Text Generation, Editing, and Summarization
AS6 Level 5
MR6, AS6-05-10
Abstract:
Text generation has been a longstanding NLP task. Before the era of deep learning, the generation methods are basically rule-based or template-based at the levels of sentence, paragraph, and document. These previous studies indeed resulted in some applicable systems for weather broadcast, sports news report, etc. Owing to the development of deep learning techniques, particularly sequence to sequence models, researchers have achieved tremendous progress on different text generation tasks in the last few years. In this talk, I will review some works that attract much attention in the research community, and introduce some of my recent works with more details. These works cover the text generation in four scenarios: (1) Abstractive text summarization, a task that neatly falls in the capability scope of the sequence to sequence models; (2) Question answering (QA), where a natural application of text generation is for answer production. Some recent works explore to generate questions for SQuAD-like tasks, or distractors for multi-choice questions in RACE, with the motivations of helping dataset preparation and generating large-scale pseudo training data; (3) Text style transfer, which meets some application cases where we need to edit an input sentence towards a given style (e.g. from informal to formal, from positive to negative); (4) Knowledge-based text generation, aiming at generating text that complies with the given knowledge, such as factual triples or common-sense knowledge. Finally, I will share with you my ideas about interesting and promising text generation directions.
Biodata:
Dr. Lidong Bing is a Research Scientist at R&D Center Singapore, Alibaba DAMO Academy, where he leads the NLP team. His research interests include Sentiment Analysis, Text Generation/Summarization, Information Extraction, Knowledge Base, Machine Translation, etc. Dr. Bing has published about 50 papers in the last few years at top-tier conferences and journals. Prior joining Alibaba, Dr. Bing was a Researcher at Tencent AI Lab. He received his Ph.D. degree from The Chinese University of Hong Kong and was a Postdoc Research Fellow at Carnegie Mellon University. For more information, refer to his homepage: http://www.cs.cmu.edu/~lbing/.
R&D Center Singapore is an international R&D team with a focus on developing cutting-edge speech and language processing technologies, including speech recognition, speech synthesis, natural language processing, and machine translation. To serve Alibaba's globalization strategy, the R&D Center Singapore is paying special attention to the technology developments related to the areas of multilingual speech and language processing.