ConceptNet: A Wonderful Semantic World By Bijoy Arif The Development of the Space-Time View of Quantum Electrodynamics “We have a habit in writing articles published in scientific journals to make the work as finished as possible, to cover all the tracks, to not worry about the blind alleys or to describe how you had the wrong idea first, and so on. So there isn't any place to publish, ...” Richard P. Feynman, Noble Lecture Presentation Plan Part 1 1. Introduction 2. Background Knowledge 3. ConceptNet and Its Counterparts Part 2 4. Building of ConceptNet 5. Structure of ConceptNet 6. Applications of ConceptNet 7. Present ConceptNet Presentation Plan (cont...) Part 3 8. ConceptNet in Windows 9. ConceptNet Modules 10. Demo of ConceptNet 11. Quick Review 12. My View of ConceptNet Part 1 Introduction: What is ConceptNet? > ConceptNet is a semantic network to give common sense knowledge(concept) to machine. > Simply means a tool so that computers can understand daily usage English. Features > Python based SQL toolkit > Maintain Semantic Network > Acquire data from Open Mind Common Sense Corpus > Till now Open Source Origin and Creators > Originated in MIT Media Lab > First Appeared as The ConceptNet Project v2.1 >Introduced by Hugo Liu Push Singh Ian Eslick Background Knowledge: What is Semantic Network? > A network represents semantic relation between concepts. > Semantics is the meaning of something focuses on relation between signifiers like words, phrases, signs or symbols. > Here Concepts means some abstract objects. > In Computer Science terminology, It is a directed or undirected graph consisting of vertices, which represent concepts, and edges. What is Open Mind Common Sense Corpus? > Open Mind Common Sense (OMCS) is an artificial intelligence project based at the Massachusetts Institute of Technology (MIT) Media Lab whose goal is to build and utilize a large commonsense knowledge base from the contributions of many thousands of people across the Web. > Unlike common corpus like British National Corpus, International Corpus of English Relation between OMCS and ConcepNet > The project is brainchild of Marvin Minsky, Push Singh, Catherine Havasi and others. Eventually they contributed to ConceptNet > ConceptNet is a semantic network based on the information in the OMCS database. Simply saying, OMCS is the core of ConceptNet. ConceptNet and Its counterparts > Two other popular Natural Language Processing toolkit like ConceptNet are: WordNet Cyc > ConceptNet project is inspired by these two projects. WordNet and Cyc > WordNet is large lexical database, initiated in Princeton University in mid 1980s by George A. Miller, to provide meaning and relation of English words. > Cyc is started by Cycorp Company in 1984 to create common sense knowledge in a formalized logical framework. Similarity and Difference > ConceptNet is the combination of WordNet like structure and Cyc like relation. > Extended WordNet's notion of node and repertoire in semantic network. > WordNet and Cyc are handcrafted by knowledge engineers but ConceptNet is OMCS corpus based, not manually handcrafting commonsense knowledge. Similarity and Difference (cont...) > WordNet has a lexical emphasis and employs a formal taxonomic approach. > Cyc represents commonsense in a formalized logical framework means it excels in careful deductive reasoning. > ConceptNet represents contextual common sense reasoning over real world texts. Part 2 Building of ConceptNet ConceptNet's extraction rules from semistructured OMCS: > Extraction Phase > Normalization Phase > Relaxation Phase Building of ConceptNet (cont...) > approximately fifty extraction rules are used to map OMCS's English sentences into ConceptNet binary relation assertion. > Extracted Nodes are also normalized. > Relaxation means to smooth over semantic gaps and improve the connectivity of network. Structure of ConceptNet > K Lines (1.25 million assertions) > Things (52,000 assertions) > Agents (104,000 assertions) > Events (38,000 assertions) > Spatial (36,000 assertions) > Causal (17,000 assertions) > Functional (115,000 assertions) > Affective (34,000 assertions) Structure of ConceptNet (cont...) Overall semantic network contains: > 1.6 millions assertions > over 300,000 nodes Applications of ConceptNet > Commonsense ARIA > Goose > MakeBelieve > GloBuddy > AAA- a profiling and recommendation system and many more Present ConceptNet > Originally initiated as ConceptNet 2 > It is no longer maintained > Then ConceptNet 3 was introduced > Now ConceptNet 5 is available > Developed by Rob Speer Catherine Havasi and Many others Part 3 ConceptNet in Windows > Using ConceptNet in Linux or Mac is very easy > But in Windows, need bag of tricks > Need a way to use others SQL database in Python ConceptNet in Windows (cont...) Need to download > Any Python Machine > ConceptNet.tar.gz > csc-util.tar.gz > Django.tar.gz > Simplenlp.tar.gz ConceptNet Modules Http://csc.media.mit.edu/docs/conceptnet/concept net4.html ConceptNet Demo >>> It is time to visit wonderful world Quick Review > Initiated as ConceptNet 2 > MIT Media Lab is Place of Birth > Use OMCS Corpus to create, maintain and develop ConceptNet > Maintain a large database brilliantly > Combination of WordNet like structure and Cyc like relation > Now ConceptNet 5 is available My View of ConceptNet > OMCS should be open source as well. > Must have a way to interact with OMCS to change, develop, acquire data. > ConceptNet have a way to update its database directly from interactive OMCS. > Overall it is a nice world. Thank You References [1] ConceptNet-a practical commonsense reasoning toolkit by H Liu and P Singh [2]Http://csc.media.mit.edu/docs/conceptnet/conc eptnet4.html Questions????