Artificial Intelligence, Machine Translation and
Language Processing Symposium 2019

KEYNOTE
Accelerate and Scale AI Adoption with Trust and Transparency across your Enterprise

Every business and enterprise is embracing AI. As shown in IBM’s new 2018 AI study – AI is moving beyond the hype cycle, with 82% of companies now considering AI adoption. There is wide consensus that AI is key to competitive advantage. Yet the challenge businesses face today is how they scale AI across the organization while making sure they can explain and trust the decisions and outcomes the technology is making.

Join us in this session to learn about IBM’s newly announced Trust and Transparency capabilities for AI and how they can help you solve AI’s “black box” problem by providing visibility and insights into how and why AI systems are making the decisions.

Research in Translation – What Is Exciting and Shows Promise Ahead?

The recent trend of using deep learning to solve a wide variety of problems in Artificial Intelligence has also reached machine translation – thus establishing a new state-of-the-art approach for this application. This approach is not yet settled by any means. New neural architectures are proposed and ideas coming from such diverse fields as computer vision, game playing, and speech recognition can be applied to machine translation as well.

At the practical end, we are just learning about the deployment challenges of this technology, since old methods, for example, to integrate terminology databases or domain adaptation no longer apply.

This presentation will give an overview of the latest developments in research and what this means for practical deployment.

Moving Pictures – Sub Caption: Neural MT, MAM & Cloud Orchestration for Broadcast & OTT

This talk will take you through the journey of Neural Machine Translation applied in the rapidly changing environment of subtitles for a high-capacity multi-language broadcast television company as it adapts to the post-OTT bubble. As smart broadcasters channel their bouquet from the TV screen to any screen, behind the scenes a revolution of MAM content processing automation, cloud orchestration, and data enrichment moves a once monolithic production chain into a hybrid latticework of on-premise and cloud services.

With the emergence of new complexity in the delivery chain to provide subtitles for connected set-top boxes that behave like mobile devices comes simplicity for viewers and new life for viewer engagement vital to a shrinking medium. Connected TVs and apps share a common backplane with OTT so the automation that drives content also connects data and subtitles are a rich source of meaningful data for search and recommendation. Boosting subtitle production with MT and managing the data it contains centrally as a service means unlocking its value becomes an integral part of the larger data lakes that sustain a new era in TV and OTT hybrid services.

How Active Learning is Changing AI and NLP

Machine-learned models depend on having human-annotated training data, a tremendously time-consuming and expensive undertaking. Just annotating 500,000 Korean tokens for named entity recognition takes about eight person months of annotator time (split among four annotators and a project manager).

Basis Technology is developing an active learning annotator that will significantly reduce the labor of this critical process, by both increasing the speed of annotation and reducing the number of tagged documents needed to reach required accuracy.

MAVERICK PRESENTATION
AI and Trust in a World Without Secrets

This presentation aims to help you decide how and when to abandon conventional thinking in pursuit of understanding innovation in language technologies and analytics. It is intentionally provocative, and the results could prove disruptive. For those new to Maverick discussions, these kinds of discussion deliberately expose unconventional thinking and may not always agree with mainstream opinion, but they open up new ideas for thought and discussion.

The world is in the early stages of being transformed as the use of Artificial Intelligence (AI) permeates all human endeavour. Decisions made today will determine the impact AI will have an the human race. Some of the smartest people on the planet are concerned that without the right governance, AI has the potential to be disastrous for mankind. The late Stephen Hawking declared that AI could be the “worst event in the history of our civilization” and Elon Musk has stated that “AI is a fundamental risk to the existence of human civilization”.Will AI live up to the dreams of utopian hype, or fall into the depths of dystopian anxiety? The fate of the human race is inexorably connected to the diffusion, use and acceptance of AI across all parts of society. Issues of trust and ethics are critical in shaping the AI-driven world we are hurtling toward. What are those ethical dilemmas? What can we do today to move AI Governance in the right direction for our future?

Big Data, Small World: Broader Web-Scale Provision of Parallel Corpora for European Languages

ParaCrawl is a large-scale effort to collect translation data for many languages from the web. The project received seed funding from Google and is currently funded by the European Union through the Connected Europe Facility to create the largest known parallel corpora for the official languages of the European Union, and other European languages of interest.

The project already made parallel corpora with billions of words per language pair available. ParaCrawl also develops novel methods for document alignment, sentence alignment, and corpus cleaning and releases its processing pipeline as open source tools. The project is a joint initiative of the University of Edinburgh, Johns Hopkins University, the University of Alicante, Prompsit, TAUS, and Omniscien Technologies.

An overview of ParaCrawl, it’s history, technologies, processes and resulting data will be presented followed by a panel discussion with several of the ParaCrawl consortium members.

Text.ai – Machine Learning Powered Intelligent Document Processing

A typical back office function is any financial services enterprise usually faces challenges while processing volumes of document such as invoices, letter of credit, payments etc. There may be issues regarding noisy data, inconsistent document quality, highly contextual and domain sensitive information, and ambiguous words. This leads to costly and time consuming manual process of data extraction, validation and verification. There is a distinct need to automate the document classification and processing flow in order to expedite processing time, reduce the cost of human labour as well as errors, and improve productivity of human operators by augmented solution.

Persistent Systems will explore Text.ai, a solution powered by Natural Language Processing or NLP driven entity extraction and document classification. Text.ai is instrumental to alleviate above-mentioned pain points. It provides 3 extraction models to increase the accuracy on entity extraction from documents: a rule based model, ML model and domain dictionary model. These underlying models are ensembled to increase the accuracy and coverage of extraction, to reduce the need for a fixed template documents and to provide the flexibility in document processing much like human operator process it. Text.ai has a pluggable framework to work with variety of open souce as well as commercial OCR engines. It also provides REST APIs for integration with downstream BPM or RPA processes.

EXPERT PANEL
ASK US ANYTHING

Through the magic of modern technology, all our speakers are assembled across multiple time zones from Australia to Asia, Europe and USA to answer any question you have from the conference presentations and topics.

Our expert panel will answer questions submitted by attendees. Ask us anything related to AI, MT and Language Processing. How do these technologies and tools impact your industry or business? How can you best take advantage? What are the pitfalls? What are the best practices?

ASK US ANYTHING To have your question addressed by the expert panel, fill in the ASK US ANYTHING question form.