Lecture 04: Knowledge Representation

Prof. Ray Larson & Prof. Marc Davis UC Berkeley SIMS Tuesday and Thursday 10:30 am - 12:00 am Fall 2003 Lecture 04: Knowledge Representation SIMS 202: Information Organization and Retrieval Credits to Warren Sack for some of the slides in this lecture

Today • Review of Categorization • Knowledge Representation • The Vocabulary Problem • Commonsense • Cyc • Discussion Questions • Phone Project Overview and Assignment 2 • Action Items for Next Time

Categorization • Processes of categorization are fundamental to human cognition • Categorization is messier than our computer systems would like • Human categorization is characterized by • Family resemblances • Prototypes • Basic-level categories • Considering how human categorization functions is important in the design of information organization and retrieval systems

Categorization • Classical categorization • Necessary and sufficient conditions for membership • Generic-to-specific monohierarchical structure • Modern categorization • Characteristic features (family resemblances) • Centrality/typicality (prototypes) • Basic-level categories

Properties of Categorization • Family Resemblance • Members of a category may be related to one another without all members having any property in common • Prototypes • Some members of a category may be “better examples” than others, i.e., “prototypical” members

Basic-Level Categorization • Perception • Overall perceived shape • Single mental image • Fast identification • Function • General motor program • Communication • Shortest, most commonly used and contextually neutral words • First learned by children • Knowledge Organization • Most attributes of category members stored at this level • Tends to be in the “middle” of a classification hierarchy

Information Hierarchy Wisdom Knowledge Information Data

Today’s Thinkers/Tinkerers

The Birth of AI • Rockefeller-sponsored Institute at Dartmouth College, Summer 1956 • John McCarthy, Dartmouth (->MIT->Stanford) • Marvin Minsky, MIT (geometry) • Herbert Simon, CMU (logic) • Allen Newell, CMU (logic) • Arthur Samuel, IBM (checkers) • Alex Bernstein, IBM (chess) • Nathan Rochester, IBM (neural networks) • Etc.

Definition of AI “... artificial intelligence [AI] is the science of making machines do things that would require intelligence if done by [humans]” (Minsky, 1963)

The Goals of AI Are Not New • Ancient Greece • Daedalus’ automata • Judaism’s myth of the Golem • 18th century automata • Singing, dancing, playing chess? • Mechanical metaphors for mind • Clock • Telegraph/telephone network • Computer

Some Areas of AI • Knowledge representation • Programming languages • Natural language understanding • Speech understanding • Vision • Robotics • Planning • Machine learning • Expert systems • Qualitative simulation

AI or IA? • Artificial Intelligence (AI) • Make machines as smart as (or smarter than) people • Intelligence Amplification (IA) • Use machines to make people smarter

Furnas: The Vocabulary Problem • People use different words to describe the same things • “If one person assigns the name of an item, other untutored people will fail to access it on 80 to 90 percent of their attempts.” • “Simply stated, the data tell us there is no one good access term for most objects.”

The Vocabulary Problem • How is it that we come to understand each other? • Shared context • Dialogue • How can machines come to understand what we say? • Shared context? • Dialogue?

Vocabulary Problem Solutions? • Furnas et al. • Make the user memorize precise system meanings • Have the user and system interact to identify the precise referent • Provide infinite aliases to objects • Minsky and Lenat • Give the system “commonsense” so it can understand what the user’s words can mean

Lenat on the Vocabulary Problem • “The important point is that users will be able to find information without having to be familiar with the precise way the information is stored, either through field names or by knowing which databases exist, and can be tapped.”

Minsky on the Vocabulary Problem • “To make our computers easier to use, we must make them more sensitive to our needs. That is, make them understand what we mean when we try to tell them what we want. […] If we want our computers to understand us, we’ll need to equip them with adequate knowledge.”

Commonsense • Commonsense is background knowledge that enables us to understand, act, and communicate • Things that most children know • Minsky on commonsense: • “Much of our commonsense knowledge information has never been recorded at all because it has always seemed so obvious we never thought of describing it.”

Commonsense Example • “I want to get inexpensive dog food.” • The food is not made out of dogs. • The food is not for me to eat. • Dogs cannot buy their own food. • I am not asking to be given dog food. • I am not saying that I want to understand why some dog food is inexpensive. • The dog food is not more than $5 per can.

Engineering Commonsense • Use multiple ways to represent knowledge • Acquire huge amounts of that knowledge • Find commonsense ways to reason with it (“knowledge about how to think”)

Multiple Representations • Minksy • “I think this is what brains do instead: Find several ways to represent each problem and to represent the required knowledge. Then when one method fails to solve a problem, you can quickly switch to another description.” • Furnas • “But regardless of the number of commands or objects in a system and whatever the choice of their ‘official’ names, the designer must make many, many alternative verbal access routes to each.”

CYC • Decades long effort to build a commonsense knowledge-base • Storied past • 100,000 basic concepts • 1,000,000 assertions about the world • The validity of Cyc’s assertions are context-dependent (default reasoning)

Cyc Examples • Cyc can find the match between a user's query for "pictures of strong, adventurous people" and an image whose caption reads simply "a man climbing a cliff" • Cyc can notice if an annual salary and an hourly salary are inadvertently being added together in a spreadsheet • Cyc can combine information from multiple databases to guess which physicians in practice together had been classmates in medical school • When someone searches for "Bolivia" on the Web, Cyc knows not to offer a follow-up question like "Where can I get free Bolivia online?"

Cyc Applications • Applications currently available or in development • Integration of Heterogeneous Databases • Knowledge-Enhanced Retrieval of Captioned Information • Guided Integration of Structured Terminology (GIST) • Distributed AI • WWW Information Retrieval • Potential applications • Online brokering of goods and services • "Smart" interfaces • Intelligent character simulation for games • Enhanced virtual reality • Improved machine translation • Improved speech recognition • Sophisticated user modeling • Semantic data mining

Fundamentals Top Level Time and Dates Types of Predicates Spatial Relations Quantities Mathematics Contexts Groups "Doing" Transformations Changes Of State Transfer Of Possession Movement Parts of Objects Composition of Substances Agents Organizations Actors Roles Professions Emotion Propositional Attitudes Social Biology Chemistry Physiology General Medicine Cyc’s Top-Level Ontology • Materials • Waves • Devices • Construction • Financial • Food • Clothing • Weather • Geography • Transportation • Information • Perception • Agreements • Linguistic Terms • Documentation http://www.cyc.com/cyc-2-1/toc.html

OpenCYC • Cyc’s knowledge-base is now coming online • http://www.opencyc.org/ • How could Cyc’s knowledge-base affect the design of information organization and retrieval systems?

Discussion Questions (Furnas) • Alison Billings & Vijay Viswanathan on Furnas • Are unlimited alias indexes an effective design solution to the problem of precision in "term based" searches? Is it possible to implement such a system that could maintain an accurate relation (category) to the designer’s “armchair” term with the existence of polysemy? Would the adaptive nature of this solution propagate an all inclusive alias category which could include all accessible information in a particular index?

Discussion Questions (Furnas) • Alison Billings & Vijay Viswanathan on Furnas • Since the publishing of this article in 1987 the technological advances in information retrieval in the past 16 years have been profound. Is the Vocabulary-Problem still a major issue in Human-System Communication? Furnas, et al., provide some solutions to the Vocabulary Problem such as “unlimited aliasing”, “keyword harvesting”, and “adaptive indices.” But now there are WYSIWYG interfaces such as Windows that may reduce the need for command line word choices, search engines that harvest the content from web pages, or services like Google that put out “Did you mean xxxxx?” when search results are sparse. Has the Vocabulary Problem been solved?

Discussion Questions (Minsky) • Joseph Hall on Minsky • Minsky talks a lot about commonsense. How would you define what is within the commonsense? Do you think that commonsense would be easy or difficult to teach to a computer? Why? Is commonsense a cross-cultural, basic-level category in the sense of what Lakoff described? Or is it more culturally specific (like "Don't step in front of moving traffic.") and thus harder to define? How would culturally-dependent definitions of "commonsense" complicate Minsky's theory? • Are machines that learn such a good thing? For example, I would like my computer to learn certain things (like how to fix common errors) but not others (like how to play the stock market with my bank account). Are ethics (cyber and otherwise) to be programmed into learning computers?

Discussion Questions (Minsky) • Joseph Hall on Minsky • What Minsky describes is all fine and dandy... but there seems to be a rather large gap between the machines of today and the machines he is postulating. To learn, machines would not only have to be able to note (and take action) when they are deviating from "operational parameter space" (malfunctioning, blue screen of death, etc.) but be able to decide on and implement a solution to the problem at hand from a different direction and/or using a different technique, quickly.

Discussion Questions (Minsky) • Joseph Hall on Minsky • Do you think that building such a commonsense-aware machine is possible today? (That is, is Minsky's model of a commonsense-based machine a reasonable *goal* or just an ideal?) If not, what are some of the impediments to the realization of one of Minsky's machines? • Do user expectations (reasonable or not) of what a computer should be doing factor into this at all?

Discussion Questions (Lenat) • Rebecca Shapley on Lenat • What does this article imply for best-practices in information organization & retrieval? How would you articulate the potential for a commonsense knowledgebase to revolutionize information retrieval? Does the premise of a commonsense-base feeding efforts at machine learning or natural language understanding make sense to you? Which potential applications Lenat mentions are compelling to you? • This article is from 1995 - do we hear anything more about this CYC? Did it revolutionize things? Why does Minsky call for a huge commonsense knowledgebase in 2000 when CYC was nearly complete in 1995?

Discussion Questions (Lenat) • Rebecca Shapley on Lenat • How would you apply the conduit metaphor & toolmaker's paradigms to describe, or perhaps critique, the CYC project? • If CYC is 'automating the whitespace in documents' - capturing the context for information, how would you describe the context it is capturing? How would you describe where the captured context is no longer applicable? How do you feel about the notion that 10+ people in Palo Alto CA were able to describe your context? Do you trust them with that task? Do you consider it necessary that some shared automated context be created? What challenges do you see for their ostensible goal, or limitations do you see to their approach?

Discussion Questions (Lenat) • Rebecca Shapley on Lenat • Anything in particular you can imagine yourself unwilling to have represented a particular way in the commonsensebase? Let's say you believe in reincarnation but the assertions in the commonsensebase don't leave any room for this idea, and how to interpret what you might say to a bereaved friend. How do you feel about the ability to 'automatically' interpret your expression being left out? Does it make you feel invisible, relieved, angry? What would be necessary to have it be culturally sensitive, and would that be encodable?

Discussion Questions (Lenat) • Rebecca Shapley on Lenat • What can you piece together about how CYC is implemented, how it makes decisions? What questions do you still have about how it works? • Do you think the tone of the article was influenced by the fact that Lenat was writing as President of Cycorp? • So, can this common-sense-base 'think'? Is it intelligent? Why and why not?

Assignment 0 Check-In • Deliverables • Personal web page • Assignments page • Email address • Focus statement • Online Questionnaire

Phone Project Overview • In this project we will be creating, sharing, and reusing mobile media and metadata • You and your Project Group will design application use scenarios and develop and refine metadata frameworks for your photos • Some of you may even choose to develop retrieval applications for the photo database in the second half of the course • We will be using the Nokia 3650 mobile media phone and software developed by Garage Cinema Research

Phone Project Overview • In the SIMS 202 Phone Project you and your Project Group will • Experience the actual process of information organization and retrieval (especially as regards metadata creation and use) • Work in small, focused teams performing a variety of tasks in image acquisition, description, and application design • Develop an ongoing resource for SIMS (an annotated photo database) that can be used for internal research and teaching, as well as for external promotional and informational purposes

Phone Project Requirements • Create engaging and useful application scenarios and photos for use by your team and the entire class • The photos you take and the applications you will design to use them should be interesting and useful to you and your colleagues • Create a shared, reusable resource of annotated photos • Design your metadata such that all photos are accessible not only for the needs of your particular application, but also for the reusability of your photos and metadata by other applications

Phone Project Assignments • Photo Use Scenario – Application Idea (Assignment 2) • You will brainstorm and storyboard an application for a mobile media device that accesses a server and facilitates the creation, sharing, and reuse of media and metadata. You will develop user personas and scenarios of how the application works and how the user experiences it. • Photo Capture and Annotation (Assignment 3) • With the goals of your application and the overall goals of the class project in mind, each group member is required to take at least 5 pictures relevant to the scenario you specified in the prior assignment. You will also get hands-on experience in annotating photos using the Mobile Media Metadata (MMM) framework, an application available on the mobile phones. You will also identify strengths and weaknesses of MMM framework.

Phone Project Assignments • Photo Metadata Design (Assignment 4) • Having your application and the overall project goals in mind, you will design a suitable metadata framework to annotate the photos in the collection. You will also annotate more photos using your metadata framework.

Lecture 04: Knowledge Representation