◆  Industrial Talk 1, by Junlan Feng

Topic: Beyond Open Source: Jiutian BiSheng Platform

Date and Time: Oct. 16, 2021, 10:00­-10:15

Meeting Room: Haitian Grand Ballroom 1, 1st floor

Short Bio: Junlan Feng Vice Chairman of the China Artificial Intelligence Industry Alliance, Chief Scientist at China Mobile，General Manager of AI and Intelligent Operation R&D Center，Board Chair of Linux Foundation Network.Dr. Feng received her Ph.D. on Speech Recognition from Chinese Academy of Sciences, and joined AT&T Labs Research in 2001, as a principal researcher on Speech recognition, language understanding and data mining until 2013. Dr. Feng has led R&D of China Mobile on artificial intelligence and intelligent operation since September 2013. She is an IEEE senior member. She had served as an IEEE speech and language committee member and an IEEE industry committee member. She is a frequent reviewer and organizer for major data mining, speech, and natural language international conferences and journals. Dr. Feng has over 100 professional publications and co-authored a book. She holds 51 issued U.S and international patents and 3 issued Chinese patents. She has 46 pending patent applications.Dr. Feng has won more than 20 domestic and foreign R&D awards, including AT&T CTO Award in 2009. “JiuTian”, the AI platform developed by her team won the single product gold award of China Mobile in 2019, the second prize of scientific and technological achievements of China Institute of Communications, the third prize of scientific and technological progress of China Institute of Electronics, and the 2018 Innovation Project Award of Deep Integration of Artificial Intelligence and Real Economy by the Ministry of Industry and Information Technology.

Abstract: Open-source software(OSS) has made fundamental contribution to nearly every issue in AI, though it is largely absent from open discussions. There are thousands of implementations of machine learning and deep learning algorithms/frameworks . Examples are Scikit-learn, R , Tensorflow, PyTorch, PaddlePaddle, Spark, Hadoop, etc. They are not simply some of the tools, they are the best AI/BigData tools. What is possible beyond Open Source? We believe Open Platform, Open Development, Open Education are part of the answer. Jiutian BiSHeng Platform(JBP) is designed for students, researchers and teachers. It provides workable instances for popular open source projects, particularly AI projects and saves developers’ time from setting it up, sorting out mismatches between OS versions and open source projects, fixing known and unknown bugs commonly seen in open source software. It provides tools and computing resources to support teachers to teach in a digital coding room, where the teacher and the students in the same course can share source codes and co-development environment. It provides students an open platform to conveniently learn and research. This talk will explain how JBP works.

◆  Industrial Talk 2, by Min Chu

Topic: Scenario Driven Innovation-Modeling and Improving the Sales Process by Discourse Analysis and Training-bot

Date and Time: Oct. 16, 2021, 11:45­-12:00

Meeting Room: Haitian Grand Ballroom 1, 1st floor

Short Bio: Dr. Chu joined AISpeech as a vice present in 2017. She built up AIspeech Beijing R&D center from scratch. The R&D center focused on developing key technologies in knowledge-driven spoken dialogue systems and exploring new business and application opportunities, and later incubated the Business Unit for Intelligent Services. Before joining AISpeech, Dr. Chu spent 8 years with Alibaba, leading various R&D efforts in big data, search and speech technologies and products. She initiated the iDST speech interaction team which supports the speech needs within Alibaba and Alipay (Yun OS, Alipay, Taobao, DingDing etc). Before Alibaba, Dr. Chu worked with MSRA for about 10 years. Her main research interests are in ASR, TTS, NLP, machine learning, big data etc. She has published 100+ academic papers and applied 50+ patents.

Abstract: Businesses that sale high value products such as cars, apartments or houses often spend a lot in advertising through various channels and the cost for win a customer keeps increasing. However, the sales management team normally has limited method to monitor the in-house sales process, which is the last and most important stage for closing the deal. They don't know whether the salesperson is friendly and professional enough, or whether they have overcommitments. This presentation introduces our innovative solution for such scenarios, which digitalizes the sales process, constructs a three-tier sales model, discovers pieces of outstanding examples, builds up training bots with these examples and setups evaluation matrix for the sales process. Many unseen challenges emerge when we go deeper in the frontier, some are solved or partially solved, more are waiting for deep exploring.

◆  Industrial Talk 3, by Songfang Huang

Topic: AliceMind – Technologies, Platform, and Applications

Date and Time: Oct. 17, 2021, 10:00-10:15

Meeting Room: Online

Short Bio: Dr. Songfang Huang is senior staff algorithm engineer of Language Technologies Lab, Alibaba DAMO Academy. He leads R&D on large scale pre-trained language models and its applications. His team develops AliceMind, Alibaba’s Collection of Encoder-decoders from Machine Intelligence of DAMO, which has been widely used internally and externally. Prior to Alibaba, he was affiliated with IBM Research (Watson and China) working on speech and language processing. He obtained his PhD from The University of Edinburgh.

Abstract: In this talk, we present AliceMind, Alibaba’s efforts on large scale pre-trained language models. We will introduce our technology innovations on generative, multilingual, and multimodal pre-trained language models. In particular, technologies on how to build and serve extra-large pre-trained language models will be mentioned. We will also talk about AliceMind Platform, and illustrate several business application scenarios of AliceMind in digital economy.

◆  Industrial Talk 4, by Yang Wu

Topic: Applications of Vision and/or Language Technologies in Video Content Production

Date and Time: Oct. 17, 2021, 11:45-12:00

Meeting Room: Haitian Grand Ballroom 1, 1st floor

Short Bio: Yang Wu received a BS degree and a Ph.D degree from Xi\'an Jiaotong University in 2004 and 2010, respectively. He is currently an expert researcher with Tencent ARC Lab. From Jul. 2019 to May 2021, he was a program-specific senior lecturer with Kyoto University. He was an assistant professor of the NAIST International Collaborative Laboratory for Robotics Vision from Dec.2014 to Jun. 2019. Before that, he was a program specific researcher with Kyoto University. His research is in the fields of computer vision, multimedia, and machine learning, with special interests in real applications. He has published more than 80 quality papers, including a AutoML best paper, two best student papers (BMVC and ICPR), works of three championships of global open challenges, and many oral presentations at major international conferences (ICCV, ECCV, BMVC, etc.).

Abstract: Video is becoming the major media in content production, distribution, and consumption in our daily lives. It has the advantanges of being able to include visual, audio, and textual information in the same media, in line with our natural experiences. While video distribution and consumption are growing explosively, video production, especially high-quality content generation, still faces great challenges due to the requirement of complex or even professional skills. This talk shares the latest progresses in Tencent ARC lab on building the fundamental components for video content production, especially those using cutting-edge vision and/or language technologies.