Never-ending learning

• DOI: 10.1145/3191513
• Publication Date: 2018-04-25T12:22:17Z
• Authors: T. Mitchell; W. Cohen; E. Hruschka; P. Talukdar; B. Yang; J. Betteridge; A. Carlson; B. Dalvi; M. Gardner; B. Kisiel; J. Krishnamurthy; N. Lao; K. Mazaitis; T. Mohamed; N. Nakashole; E. Platanios; A. Ritter; M. Samadi; B. Settles; R. Wang; D. Wijaya; A. Gupta; X. Chen; A. Saparov; M. Greaves; J. Welling
• Keywords: 0202 electrical engineering, electronic engineering, information engineering; 006; 02 engineering and technology; Computer Science & Automation; 004

Whereas people learn many different types of knowledge from diverse experiences over years, and become better learners time, most current machine learning systems are much more narrow, just a single function or data model based on statistical analysis set. We suggest that than computers precisely because this difference, we key direction for research is to develop software architectures enable intelligent agents also knowledge, continuously time. In paper define never-ending paradigm learning, present one case study: the Never-Ending Language Learner (NELL), which achieves number desired properties learner. NELL has been read Web 24hrs/day since January 2010, so far acquired base with 120mn diverse, confidence-weighted beliefs (e.g., servedWith(tea,biscuits) ), while thousands interrelated functions continually improve its reading competence learned reason infer new it not yet those has, inventing relational predicates extend ontology uses represent beliefs. describe design NELL, experimental results illustrating behavior, discuss both successes shortcomings as study in learning. can be tracked online at http://rtw.ml.cmu.edu, followed Twitter @CMUNELL.