What company do words keep? Revisiting the distributional semantics of J.R. Firth & Zellig Harris
Not on the map. Why?
Abstract (arXiv)
The power of word embeddings is attributed to the linguistic theory that similar words will appear in similar contexts. This idea is specifically invoked by noting that "you shall know a word by the company it keeps," a quote from British linguist J.R. Firth who, along with his American colleague Zellig Harris, is often credited with the invention of "distributional semantics." While both Firth and Harris are cited in all major NLP textbooks and many foundational papers, the content and differences between their theories is seldom discussed. Engaging in a close reading of their work, we discover two distinct and in many ways divergent theories of meaning. One focuses exclusively on the internal workings of linguistic forms, while the other invites us to consider words in new company - not just with other linguistic elements, but also in a broader cultural and situational context. Contrasting these theories from the perspective of current debates in NLP, we discover in Firth a figure who could guide the field towards a more culturally grounded notion of semantics. We consider how an expanded notion of "context" might be modeled in practice through two different strategies: comparative stratification and syntagmatic extension
Key facts. Cited by 8 papers in the map (data through 14 Oct 2026). Builds on 29 of its 151 references that are in the map. First posted on arXiv 16 May 2022.
Builds on
29 of 151 references.
References · 29
- 2017Attention Is All You Needcited by 60,361
- 2018BERT: Pre-training of Deep Bidirectional Transformers for Language Understandingcited by 40,399
- 2022Hierarchical Text-Conditional Image Generation with CLIP Latentscited by 6,585
- 2022PaLM: Scaling Language Modeling with Pathwayscited by 5,326
- 2014Neural Machine Translation by Jointly Learning to Align and Translatecited by 4,548
- 2013Distributed Representations of Words and Phrases and their Compositionalitycited by 3,474
- 2018Deep contextualized word representationscited by 2,290
- 2016Enriching Word Vectors with Subword Informationcited by 1,583
- 2020Language (Technology) is Power: A Critical Survey of "Bias" in NLPcited by 956
- 2014Distributed Representations of Sentences and Documentscited by 678
- 2013Speech Recognition with Deep Recurrent Neural Networkscited by 669
- 2019What do you learn from context? Probing for sentence structure in contextualized word representationscited by 438
17 more
- 2019Social Bias Frames: Reasoning about Social and Power Implications of Languagecited by 380
- 2017Word Embeddings Quantify 100 Years of Gender and Ethnic Stereotypescited by 253
- 2020Experience Grounds Languagecited by 242
- 2016A Persona-Based Neural Conversation Modelcited by 223
- 2016Diachronic Word Embeddings Reveal Statistical Laws of Semantic Changecited by 221
- 2018Gender Bias in Neural Natural Language Processingcited by 211
- 2020A Survey of the State of Explainable AI for Natural Language Processingcited by 206
- 2018Diachronic word embeddings and semantic shifts: a surveycited by 92
- 2018The Geometry of Culture: Analyzing Meaning through Word Embeddingscited by 87
- 2019Leveraging Contextual Embeddings for Detecting Diachronic Semantic Shiftcited by 37
- 2020A Framework for the Computational Linguistic Analysis of Dehumanizationcited by 31
- 2021Characterizing English Variation across Social Media Communities with BERTcited by 20
- 2020Language (Re)modelling: Towards Embodied Language Understandingcited by 16
- 2020(Re)construing Meaning in NLPcited by 16
- 2020The Unstoppable Rise of Computational Linguistics in Deep Learningcited by 12
- 2021Identification of Biased Terms in News Articles by Comparison of Outlet-specific Word Embeddingscited by 9
- 2020Enriching Word Embeddings with Temporal and Spatial Informationcited by 2
Built on by
Citing papers per year: 2022: 1; 2023: 2; 2024: 3; 2025: 0; 2026: 2.
8 papers.
- 2024Reducing the Footprint of Multi-Vector Retrieval with Minimal Performance Impact via Token Poolingcited by 25
- 2022Norm of Word Embedding Encodes Information Gaincited by 21
- 2024Can You Learn Semantics Through Next-Word Prediction? The Case of Entailmentcited by 9
- 2024From Frege to chatGPT: Compositionality in language, cognition, and deep neural networkscited by 7
- 2023Computational modeling of semantic changecited by 2
- 2023Toward a Critical Toponymy Framework for Named Entity Recognition: A Case Study of Airbnb in New York Citycited by 1
- 2026A visual observation on the geometry of UMAP projections of the difference vectors of antonym and synonym word pair embeddingscited by 0
- 2026Gyan: An Explainable Neuro-Symbolic Language Modelcited by 0
- 2026Gyan: An Explainable Neuro-Symbolic Language Model6 May 2026
- 2026A visual observation on the geometry of UMAP projections of the difference vectors of antonym and synonym word pair embeddings25 Mar 2026
- 2024Reducing the Footprint of Multi-Vector Retrieval with Minimal Performance Impact via Token Pooling23 Sep 2024
- 2024From Frege to chatGPT: Compositionality in language, cognition, and deep neural networks24 May 2024
- 2024Can You Learn Semantics Through Next-Word Prediction? The Case of Entailment21 Feb 2024
- 2023Toward a Critical Toponymy Framework for Named Entity Recognition: A Case Study of Airbnb in New York City23 Oct 2023
- 2023Computational modeling of semantic change13 Apr 2023
- 2022Norm of Word Embedding Encodes Information Gain19 Dec 2022