Jon Juett April 21, 2014
Selected very recent papers Includes some student level event / conference papers UM Health Counseling Program Correctly respond to emotion Follow tree-based conversation Detect deception, infer progress, etc.
First automated system by Green et al in 1961 Early systems domain specific Open domain problem initially too large Internet makes more data sets available and searches more common Database querying becomes faster Open domain QA research resumes
Computationally inexpensive Current bottleneck in initial search Can scale with dataset growth Especially Internet Fully automated Except if cooperative steps with user included Finds short, correct answers User-friendly
Like a Google Search Uses primitive techniques Bag of Words Easier task Returns entire likely relevant documents Leaves answer extraction step to user
In an open domain, cannot rely on domain specific knowledge General world knowledge only Conglomeration of components Many systems overlap, plug-in techniques
Cooperative Component Not restricted to literal answers Allows clarification of question for better answer Can build up information on user interests, system may even lead conversation to do this
Database approaches Semantic via Resource Description Framework (RDF) Triples ▪ (subject, verb, object) ▪ Answer true/false or factoid ▪ Efficiently increase answer database Template matching ▪ Use machine learning to cluster questions ▪ Templates can correct for typos and abbreviations ▪ Only need to store an answer to each cluster ▪ Novel questions matched to closest cluster
Question WordGoogle Answer RatioSemantic Web Answer Ratio What When Where Which Who How0.50 How old How long0.98 How many0.50 Najmi et al, 2013
Response to unnatural feel of non-affective dialogue systems Sense user emotional state Punctuation, word choice, capitalization, emoticons Simulate emotions Speech synthesis, emoticons, nonverbal behavior Attempt to influence user emotion Extend to poll online communities Composed of perception layer, control layer, and actuator-communicator layer
Complex automated evaluation beyond state of the art Multi-step reasoning Real-time question answering Language of user, query, data Answer extraction Customize to user preferences
Fader, Anthony, Zettlemoyer, Luke, Etzioni, Oren, "Paraphrase-Driven Language for Open Question Answering," University of Washington ACL Melo, Dora, Rodriues, Irene Pimenta, Nogueira, Vitor Beires, "A Review on Cooperative Question-Answering Systems," Najmi, E.; Hashmi, K.; Khazalah, F.; Malik, Z., "Intelligent Semantic Question Answering System," Cybernetics (CYBCONF), 2013 IEEE International Conference on, vol., no., pp.255,260, June 2013 Skowron, Marcin, "Affect Listeners: Acquisition of Affective States by Means of Conversational Systems," Development of Multimodal Interfaces: Active Listening and Synchrony, Springer Berlin Heidelberg, pp January Tyagi, Deepali, Joshi, Tejas, Ghule, Dhanashri, Ameya, Joshi, "An Interactive Answering System using Template Matching and SQL Mapping for Natural Language Processing," International Journal of Advance Research in Computer Science and Management Studies, vol. 2, issue 2, pp , February 2014