ABSTRACT We address the problem of disambiguating large scale catalogs through the definition of an unknown artist clustering task. We explore the use of metric learning techniques to learn artist embeddings directly from audio, and using a dedicated homonym artists dataset, we compare our method with a recent approach that learn similar embeddings using artist classifiers. While both systems have the ability to disambiguate unknown artists relying exclusively on audio, we show that our system is more suitable in the case when enough audio data is available for each artist in the train dataset. We also propose a new negative sampling method for metric learning that takes advantage of side information such as music genre during the learning phase and shows promising results for the artist clustering task. 1. INTRODUCTION 1.1 Motivation With contemporary online music catalogs typically proposing dozens of millions of recordings, a major problem is the lack of an universal and reliable mean to identify music artists. Contrarily to albums' and tracks' ISRC As a direct consequence, the name of an artist remains its defacto identifier in practice although it results in common ambiguity issues. For example, name artist collisions (e.g. Bill Evans is the name of a jazz pianist but also the name of a jazz saxophonist and the name of a blackgrass banjo player) or artist aliases (e.g. Youssou N'Dour vs. Youssou Ndour, Simon & Garfunkel vs Paul Simon and Art Garfunkel, Cat Stevens vs Yusuf Islam) are usual.

Many types of data can be represented as graphs. Edges may correspond to actual links in the data (e.g., users connected by some social network) or to levels of similarity induced from the data (e.g., users having liked a large common set of movies). The resulting graph is typically sparse in the sense that the number of edges is much lower than the total number of node pairs, which makes the data hard to exploit. A standard approach to the analysis of sparse graphs consists in embedding the graph in some vectorial space of low dimension, typically much smaller than the number of nodes [15, 19, 4]. Each node is represented by some vector in the embedding space so that close nodes in the graph (linked either directly or through many short paths in the graph) tend to be represented by close vectors in terms of the Euclidian distance.

ABSTRACT We explore active learning (AL) utterance selection for improving the accuracy of new underrepresented domains in a natural language understanding (NLU) system. Moreover, we propose an AL algorithm called Majority-CRF that uses an ensemble of classification and sequence labeling models to guide utterance selection for annotation. Experiments with three domains show that Majority-CRF achieves 6.6%-9% relative error rate reduction compared to random sampling with the same annotation budget, and statistically significant improvements compared to other AL approaches. Additionally, case studies with human-in-the-loop AL on six new domains show 4.6%-9% improvement on an existing NLU system. Index Terms-- Active Learning, Domain Expansion, Natural Language Understanding, Intelligent Virtual Assistants 1. INTRODUCTION Intelligent virtual assistants (IVA) with natural language understanding (NLU), such as Amazon Alexa, Apple Siri, Google Assistant, and Microsoft Cortana, are becoming increasingly popular. For IVA, NLU is a distinct component of spoken language understanding (SLU) [1], in conjunction with automatic speech recognition (ASR) and dialog management (DM). ASR produces a token sequence from speech, which is passed to NLU for both classifying the action or "intent" that the user wants to invoke (e.g.

Holden Foreman '21 is a Managing Editor for the News section. He hails from the city of St. Louis, Missouri and is studying electrical engineering, computer science and economics.

A major new Harry Potter video game may be in development, after leaked footage appeared online. The 72 seconds of footage shows players casting spells, cavorting with magical beasts, fighting goblins and exploring what appears to be Hogwarts. Gaming journalists took to social media to speculate about the authenticity of the leak, with some suggesting that a big-budget Harry Potter game is long overdue. The leaked video was posted to YouTube by someone using the name RastaPasta, together with a description of the game. "Set in the 19th Century (1800's) Wizarding World, this 3rd person open-world action RPG game centers around your character with unique abilities who has eared a late acceptance to Hogwarts School of Witchcraft and Wizardry. You are a newly arrived 5th year student to Hogwarts that demonstrates a latent gift for magic with a unique ability to track and identify remnants of a pottant ancient power," it states.

Most of us can position ourselves somewhere on The AI Fear Continuum. For nearly 100 years, science fiction writers have prepared us for the worst as they explored scenarios from Hollywood 1927's Metropolis to adaptations of Isaac Asimov's I, Robot and Philip K Dick's Do Androids Dream Electric Sheep (famously the basis for Blade Runner). We're ready to be tricked (Alien), enslaved (The Matrix), killed (West World) or even have our whole species wiped out (Terminator). Conversely, news stories about AI feel overwhelmingly positive, informing us that AI is saving the day in sectors as diverse as health care, law and climate change. Marketing professionals have two key points to consider; claims that AI can help to make marketing more creative and effective, and claims that it will make many of our job roles surplus to requirements.

We may take the #hashtag for granted today, but it didn't emerge fully formed from Biz Stone's head. The Large Hadron Collider hasn't collapsed (or collapsed the space-time continuum), but that wasn't a given when scientists first turned the thing on. Nobody thought the sweaty geeks who sent their supposedly self-driving cars into concrete barriers instead of across the Mojave Desert would soon threaten to upend the way we move through the world. Oh, and remember the Microsoft trial? When we started planning our 25th birthday party more than a year ago, we knew that not all readers (okay, not even most readers) would have been following us since day one, and they certainly wouldn't recall every story we've told.

Asia-based accelerator program Zeroth is getting a major infusion of capital after digital media company Animoca Brands agreed to invest over $3 million into its businesses. Animoca is listed on the ASX with a market cap of around $40 million. It is best known for its range of mobile games which include the Doraemon and Garfield brands but it has been pushing to broaden its focus into artificial intelligence, blockchain and more. The relationship is not new. Animoca previously invested US$1 million (A$1.39 million) in Hong Kong-based Zeroth last December, and now it is following up to take a majority stake in Zeroth's operational business and also joining its fund as an LP. According to an announcement, Animoca is paying up to US$1.08 million (A$1.5 million) for a 67 percent share of Venture Classic Limited -- Zeroth's operational business -- in addition to a $2 million commitment to Zeroth's fund, which it will join as an LP.

CBS All Access is streaming the first of its Short Treks tidbits to hold Discovery fans over until season two, while Hulu premieres Into the Dark, Amazon Prime has a new season of The Man in the High Castle and Netflix leads off a slew of releases with Big Mouth season two and Dancing Queen. For gamers there's Assassin's Creed: Odyssey, Mega Man 11 and the wide release of Forza Horizon 4. Look after the break to check out each day's highlights, including trailers and let us know what you think (or what we missed). Richard's been tech-obsessed since first laying hands on an Atari joystick.

In this era of 24 hour news cycles where news breaks more often on social media than on the front page, newsrooms and journalists have been able to work more efficiently to stay on top of important stories by using Artificial Intelligence. AI has long been seen as away of relegating the more mundane, time-consuming tasks away from the man to the machine, allowing journalists more time to do the thing we rely on them for: reporting. But things are changing, and AI is developing with an aim of doing more to the newsroom than helping achieve maximum productivity. Some early adopters in the AI efficiency space were technologies like the BBC News' Juicer Project. Juicer launched in 2012 with the aim to "provide a source of tagged News content to power prototypes."