Knowledge Vault 5 /22 - CVPR 2017
Extracting Social Meaning from Language
Dan Jurafsky
< Resume Image >

Concept Graph & Resume using Claude 3 Opus | Chat GPT4o | Llama 3:

graph LR classDef innovation fill:#f9d4d4, font-weight:bold, font-size:14px classDef nlp fill:#d4f9d4, font-weight:bold, font-size:14px classDef language fill:#d4d4f9, font-weight:bold, font-size:14px classDef society fill:#f9f9d4, font-weight:bold, font-size:14px A[Extracting Social Meaning
from Language] --> B[Innovation: borrowing, extending
neighbors ideas. 1] A --> C[1988 NLP revolution:
speech researchers. 2] C --> D[Pollination model:
neighboring fields inspire. 3] A --> E[Police: less respect,
black communities. 4] E --> F[Respect detection:
leveraging social theories. 5] E --> G[NLP: improve police-community
interactions. 6] A --> H[Cheap menus: vague
positive words. 7] H --> I[Appealing vegetable descriptions
increased consumption. 8] H --> J[Persuasive product descriptions
boost sales. 9] H --> K[One-star reviews:
trauma-like language. 10] H --> L[Positive reviews: food
as sex/drugs. 11] A --> M[Language reveals psychology, norms. 12] M --> N[Social science aids
NLP interpretation. 13] A --> O[CS research: understand,
improve society. 14] A --> P[Interdisciplinary collaboration
drives innovation. 15] class A,B,P innovation class C,D,E,F,G nlp class H,I,J,K,L,M,N language class O society


1.- Innovation happens at interstices by borrowing and extending ideas from neighbors, as exemplified by the history of ketchup.

2.- The statistical revolution in NLP in 1988 was driven by speech recognition researchers presenting at NLP conferences.

3.- A "pollination model" of scientific innovation: people from neighboring fields present papers, students get excited, math stays, people leave.

4.- Police officers speak with less respect toward black community members than white ones, even in routine traffic stops.

5.- Respect can be automatically detected in language, requiring little training data by leveraging rich social science domain theories.

6.- Studying police-community interactions with NLP can help improve training and understand racial disparities in treatment.

7.- Cheap restaurant menus use vague positive words, mid-priced ones use sensory adjectives, expensive ones are short and use uncommon words.

8.- Describing vegetables with appealing indulgent words led to more consumption than plain descriptions in a dining hall study.

9.- Product descriptions using appeals to authority, Japanese politeness markers, and providing information lead to higher sales on Rakuten.

10.- One-star Yelp reviews resemble trauma narratives, with "we/us" pronouns, past tense, negative words, and focus on people rather than food.

11.- Positive Yelp reviews of expensive restaurants use sexual metaphors; cheap restaurant reviews describe food as drugs, especially by women.

12.- Everyday language, like menus and reviews, provide crucial cues about psychology, society, and subconscious norms.

13.- Social science domain knowledge is important for interpreting NLP results and developing models that capture complex social phenomena.

14.- Computer science research should strive to better understand society and make a positive impact on the world.

15.- Collaboration across disciplines, despite challenges for academic careers, can drive important innovations.

Knowledge Vault built byDavid Vivancos 2024