We will describe & give a live demo of a Conversational AI system built for the 2018 Amazon Alexa Prize which came third among 8 teams selected from >200 international entrants. The system consists of an ensemble of conversational bots, combining rule-based and machine learning systems.
------------------------------------------------------------------------------------
SPEAKERS
Arash Eshghi - Heriot-Watt University
------------------------------------------------------------------------------------
AGENDA
17:30 - Light Refreshments (Served in Room 9020a)
18:30 - Talk
19:30 - The event ends
------------------------------------------------------------------------------------
SYNOPSIS
Arash will describe & give a live demo of a Conversational AI system built for the 2018 Amazon Alexa Prize which came third among 8 teams selected from >200 international entrants. The system consists of an ensemble of conversational bots, combining rule-based and machine learning systems.
The various features of the system include:
(1) A Natural Language Understanding (NLU) pipeline;
(2) TYhe use of topic ontologies and Named Entity Linking to enable the user to navigate and search interactively through a web of related information; rendering Alana in part an interactive NL interface to linked information on the web;
(3) System generated clarification questions to interactively disambiguate between Named Entities as part of NLU;
(4) A new profanity & abuse detection model with rule-based mitigation strategies;
(5) Response retrieval from Reddit.
Arash will also present some of the ablation studies that measure the performance contributions of specific features (e.g. use of Ontology-bot, Reddit-bot, rule-based systems, etc). Arash and his team found that these features all increased overall system performance. Their final score, namely averaged user ratings over the whole semi-finals period, was 3.4 (out of 5). The team were also able to achieve long dialogues (average around 11 turns and 2.20 minutes) during the semi-finals period.
------------------------------------------------------------------------------------
SPEAKER BIOGRAPHY
Arash Eshghi - Heriot-Watt University
Arash Eshghi is a Lecturer at the Computer Science Department, Heriot-Watt University, and a member of the Interaction Lab. With over 8 years of postdoctoral experience, Arash’s research is at the interface of Computer Science, Cognitive Science & Computational Linguistics, with a focus on modelling the various intricate processes involved in everyday human interaction, and using such models to build better, more naturally interactive conversational agents. For this, he has combined formal grammars of dialogue with machine learning techniques for automatically bootstrapping conversational AI.
------------------------------------------------------------------------------------
For overseas delegates who wish to attend the event please note that BCS does not issue invitation letters.
------------------------------------------------------------------------------------
THIS EVENT IS BROUGHT TO YOU BY:
BCS South Yorkshire Branch
Visit www.bcs.org/category/18932
