• Responds to requests for device actions including calls, SMS, email, maps, and device controls.
• Utilizes AIML 2.0 features including Zero+ wildcards, AIML Sets, Maps and a Triple-Store knowledge base.
• Contains an ontology of facts about everyday objects.
• Positive, friendly, gentle personality easily adaptable for commercial applications.

How does ALICE 2.0 differ from original ALICE?

The original ALICE chatbot, under development since 1995, is one of the most popular chat bots.  Based on AIML 1.0, the ALICE chatbot contains about 100,000 units of knowledge. ALICE has won numerous awards in Turing Test and chatbot contests, and in fact is the most awarded chatbot.   

The original ALICE is however a bit dated.  Many developers found the personality responses of original bot unsuitable for applications such as help desk, FAQ bots, and sales.  Moreover many of the responses were hard-wired to what are now, 15 years after they were written, obsolete cultural references.  ALICE’s personality reflects the mindset of a web-based entity.  Some responses assume the user is sitting in front of computer, using a web browser, and typing rather than speaking.   Finally, much of the knowledge in the ALICE brain is “Wikipedia-style” facts.

ALICE 2.0 repairs the shortcomings of the original ALICE.  Because ALICE 2.0 can obtain factual information from external services and other bots, the number of AIML categories required is reduced.  These external sources provide up-to-date cultural references and information about current events.  Also making use 2.0 features means that, in many cases, the work done by hundreds or thousands of AIML 1.0 categories can be compressed into a single category. The personality of ALICE 2.0 is positive and friendly, and designed to play the role a mobile device robot entity.

Where can I get ALICE 2.0?

ALICE 2.0 is available in the CallMom BASIC app on Google Play.   The CallMom BASIC - ALICE 2.0 app offers all the features of ALICE 2.0 as well as enhanced privacy controls.  Because the bot resides directly on the device, the user has more choices about whether his interactions and personal information are recorded on a remote server.

Chatbots 3.3 Conference Report

The Chatbots 3.3 conference, held in Philadelphia on March 23, was once again this year a huge success.  Sponsored by the ALICE AI Foundation, Pandorabots and Seed Philly, this was the fourth annual Chatbots conference held in Philadelphia. We enjoyed a collection of high quality presentations covering legal and business issues, mobile apps, research topics, new languages and technologies, animated avatars, and even an application of chatbots in the field of psychiatry.  About 25 people attended the event, and many more participated via live webcast.  You can view additional photos on Flickr by Wolf Paulus.

The final program of speakers is posted here, along with links to the presentations we have received so far.

• Brad Danenberg, Seed Philly, Venture Capital Opportunities for Chatbot Technology
• Bill DeSmedt, Alliance Global Services, MetaLang: A Knowledge Based Conversational Agent
• Mohan Embar (Loebner Prize Winner, 2012), Mohan Embar, Inc., Your Chatbot: Moon Shot or Low-Hanging Fruit?
• Dave Morton, Geek Cave Creations, Building Chatbots and Apps in a Web-Based World
• Noah Petherbridge, DreamHost, Chatbots in Web Applications with RiveScript
• Wolf Paulus, Inuit, Voice enabling Chatbots
• Jeremy Spiegel, MD, Casco Bay Medical, Feelbetter: Bringing a Psychiatrist Chatbot to Life
• Chris Shaw, Xprevo, The Arts, Collaboration, and Narrative Interactivity
• Francis X. Taney, Esq, Stevens & Lee, Legal and IP issues for Botmasters
• Richard Wallace (Loebner Prize Winner 2000, 2001 and 2004), ALICE AI Foundation, AIML 2.0 Chatbot Technology for a Mobile Era

Live video streaming of the conference was provided by Altify:Software.  We had more than 200 viewers during thr course of the event, giving this conference by far the largest reach in the Chatbots 3.x series.  Viewers were able to submit questions for the speakers to the moderator in real-time.  You can view the videos on the Ustream.tv aimlinstructor channel.

The organizers of the Chatbots 3.3 conference would like to thank Altify:Software for video streaming support, Universal Arts for camera and audio work, and especially Seed Philly for providing the conference space.  We look forward to seeing you at the next conference!

Chatbots 3.3 Conference March 23

The sponsors of the Chatbots 3.x conference series are pleased to announce the 4th annual conference on chatbot and virtual assistant technology.  The Chatbots 3.3 conference will be held on March 23, 2013 in Philadelphia, Pennsylvania USA.  This years conference will emphasize all the latest developments in mobile apps and virtual assistant technology.  We’ve already attracted an impressive lineup of speakers, including three former Loebner Prize winners.

Latest updated Speaker roster:

• Kevin Copple (Loebner Prize Winner 2002), Tianjin Ring-Top Petroleum Manufacturing Co., Ltd., The Role of Chabots and their Authors in the coming Singularity
• Brad Danenberg, Seed Philly, Venture Capital Opportunities for Chatbot Technology
• Bill DeSmedt, Alliance Global Services, TBD
• Mohan Embar (Loebner Prize Winner, 2012), Mohan Embar, Inc., Your Chatbot: Moon Shot or Low-Hanging Fruit?
• Karsten Fluegge, Pannous, Mobile Apps and Web Services
• Dave Morton, Geek Cave Creations, Building Chatbots and Apps in a Web-Based World
• Noah Petherbridge, DreamHost, Chatbots in Web Applications with RiveScript
• Jeremy Spiegel, MD, Casco Bay Medical, Chatbots in Psychiatry
• Francis X. Taney, Esq, Stevens & Lee, Legal and IP issues for Botmasters
• Richard Wallace (Loebner Prize Winner 2000, 2001 and 2004), ALICE AI Foundation, AIML 2.0

Registration: http://www.eventbrite.com/event/3126769251

Program AB -- AIML 2.0 Reference Implementation

Motivated by the technological and social changes of this mobile era, the ALICE A.I. Foundation has released a new, free, open source Java implementation of the AIML 2.0 Draft Specification.   Following the tradition of naming AIML interpreters after letters of the alphabet (Program A (SETL), Program B (Java), Program D (Java), Program E (PHP), J-Alice (C++), Program N (C++), Program O (PHP/MySQL), Program P (Pascal),  Program Q (C++, Qt), Program R (Ruby), Program V (Perl), Program W (Java), Program Y/PyAIML (Python), and Program Z (Lisp)), this new system is called Program AB to indicate a "fresh start" with AIML 2.0.

Program AB serves as the reference implementation of AIML 2.0 and is currently the primary platform for experimenting with new features.  The program is structured to enable developers to easily extend AIML with custom tags.  Program AB implements AIML 2.0 features including Sets and Maps, Zero+ wildcards, and the capability to connect to remote bots and web services through the new  tag.

Significantly, Program AB implements some memory and storage optimizations that make it possible to run a sophisticated chatbot on a mobile device or embedded system.   Versions of Program AB have already been tested on Android phones and tablets, as well as the Raspberry Pi single-board computer.  Embedding the artificial intelligence engine in such devices provides the possibility of creating virtual assistants that work without an internet connection, as well as enhancing user privacy because the application does not rely on transmitting conversation data to a remote server.

Program AB is available for beta testing at https://code.google.com/p/program-ab/.

AIML 2.0 draft specification released

The ALICE A.I. Foundation has released a draft specification of a major uprgade to AIML (Artificial Intelligence Markup Language), version 2.0 of the language.   AIML is a widely adopted standard for creating chat bots and mobile virtual assistants like ALICE, Mitsuku, English Tutor, The Professor and many more.  A decade or more of experience building AIML bots has informed the need for new features in the language.  The need for new features however is tempered by the goal of keeping AIML as simple as possible, especially for non-programmers.

AIML 2.0 combines Pandorabots extensions to the language, the Out-of-Band (OOB) tags that implement actions on mobile devices, and a collection of new AIML features.   AIML 2.0 includes a new tag so that the bot can access responses from remote web services and even other AIML bots.  The AIML 2.0 draft includes new wildcards, conditional loops, local variables, and the ability to specify attribute values in tags.  One significant advance, AIML Sets and Maps, is described in a companion document.

The AI Foundation will shortly be releasing a new open source AIML interpreter in Java to serve as a reference implementation of AIML 2.0.  To join the discussion about AIML 2.0 please visit forum.alicebot.org and the chatbots.org forum.

What’s new in AIML 2.0?

• Zero+ wildcards: new wildcards that match 0 or more words.
• Highest priority matching: select certain words to have top matching priority.
• Migrating from attributes to tags: more dynamic control of attribute values.
• AIML Sets: match inputs with sets of words and phrases.
• AIML Maps: map set elements to members of other sets.
• Loops: Iterations in conditional expressions.
• Local variables: variables with scope limited to one category.
• Sraix: access external web services and other Pandorabots.
• Pandorabots extensions
• date: formatted date and time.
• request: access previous input request history.
• response: access previous bot response history.
• unbound predicates: check if a predicate has been set or not.
• learn: learn new AIML categories.
• learnf: learn new AIML categories and save in a file.
• explode: split words and phrases into individual characters.

• OOB (Out of Band) Tags: AIML extension for mobile device control

Full document: AIML 2.0 Draft Specification

Guild 3D Studio Virtual Assistant Denise

The ALICE A.I. Foundation recently received a review copy of the virtual assistant Denise from Guile 3D studio.   More than just a chatbot, the Denise system is a complete desktop assistant that includes a customizable talking animated avatar, is integrated with PC functions and web services, and even includes a full AIML development environment to create and modify the chat bot personality.   The Denise bot can be deployed on a PC desktop, or integrated into commercial applications such as kiosks and office systems.

Denise is the creation of Brazilian developer Guile Lindroth, who was inspired while working 

with a person with a serious upper body disability.  He decided to start to build a "Virtual Nurse" application, that could help her to operate a computer. Lindroth soon noticed that not only disabled users, but many others as well, were in search of virtual assistant technology.   To meet the need, he says, "I've joined my passion for 3D graphics with my skills as a System Annalist...to develop my own graphic engine that could show more realistic avatars as well as a new artificial intelligence interpreter."  

The Denise software implements the full AIML 1.1 standard as well as some extensions.  Unlike some chat bot systems that have been integrated with browsers, the Denise system includes its own integrated web browser.  The AIML extensions make it possible to control events in this browser, such as looking up information on Wikipedia.  Other extensions provide for integration with third-party applications including Skype, Email and Facebook.  It is even possible to connect Denise to an AIML bot running remotely on Pandorabots.

When you install and launch Denise, she initiates an interview to learn a little about you.  She remembers your name, age, gender and optionally much more information.   The system includes an integrated speech recognition system from Nuance, and Denise takes the user through a speech recognition training session to improve the accuracy of the recognizer.   After training, the user may begin a conversation with Denise, asking for information, send and receive emails, translating languages, or just enjoy a chat with the Denise personality.  The platform also includes an integrated AIML editor, so that the user can add his own custom voice commands. 

Guile Lindroth was able to secure an investment to launch his company and the Denise project.  He says his biggest challenge is "keeping the dream alive...(while) competing against giants."   Lindroth sees three major areas of technical advancement: improved realism of avatars, improved speech recognition, and improved artificial intelligence.  Because the Denise system is designed to help computer users complete tasks, Lindroth says it doesn't have to be "out of this world AI", but AI that is good enough to understand and assist users.  Forseeing a time when ubiquitous, learning virtual assistants will control home automation systems, serve us in our cars, keep us informed and connected, and act as virtual office assistants at work, Lindroth says "We are actually not far from this day."

Denise is available in three versions from Guile 3D.

Answer Devices Releases Mobile Assistant Benchmark

Our friends at Answer Devices have released a very useful data set for anyone developing or testing a mobile virtual assistant like Siri or CallMom.  The "1000 questions every mobile assistant should be able to answer" includes sample inputs that cover a range of topic areas including apps, contacts, games, knowedge, dialing, SMS, maps, search.  Borrowing from published transcripts with popular virtual agents, Youtube videos, CallMom log files and even Apple TV commercials, the sample data is designed to test both speech recognition and response accuracy.  

An Apple critic, Gene Munster of Piper Jaffray, published a study of Siri in June in which he tested the app with 1600 sample inputs in both quiet and noisy background environments.  Although the reported result was about 65% accuracy, it was difficult to verify because Munster did not make the sample data  public.  As far as we know, the Answer Devices data set is the only public benchmark test suite available to compare mobile virtual assistants.

Pandorabots has offered to test any virtual assistant that has an API with the Answer Devices data.  If you are developing a virtual assistant and would like Pandorabots to run the test for you, please contact info@pandorabots.com.