Asli Celikyilmaz: Learning Structure in Text Generation

Abstract: Automatic text generation enables computers to summarize text, describe pictures to visually impaired, write stories or articles about an event, have conversations in customer-service, chit-chat with individuals, and other settings, and customize content based on the characteristics and goal of the human interlocutor. Neural text generation (NLG) – using neural network models to generate coherent text – have seen a paradigm shift in the last years, caused by the advances in deep contextual language modeling (e.g., LSTMs, GPT, GPT2) and transfer learning (e.g., ELMo, BERT). While these tools have dramatically improved the state of NLG, particularly for low resources tasks, state-of-the-art NLG models still face many challenges: a lack of diversity in generated text, commonsense violations in depicted situations, difficulties in making use of factual information, and difficulties in designing reliable evaluation metrics. In this talk I will discuss existing work on text only transformers that specifies how to generate long-text with better discourse structure and narrative flow, generate multi-document summaries, build automatic knowledge graphs with commonsense transformers as text generators. I will conclude the talk with a discussion of current challenges and shortcomings of neural text generation, pointing to avenues for future research. Biography: Asli Celikyilmaz is a Principal Researcher at Microsoft Research in Redmond, Washington. She is also an Affiliate Professor at the University of Washington. Her research interests are mainly in deep learning and natural language, specifically on language generation with long-term coherence, language understanding, language grounding with vision, and building intelligent agents for human-computer interaction She has received several “best of” awards including NAFIPS 2007, Semantic Computing 2009, and CVPR 2019.