Christopher D. Manning is a computer scientist and the Thomas M. Siebel Professor at Stanford University, where he directs the Stanford Artificial Intelligence Laboratory (SAIL) and has been one of the field's most influential figures in Natural language processing for more than two decades.
Manning co-authored two of the field's standard graduate textbooks, "Foundations of Statistical Natural Language Processing" (1999, with Hinrich Schutze) and "Introduction to Information Retrieval" (2008, with Prabhakar Raghavan and Hinrich Schutze), which trained a generation of NLP researchers during the field's shift from hand-written linguistic rules toward statistical and, later, neural methods.
GloVe and word embeddings
In 2014, Manning's group, including Jeffrey Pennington and Richard Socher, published GloVe (Global Vectors for Word Representation), an Embedding method that learned dense vector representations of words from global word co-occurrence statistics across a large corpus. GloVe became one of the two dominant word embedding methods of the mid-2010s alongside Word2vec, and both were widely used as an input representation for downstream Natural language processing systems before being largely superseded by contextual representations learned inside models such as BERT and later large language models.
From statistical to neural NLP
Manning's lab was also active in early neural approaches to parsing, sentiment analysis and machine translation, contributing to the field's broader transition from statistical methods toward Deep learning-based Natural language processing over the 2010s. As director of SAIL, Manning has overseen research spanning core NLP, Computer vision, Robotics and machine learning theory, and Stanford's AI programs have produced a large number of researchers who went on to senior roles at major AI labs including OpenAI and Google DeepMind.
Standing in the field
Manning is widely regarded as one of the researchers who bridged the linguistics-oriented and statistics-oriented traditions in NLP, and later the statistical and neural traditions, publishing consistently across each shift in dominant methodology rather than being identified with only one era. He has continued to comment publicly on how the rise of large language models has reshaped what counts as an open research problem in NLP, noting that many long-standing subfields, such as parsing and part-of-speech tagging, were effectively solved or subsumed by general-purpose pretrained models.