General Inquirer

Lexicon categorising words across multiple psycholinguistic 🧠 dimensions including semantics, values, motivations and much more.

Summary

Composition:

• Approximately 11.8k unigram lexicon entries

• Each lexicons are “mapped” under 182 semantic and linguistic categories

• Multi-Class Multi-Label (NOTE: For simplicity of application, the processed dictionary currently labels lexicons with only ‘Positive’ or ‘Negative’ tags. )

Creation Methodology:

• Manual dictionary construction starting in 1960s, coming from multiple sources including Harvard IV-4, Lasswell Values and Social Cognition research

  • Harvard IV-4: Developed in the 1960s and 1970s to represent concepts from major social science theories at the time, including the work of theorists like Lasswell, Bales, Parsons and Osgood

  • Lasswell Values: Created by Lasswell and colleagues to represent values across domains such as power and affection.

• The main aim was to create valid representations of specific constructs so experts would agree on category entries.

• Several iterations and expansions over decades, with rigorous multi-cycle human validation

Evaluation: The General Inquirer does not specify the exact evaluation methods. However, the passage states ‘for some studies, overall Inquirer codings correlated with those of expert manual coders about as well as expert coders correlated with one another’. This suggests the Inquirer correlations with experts were reasonably strong, and in the same range as inter-coder reliability.

Usage Guidance: Provides broad psycholinguistic assessment. Combine with modern lexicons for sentiment tasks. Access via sentibank.archive.load().dict(“HarvardGI_v2000”)

📋 Introduction

Arising from the work of Stone et al. (1962), the Harvard General Inquirer serves as a fundamental ‘mapping tool,’ categorizing each lexicon based on three primary sources. The sources include Harvard IV-4, Lasswell Value, and Semin and Fiedler (1988). The categories provide rich semantic classifications of words that can enable more detailed text analysis. With over 180 categories, the Inquirer dictionaries offer lexical resources beyond basic sentiment or topical classification.

📚 Original Dictionary

Harvard IV-4

The Harvard IV-4 categories cover a wide range of lexical classifications including semantic dimensions, social institutions, roles and relations, cognition, communication, and more. These move beyond basic sentiment to enable more nuanced text analysis.

Category	Description	Sub-categories
1. Osgood Dimensions	Words categorised along semantic differential dimensions	Pstv, Ngtv, Strong, Weak, Active, Passive
2. Pleasure/Pain	Words related to emotions, feelings and arousal	Pleasur, Pain, Feel, Arousal, EMOT, Virtue, Vice
3. Over/Understatement	Exaggerated versus tentative language	Ovrst, Undrst
4. Institutions	Language associated with specific social institutions	Academ, Doctrin, Econ@, Exch, Exprsv, Sports, Arts, Legal, Milit, Polit@, Relig
5. Social Relations	Words about social roles, groups and interactions	Role, COLL, Work, Ritual, SocRel
6. Social Categories	Lexical categories based on social/demographic groups	Race, Kin@, MALE, Female, NonAdlt, HU, ANI
7. Places	Words referring to locations and spatial relationships	Place, Social, Region, Route, Aquatic, Land, Sky
8. Objects	Lexical classifications of objects and artefacts	Object, Tool, Food, Vehicle, BldgPt, ComnObj, NatObj, BodyPt
9. Communication	Words related to communication modes and media	ComForm, Say
10. Motivation	Language related to needs, goals and achievement	Need, Goal, Try, Means, Persist, Complet, Fail
11. Processes	Words associated with processes, change and movement	NatrPro, Begin, Vary, Increas, Decreas, Finish, Stay, Rise, Exert, Fetch, Travel, Fall
12. Cognition	Lexical categories related to thought, knowledge and reasoning	Think, Know, Causal, Ought, Perceiv, Compare, Eval@, Solve, Abs@, Quality, Quan, FREQ, DIST, Time@, Space, Rel, COLOR
13. Pronouns	References to 1st, 2nd and 3rd person points of view	Self, Our, You, Name
14. Yes/No	Words indicating agreement, disagreement, negation and interjection	Yes, No, Negate, Intrj

New categories based on social cognition

The categories are based on social cognition research focused specifically on categorising verbs and adjectives: Verbs based on whether they are interpretative, descriptive or state; Adjectives based on interpersonal versus individual. These categories reveal perspective and social dynamics.

Category	Description	Sub-categories
15. Verb Types	Classification of verbs based on interpretive explanation of an action, description of an action, and description of mental/emotional states	IAV, DAV, SV
16. Adjective Types	Categorization of adjectives as interpersonal vs individual	IPadj, IndAdj

Lasswell Value

The Lasswell value categories provide a sociological classification of language into four deference domains - power, rectitude, respect, and affiliation - and four welfare domains - wealth, well-being, enlightenment and skill. This provides resources for understanding values and motivations.

Category	Description	Sub-categories
17. Power	Words related to influence, authority and politics	PowGain, PowLoss, PowEnds, PowAren, PowCon, PowCoop, PowAuPt, PowPt, PowDoct, PowAuth, PowOth, PowTot
18. Rectitude	Moral and ethical language	RcEthic, RcRelig, RcGain, RcLoss, RcEnds, RcTot
19. Respect	Words associated with status, honour, recognition and prestige	RspGain, RspLoss, RspOth, RspTot
20. Affection	Language related to relationships and emotions, with a particular focus of love and friendship	AffGain, AffLoss, AffPt, AffOth, AffTot
21. Wealth	Economically oriented words	WltPt, WltTran, WltOth, WltTot
22. Well-Being	Words related to health, safety and security	WlbGain, WlbLoss, WlbPhys, WlbPsyc, WlbPt, WlbTot
23. Enlightenment	Words associated with knowledge, insight and understanding of personal and cultural relations	EnlGain, EnlLoss, EnlEnds, EnlPt, EnlOth, EnlTot
24. Skill	Language related to competencies and professions, with a particular focus on the arts	SklAsth, SklPt, SklOth, SklTot
25. Other	Additional Lasswell lexical categories	TrnGain, TrnLoss, TranLw, MeansLw, EndsLw, ArenaLw, PtLw, Nation, Anomie, NegAff, PosAff, SureLw, If, NotLw, TimeSpc, FormLw

from sentibank import archive 

load = archive.load()
harvard_gi = load.origin("HarvardGI_v2000")

General Inquirer

Entry	Source	Positiv	Negativ	Pstv	Affil	Ngtv	Hostile	Strong	Power	Weak	Submit	Active	Passive	Pleasur	Pain	Feel	Arousal	EMOT	Virtue	Vice	Ovrst	Undrst	Academ	Doctrin	Econ@	Exch	ECON	Exprsv	Legal	Milit	Polit@	POLIT	Relig	Role	COLL	Work	Ritual	SocRel	Race	Kin@	MALE	Female	Nonadlt	HU	ANI	PLACE	Social	Region	Route	Aquatic	Land	Sky	Object	Tool	Food	Vehicle	BldgPt	ComnObj	NatObj	BodyPt	ComForm	COM	Say	Need	Goal	Try	Means	Persist	Complet	Fail	NatrPro	Begin	Vary	Increas	Decreas	Finish	Stay	Rise	Exert	Fetch	Travel	Fall	Think	Know	Causal	Ought	Perceiv	Compare	Eval@	EVAL	Solve	Abs@	ABS	Quality	Quan	NUMB	ORD	CARD	FREQ	DIST	Time@	TIME	Space	POS	DIM	Rel	COLOR	Self	Our	You	Name	Yes	No	Negate	Intrj	IAV	DAV	SV	IPadj	IndAdj	PowGain	PowLoss	PowEnds	PowAren	PowCon	PowCoop	PowAuPt	PowPt	PowDoct	PowAuth	PowOth	PowTot	RcEthic	RcRelig	RcGain	RcLoss	RcEnds	RcTot	RspGain	RspLoss	RspOth	RspTot	AffGain	AffLoss	AffPt	AffOth	AffTot	WltPt	WltTran	WltOth	WltTot	WlbGain	WlbLoss	WlbPhys	WlbPsyc	WlbPt	WlbTot	EnlGain	EnlLoss	EnlEnds	EnlPt	EnlOth	EnlTot	SklAsth	SklPt	SklOth	SklTot	TrnGain	TrnLoss	TranLw	MeansLw	EndsLw	ArenaLw	PtLw	Nation	Anomie	NegAff	PosAff	SureLw	If	NotLw	TimeSpc	FormLw	Othtags	Defined
Loading... (need help?)

🧹 Processed Dictionary

First-Pass Processing

Drawing on the three resources, the General Inquirer introduced a new category called ‘Valence’, which provides positive and negative classifications to enable simple sentiment analysis. The Positive category contains 1,915 entries compiled from the source dictionaries, while the Negative category has 2,291 entries. By aggregating multiple existing lexicons, the Valence category offers broad coverage for classifying sentiment.

From the 4,206 lexicons, there were 381 unigrams with duplicate entries^[1]. The majority of duplicates displayed consistent sentiment labels, allowing them to be consolidated into a single entry. However, 16 entries with conflicting labels^[2] were deleted due to their contextual ambiguation with sentiment signals. Consequently, the refined dictionary now comprises 3,610 entries.

Note

^[1] Below is the frequency table showing the distribution for duplicate entries.

Duplicates	Unigrams
2	255
3	77
4	32
5	12
6	3
7	2

^[2] The entries with a conflicting labels were: ‘arrest’, ‘board’, ‘deal’, ‘even’, ‘fine’, ‘fun’, ‘hand’, ‘help’, ‘hit’, ‘laugh’, ‘make’, ‘matter’, ‘mind’, ‘order’, ‘particular’ and ‘pass’.