1 / 25

ConceptNet

ConceptNet. Outline. What is commonsense? Representative research on commonsense Open Mind Common Sense (OMCS) ConceptNet LiftNet StoryNet Practice Summary. What is commonsense?.

edmund
Download Presentation

ConceptNet

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. ConceptNet

  2. Outline • What is commonsense? • Representative research on commonsense • Open Mind Common Sense (OMCS) • ConceptNet • LiftNet • StoryNet • Practice • Summary

  3. What is commonsense? • Beliefs or propositions that seem, to most people, to be prudent and of sound judgment, without dependence upon esoteric knowledge • Exhibiting native good judgment • Arrive home at a reasonable hour • Commonsense scholarship on the foibles of a genius • Unlearned and commonsensical countryfolk were capable of solving problems that beset the more sophisticated • Commonsense reasoning is the branch of Artificial intelligence concerned with replicating human thinking • Developing adequately broad and deep commonsense knowledge bases. • Developing reasoning methods that exhibit the features of human thinking, including: • The ability to reason with knowledge that is true by default • The ability to reason rapidly across a broad range of domains • The ability to tolerate uncertainty in your knowledge • Developing new kinds of cognitive architectures that support multiple reasoning methods and representations

  4. Projects to Collect Commonsense • Cyc • Started in 1984 by Dr. Doug Lenat • Developed by CyCorp, with 3.2 millions of assertions linking over 280.000 concepts and using thousands of micro-theories. • Cyc-NL is still a “potential application”, knowledge representation in frames is quite complicated and thus difficult to use. • Open Mind Common Sense Project • Started in 2000 at MIT by Push Singh • WWW collaboration with over 20,123 registered users, who contributed 812,769 items • Used to generate ConceptNet, very large semantic network. • Other such projects • HowNet (Chinese Academy of Science) • FrameNet (Berkley)

  5. Open Mind Common Sense (OMCS) • 750k NL assertions from 15k contributors (Initial stage) • ConceptNet • A semantic net built from these • 20 link types

  6. ConceptNet • Common sense knowledge base with NLP capability • Much needs to be examined • Uncontrolled vocabulary, can be biased in terms of content; but seems quite reliable knowledge • Extracted automatically from common sense knowledge expressed in semi-structured NL sentences from OMCSNet (open mind common sense) – applying about 50 extraction rules • ”The Effect of [falling off a bike] is [you get hurt].” • ”A lime is a very sour fruit” at OMCS is extracted into two assertions • IsA (lime, fruit) • PropertyOf (lime, very sour) • Commonsense knowledge covering aspects of everyday life • Spatial • Physical • Social • Temporal • psychological

  7. ConceptNet Example

  8. ConceptNet Function • Inference • Spreading activation: Node-activation radiating outward from an origin code • GetContext (node) • GetAnalogousConcept (node) • Graph traversal: • FindPathBetweenNodes (node1, node2) • Support • Topic sensing • Query expansion • Semantic similarity of words • Lexical generalization • Thematic generalization

  9. Snapshot of ConceptNet

  10. ConceptNet Performance

  11. ConceptNet Application • Commonsense ARIA • Observes a user writing an e-mail and proactively suggests photos relevant to the user’s story • Bridges semantic gaps between annotations and the user’s story • GOOSE • A goal-oriented search engine for novice users • Generate the search query • MAKEBELIEVE • Story-generator that allows a person to interactively invent a story with the system • Generate causal projection chains to create storylines • GloBuddy: A dynamic foreign language phrasebook • AAA: Rrecommends products from Amazon.com by using ConceptNet to reason about a person’s goals and desires,creating a profile of their predicted tastes.

  12. Reasoning in LifeNet • LifeNet : A large-scale temporal graphical model expressed in terms of egocentric propositions of the form • I am at a restaurant • I eat a sandwich • It is 3 pm • It is raining outside • I feel frightened • Temporal reasoning • Prediction : Guess what might be true in the next moment • Elaboration : Guess what else might be true now • Explanation : Guess at what happened prior to the current event • Projection : Guess what series of events might follow • Filtering : Filter unlikely current states or events • Fixed-lag smoothing : Filter unlikely past states or events

  13. StoryNet • StoryNet builds on LifeNet and ConceptNet • ConceptNet lays out the possibilities for ordering elements • I want to drive a car • I ned gasoline • Gasoline can be found in a plane • A plane can be found in the sky

  14. Collaboartion of Three Nets

  15. Demos • Video of Henry Lieberman's lecture on Applying Common Sense Reasoning in Interactive Applications: http://helix.media.mit.edu/ramgen/insite/exa/2003/lieber-2003-02-26.rm • http://web.media.mit.edu/~lieber/Lieberary/Mondrian/Knowacq.mov • A User Interface for Knowledge Acquisition from Video • http://agents.media.mit.edu/projects/voice/ • CS reasoning for better voice recognition • ConceptNet - to disambiguate phonetically similar words and improve overall recognition accuracy • http://web.media.mit.edu/~lieber/Lieberary/Lieberary.html

  16. Practice Installation • 파이썬 설치 (python-2.4.3.msi) • 컨셉넷 설치 (ConceptNet2.1.zip) • Montylingua: 문장 분석 모듈 • ConceptNet • 기본 실행 소스 (*.py) • ConceptNetGUI.py • ConceptNetXMLRPCServer.py • ConceptNetDB.py • ConceptNetNLTools.py • 기본 DB 소스 (*.txt) • predicates_concise_kline.txt • predicates_concise_nonkline.txt • predicates_nonconcise_kline.txt • predicates_nonconcise_nonkline.txt

  17. Practice ConceptNetGUI.py

  18. Practice ConceptNetXMLRPCServer.py import sys import ConceptNetDB import DocXMLRPCServer pred_filename = "predicates.txt" if len(sys.argv)>0 and sys.argv[-1][-1*len('.py'):].lower()!='.py': pred_filename = sys.argv[-1] print "Syntax: python ConceptNetXMLRPCServer.py [predicates_file]" print "Loading Predicates from %s..."%pred_filename c =ConceptNetDB.ConceptNetDB(None,pred_filename) print "Starting XML-RPC Server" port = 8000 xmlrpc = DocXMLRPCServer.DocXMLRPCServer(('',port)) print "Now serving on localhost port %s!"%str(port) xmlrpc.register_introspection_functions() xmlrpc.register_instance(c) xmlrpc.register_instance(c.nltools) xmlrpc.serve_forever() • ConceptNetDB.py (http://주소:port), ex: http://165.132.140.237:8000 • ConceptNetNLTools.py (http://주소/port), ex: http://165.132.140.237:8001

  19. Practice Wrapper Modules • Python: 기본 동작 시스템 • XML-RPC: 모듈간 통신 모듈 • C# • 소스 • CookComputing.XmlRpc.dll • ConceptNetEx2.exe • 기본 제공 함수 • guess_mood, guess_topic, guess_concept, summarize_document, tag, get_analogous_concepts, get_context, get_all_projections, project_affective, project_consequences, project_details, project_spatial • C++: Pipe 연결 (XML RPC는 Visual C++ 지원 안함) • 소스: concept_cpp_interface.h • 사용방법 • #include "concept_cpp_interface.h“ • ConceptNetCPPInterface ci; • ci.setCommend(type, query); • ci.executeCN(); • ci.mt, ci.tt, ci.ct, ci.at, ci.cxt, ci.pt, ci.aft, ci.cqt, ci.dt, ci.st • 기타방법: Python과 C++ 직접 연동해서 사용하기 (문의: 송인지)

  20. Practice Example: 장소 질의 확장

  21. Summary • Little new work on the practical commonsense reasoning • Building practical commonsense reasoning systems using unconventional techniques • Representing knowledge in natural languageDistributing knowledge acquisition to non-experts via the World Wide Web • Developing reasoning techniques that work successfully with large and imperfect knowledge bases. • Lots of possibility with commonsense reasoning

More Related