feat(sentiment): Improve scoring and analysis

* Implement english-only scoring with wink https://github.com/winkjs/wink-sentiment (AFINN, emojis)
* Implement english-only scoring with NLP.js https://github.com/axa-group/nlp.js/blob/master/docs/v3/sentiment-analysis.md (AFINN, Senticon, Pattern)
* Refactor language processing into standalone functions for future use
* Add limited multi-langauge support
  * Can run sentiment with NLP.js on english, german, spanish, and french
* Normalize all scores to range between -1 and +1
* Improve score accuracy by averaging all scores

package.json

@@ -67,6 +67,7 @@
     "mustache": "^4.2.0",
     "nanoid": "^3.3.1",
     "node-fetch": "^2.6.1",
+    "node-nlp": "^4.24.0",
     "normalize-url": "^6.1.0",
     "object-hash": "^2.2.0",
     "p-event": "^4.2.0",
@@ -92,6 +93,7 @@
     "typescript": "^4.3.4",
     "vader-sentiment": "^1.1.3",
     "webhook-discord": "^3.7.7",
+    "wink-sentiment": "^5.0.2",
     "winston": "github:FoxxMD/winston#fbab8de969ecee578981c77846156c7f43b5f01e",
     "winston-daily-rotate-file": "^4.5.5",
     "winston-duplex": "^0.1.1",

src/Common/Infrastructure/Atomic.ts

@@ -190,29 +190,40 @@ export type ActionTypes =
 /**
  * Test the calculated VADER sentiment (compound) score for an Activity using this comparison. Can be either a numerical or natural language
  *
- * Sentiment values range from extremely negative to extremely positive in a numerical range of -0.5 to 0.5:
+ * Sentiment values range from extremely negative to extremely positive in a numerical range of -1 to +1:
  *
- * * -0.5 => extremely negative
- * * -0.2 => very negative
- * * -0.5 => negative
+ * * -0.6 => extremely negative
+ * * -0.3 => very negative
+ * * -0.1 => negative
  * *    0 => neutral
- * * 0.05 => positive
- * *  0.2 => very positive
- * *  0.5 => extremely positive
+ * *  0.1 => positive
+ * *  0.3 => very positive
+ * *  0.6 => extremely positive
  *
  * The below examples are all equivocal. You can use either set of values as the value for `sentiment` (numerical comparisons or natural langauge)
  *
- * * `>= 0.05` = `is positive`
- * * `<= 0.2` = `is very negative`
- * * `< 0.05` = `is not positive`
- * * `> -0.2` = `is not very negative`
+ * * `>= 0.1` = `is positive`
+ * * `<= 0.3` = `is very negative`
+ * * `< 0.1` = `is not positive`
+ * * `> -0.3` = `is not very negative`
  *
  * Special case:
  *
- * * `is neutral` equates to `> -0.5 and < 0.5`
- * * `is not neutral` equates to `< -0.5 or > 0.5`
+ * * `is neutral` equates to `> -0.1 and < 0.1`
+ * * `is not neutral` equates to `< -0.1 or > 0.1`
  *
- * More about VADER Sentiment: https://github.com/cjhutto/vaderSentiment
+ * ContextMod uses a normalized, weighted average from these sentiment tools:
+ *
+ * * NLP.js (english, french, german, and spanish) https://github.com/axa-group/nlp.js/blob/master/docs/v3/sentiment-analysis.md
+ * * (english only) vaderSentiment-js https://github.com/vaderSentiment/vaderSentiment-js/
+ * * (english only) wink-sentiment https://github.com/winkjs/wink-sentiment
+ *
+ * More about the sentiment algorithms used:
+ * * VADER https://github.com/cjhutto/vaderSentiment
+ * * AFINN http://corpustext.com/reference/sentiment_afinn.html
+ * * Senticon https://ieeexplore.ieee.org/document/8721408
+ * * Pattern https://github.com/clips/pattern
+ * * wink https://github.com/winkjs/wink-sentiment
  *
  * @pattern ((>|>=|<|<=)\s*(-?\d?\.?\d+))|((not)?\s*(very|extremely)?\s*(positive|neutral|negative))
  * @examples ["is negative", "> 0.2"]

src/Common/LangaugeProcessing.ts

@@ -0,0 +1,267 @@
+import {Language, SentimentManager} from 'node-nlp';
+import {SentimentIntensityAnalyzer} from 'vader-sentiment';
+import wink from 'wink-sentiment';
+import {SnoowrapActivity} from "./Infrastructure/Reddit";
+import {
+    asGenericComparison,
+    GenericComparison,
+    parseGenericValueComparison,
+    RangedComparison
+} from "./Infrastructure/Comparisons";
+import {asSubmission, between, comparisonTextOp} from "../util";
+import {CMError, MaybeSeriousErrorWithCause} from "../Utils/Errors";
+import InvalidRegexError from "../Utils/InvalidRegexError";
+import {StringOperator} from "./Infrastructure/Atomic";
+
+export type SentimentAnalysisType = 'vader' | 'afinn' | 'senticon' | 'pattern' | 'wink';
+
+export const sentimentQuantifier = {
+    'extremely negative': -0.6,
+    'very negative': -0.3,
+    'negative': -0.1,
+    'neutral': 0,
+    'positive': 0.1,
+    'very positive': 0.3,
+    'extremely positive': 0.6,
+}
+
+export const sentimentQuantifierRanges = [
+    {
+        range: [Number.MIN_SAFE_INTEGER, -0.6],
+        quant: 'extremely negative'
+    },
+    {
+        range: [-0.6, -0.3],
+        quant: 'very negative'
+    },
+    {
+        range: [-0.3, -0.1],
+        quant: 'negative'
+    },
+    {
+        range: [-0.1, 0.1],
+        quant: 'neutral'
+    },
+    {
+        range: [0.1, 0.3],
+        quant: 'positive'
+    },
+    {
+        range: [0.3, 0.6],
+        quant: 'very positive'
+    },
+    {
+        range: [0.6, Number.MAX_SAFE_INTEGER],
+        quant: 'extremely positive'
+    }
+]
+
+const scoreToSentimentText = (val: number) => {
+    for(const segment of sentimentQuantifierRanges) {
+        if(between(val, segment.range[0], segment.range[1], false)) {
+            return segment.quant;
+        }
+    }
+    throw new Error('should not hit this!');
+}
+
+export interface SentimentResult {
+    comparative: number
+    type: SentimentAnalysisType
+    sentiment: string
+    weight: number
+}
+
+export interface ActivitySentiment {
+    results: SentimentResult[]
+    score: number
+    scoreWeighted: number
+    sentiment: string
+    sentimentWeighted: string
+    activity: SnoowrapActivity
+    language: string
+}
+
+export interface ActivitySentimentTestResult extends ActivitySentiment {
+    passes: boolean
+    test: GenericComparison | RangedComparison
+}
+
+export interface ActivitySentimentOptions {
+    testOn?: ('title' | 'body')[]
+}
+
+export type SentimentCriteriaTest = GenericComparison | RangedComparison;
+
+export const availableSentimentLanguages = ['en', 'es', 'de', 'fr'];
+
+export const textComparison = /(?<not>not)?\s*(?<modifier>very|extremely)?\s*(?<sentiment>positive|neutral|negative)/i;
+
+export const parseTextToNumberComparison = (val: string): RangedComparison | GenericComparison => {
+
+    let genericError: Error | undefined;
+    try {
+        return parseGenericValueComparison(val);
+    } catch (e) {
+        genericError = e as Error;
+        // now try text match
+    }
+
+    const matches = val.match(textComparison);
+    if (matches === null) {
+        const textError = new InvalidRegexError(textComparison, val);
+        throw new CMError(`Sentiment value did not match a valid numeric comparison or valid text: \n ${genericError.message} \n ${textError.message}`);
+    }
+    const groups = matches.groups as any;
+
+    const negate = groups.not !== undefined && groups.not !== '';
+
+    if(groups.sentiment === 'neutral') {
+        if(negate) {
+            return {
+                displayText: 'not neutral (not -0.49 to 0.49)',
+                range: [-1, 1],
+                not: true,
+            }
+        }
+        return {
+            displayText: 'is neutral (-0.49 to 0.49)',
+            range: [-1, 1],
+            not: false
+        }
+    }
+
+    const compoundSentimentText = `${groups.modifier !== undefined && groups.modifier !== '' ? `${groups.modifier} ` : ''}${groups.sentiment}`.toLocaleLowerCase();
+    // @ts-ignore
+    const numericVal = sentimentQuantifier[compoundSentimentText] as number;
+    if(numericVal === undefined) {
+        throw new CMError(`Sentiment given did not match any known phrases: '${compoundSentimentText}'`);
+    }
+
+    let operator: StringOperator;
+    if(negate) {
+        operator = numericVal > 0 ? '<' : '>';
+    } else {
+        operator = numericVal > 0 ? '>=' : '<=';
+    }
+
+    return {
+        operator,
+        value: numericVal,
+        isPercent: false,
+        displayText: `is${negate ? ' not ': ' '}${compoundSentimentText} (${operator} ${numericVal})`
+    }
+}
+
+const nlpAnalyzer = new SentimentManager();
+const langDetect = new Language();
+
+export const getActivitySentiment = async (item: SnoowrapActivity, options?: ActivitySentimentOptions): Promise<ActivitySentiment> => {
+
+    const {
+        testOn = ['body', 'title']
+    } = options || {};
+
+    // determine what content we are testing
+    let contents: string[] = [];
+    if (asSubmission(item)) {
+        for (const l of testOn) {
+            switch (l) {
+                case 'title':
+                    contents.push(item.title);
+                    break;
+                case 'body':
+                    if (item.is_self) {
+                        contents.push(item.selftext);
+                    }
+                    break;
+            }
+        }
+    } else {
+        contents.push(item.body)
+    }
+
+    const contentStr = contents.join(' ');
+
+    const guess = langDetect.guessBest(contentStr);
+
+    if (availableSentimentLanguages.includes(guess.alpha2)) {
+        const results: SentimentResult[] = [];
+
+        const nlpResult = await nlpAnalyzer.process(guess.alpha2, contentStr);
+
+        results.push({
+            comparative: nlpResult.comparative,
+            type: nlpResult.type as SentimentAnalysisType,
+            sentiment: scoreToSentimentText(nlpResult.comparative),
+            weight: 1
+        });
+
+        if (guess.alpha2 === 'en') {
+            const score = SentimentIntensityAnalyzer.polarity_scores(contentStr);
+            results.push({
+                comparative: score.compound,
+                type: 'vader',
+                sentiment: scoreToSentimentText(score.compound),
+                // may want to weight higher in the future...
+                weight: 1
+            });
+
+            const winkScore = wink(contentStr);
+            // normalizedScore is range of -5 to +5 -- convert to -1 to +1
+            const winkAdjusted = (winkScore.normalizedScore * 2) / 10;
+            results.push({
+                comparative: winkAdjusted,
+                type: 'wink',
+                sentiment: scoreToSentimentText(winkAdjusted),
+                weight: 1
+            })
+        }
+
+        const score = results.reduce((acc, curr) => acc + curr.comparative, 0) / results.length;
+        const sentiment = scoreToSentimentText(score);
+
+        const weightSum = results.reduce((acc, curr) => acc + curr.weight, 0);
+        const weightedScores = results.reduce((acc, curr) => acc + (curr.weight * curr.comparative), 0);
+        const weightedScore = weightedScores / weightSum;
+        const weightedSentiment = scoreToSentimentText(weightedScore);
+
+        return {
+            results,
+            score,
+            sentiment,
+            scoreWeighted: weightedScore,
+            sentimentWeighted: weightedSentiment,
+            activity: item,
+            language: guess.language
+        }
+
+    } else {
+        throw new MaybeSeriousErrorWithCause(`Cannot test sentiment for unsupported language ${guess.language}`);
+    }
+}
+
+export const testActivitySentiment = async (item: SnoowrapActivity, criteria: SentimentCriteriaTest, options?: ActivitySentimentOptions): Promise<ActivitySentimentTestResult> => {
+    const sentimentResult = await getActivitySentiment(item, options);
+
+    if (asGenericComparison(criteria)) {
+        return {
+            passes: comparisonTextOp(sentimentResult.scoreWeighted, criteria.operator, criteria.value),
+            test: criteria,
+            ...sentimentResult,
+        }
+    } else {
+        if (criteria.not) {
+            return {
+                passes: sentimentResult.scoreWeighted < criteria.range[0] || sentimentResult.scoreWeighted > criteria.range[1],
+                test: criteria,
+                ...sentimentResult,
+            }
+        }
+        return {
+            passes: sentimentResult.scoreWeighted >= criteria.range[0] || sentimentResult.scoreWeighted <= criteria.range[1],
+            test: criteria,
+            ...sentimentResult,
+        }
+    }
+}

src/Common/Typings/support.d.ts

@@ -32,3 +32,23 @@ declare module 'winston-null' {
 
     }
 }
+
+declare module 'node-nlp' {
+    export class Language {
+        guess(val: string, allowedList?: string[] | null, limit: number): {alpha3: string, alpha2: string, language: string, score: number}[];
+        guessBest(val: string, allowedList?: string[] | null): {alpha3: string, alpha2: string, language: string, score: number};
+    }
+    export class SentimentAnalyzer {
+        constructor(settings?: {language?: string})
+        getSentiment(phrase: string): Promise<{score: number, comparative: score, vote: 'string', numWords: number, numHits: number, type: string, language: string}>
+    }
+
+    export class SentimentManager {
+        process(locale: string, phrase: string): Promise<{score: number, comparative: score, vote: 'string', numWords: number, numHits: number, type: string, language: string}>
+    }
+}
+
+declare module 'wink-sentiment' {
+    function sentiment(phrase: string): {score: number, normalizedScore: number, tokenizedPhrase: any[]};
+    export default sentiment;
+}

src/Rule/SentimentRule.ts

@@ -2,123 +2,26 @@ import {Rule, RuleJSONConfig, RuleOptions} from "./index";
 import {Comment} from "snoowrap";
 import Submission from "snoowrap/dist/objects/Submission";
 import {
-    asSubmission,
     comparisonTextOp, formatNumber,
-    parseStringToRegex,
     triggeredIndicator, windowConfigToWindowCriteria
 } from "../util";
-import {Scores, SentimentIntensityAnalyzer} from 'vader-sentiment';
 
 import dayjs from 'dayjs';
-import {CMError, SimpleError} from "../Utils/Errors";
-import InvalidRegexError from "../Utils/InvalidRegexError";
+import {map as mapAsync} from 'async';
 import {
-    asGenericComparison,
     GenericComparison,
-    parseGenericValueComparison, parseGenericValueOrPercentComparison,
+    parseGenericValueOrPercentComparison,
     RangedComparison
 } from "../Common/Infrastructure/Comparisons";
 import {ActivityWindowConfig, ActivityWindowCriteria} from "../Common/Infrastructure/ActivityWindow";
-import {StringOperator, VaderSentimentComparison} from "../Common/Infrastructure/Atomic";
+import {VaderSentimentComparison} from "../Common/Infrastructure/Atomic";
 import {RuleResult} from "../Common/interfaces";
 import {SnoowrapActivity} from "../Common/Infrastructure/Reddit";
-
-export const sentimentQuantifier = {
-    'extremely negative': -0.5,
-    'very negative': -0.2,
-    'negative': -0.05,
-    'neutral': 0,
-    'positive': 0.05,
-    'very positive': 0.2,
-    'extremely positive': 0.5,
-}
-
-const sentimentQuantifierEntries = Object.entries(sentimentQuantifier);
-
-const scoreToSentimentText = (val: number) => {
-    for(let i = 0; i < sentimentQuantifierEntries.length; i++) {
-        // hit max score (extremely positive)
-        if(i + 1 === sentimentQuantifierEntries.length && val > sentimentQuantifierEntries[i][1]) {
-            return sentimentQuantifierEntries[i][0];
-        }
-
-        // if score is between (inclusive start) current and next
-        if(i + 1 <= sentimentQuantifierEntries.length && val >= sentimentQuantifierEntries[i][1] && val < sentimentQuantifierEntries[i + 1][1]) {
-            return sentimentQuantifierEntries[i][0];
-        }
-
-        // if neither of those is hit and i = 0 then min score (extremely negative)
-        if(i === 0 && val < sentimentQuantifierEntries[i][1]) {
-            return sentimentQuantifierEntries[i][0];
-        }
-    }
-    throw new Error('should not hit this!');
-}
-
-interface ActivitySentimentResult {
-    passes: boolean,
-    scores: Scores,
-    sentiment: string
-    activity: SnoowrapActivity
-}
-
-export const textComparison = /(?<not>not)?\s*(?<modifier>very|extremely)?\s*(?<sentiment>positive|neutral|negative)/i;
-
-export const parseTextToNumberComparison = (val: string): RangedComparison | GenericComparison => {
-
-    let genericError: Error | undefined;
-    try {
-        return parseGenericValueComparison(val);
-    } catch (e) {
-        genericError = e as Error;
-        // now try text match
-    }
-
-    const matches = val.match(textComparison);
-    if (matches === null) {
-        const textError = new InvalidRegexError(textComparison, val);
-        throw new CMError(`Sentiment value did not match a valid numeric comparison or valid text: \n ${genericError.message} \n ${textError.message}`);
-    }
-    const groups = matches.groups as any;
-
-    const negate = groups.not !== undefined && groups.not !== '';
-
-    if(groups.sentiment === 'neutral') {
-        if(negate) {
-            return {
-                displayText: 'not neutral (not -0.49 to 0.49)',
-                range: [-0.49, 0.49],
-                not: true,
-            }
-        }
-        return {
-            displayText: 'is neutral (-0.49 to 0.49)',
-            range: [-0.49, 0.49],
-            not: false
-        }
-    }
-
-    const compoundSentimentText = `${groups.modifier !== undefined && groups.modifier !== '' ? `${groups.modifier} ` : ''}${groups.sentiment}`.toLocaleLowerCase();
-    // @ts-ignore
-    const numericVal = sentimentQuantifier[compoundSentimentText] as number;
-    if(numericVal === undefined) {
-        throw new CMError(`Sentiment given did not match any known phrases: '${compoundSentimentText}'`);
-    }
-
-    let operator: StringOperator;
-    if(negate) {
-        operator = numericVal > 0 ? '<' : '>';
-    } else {
-        operator = numericVal > 0 ? '>=' : '<=';
-    }
-
-    return {
-        operator,
-        value: numericVal,
-        isPercent: false,
-        displayText: `is${negate ? ' not ': ' '}${compoundSentimentText} (${operator} ${numericVal})`
-    }
-}
+import {
+    ActivitySentimentTestResult,
+    parseTextToNumberComparison,
+    testActivitySentiment
+} from "../Common/LangaugeProcessing";
 
 export class SentimentRule extends Rule {
 
@@ -166,10 +69,8 @@ export class SentimentRule extends Rule {
 
     protected async process(item: Submission | Comment): Promise<[boolean, RuleResult]> {
 
-        let criteriaResults = [];
-
-        let ogResult = this.testActivity(item, this.sentiment);
-        let historicResults: ActivitySentimentResult[] | undefined;
+        let ogResult = await this.testActivity(item, this.sentiment);
+        let historicResults: ActivitySentimentTestResult[] | undefined;
 
         if(this.historical !== undefined && (!this.historical.mustMatchCurrent || ogResult.passes)) {
             const {
@@ -178,7 +79,7 @@ export class SentimentRule extends Rule {
             } = this.historical;
             const history = await this.resources.getAuthorActivities(item.author, window);
 
-            historicResults = history.map(x => this.testActivity(x, sentiment));
+            historicResults = await mapAsync(history, async (x: SnoowrapActivity) => await this.testActivity(x, sentiment)); // history.map(x => this.testActivity(x, sentiment));
         }
 
 
@@ -197,8 +98,8 @@ export class SentimentRule extends Rule {
 
         if(historicResults === undefined) {
             triggered = ogResult.passes;
-            averageScore = ogResult.scores.compound;
-            logSummary.push(`${triggeredIndicator(triggered)} Current Activity Sentiment '${ogResult.sentiment} (${ogResult.scores.compound})' ${triggered ? 'PASSED' : 'DID NOT PASS'} sentiment test '${sentimentTest}'`);
+            averageScore = ogResult.scoreWeighted;
+            logSummary.push(`${triggeredIndicator(triggered)} Current Activity Sentiment '${ogResult.sentiment} (${ogResult.scoreWeighted})' ${triggered ? 'PASSED' : 'DID NOT PASS'} sentiment test '${sentimentTest}'`);
             if(!triggered && this.historical !== undefined && this.historical.mustMatchCurrent) {
                 logSummary.push(`Did not check Historical because 'mustMatchCurrent' is true`);
             }
@@ -212,8 +113,8 @@ export class SentimentRule extends Rule {
             totalMatchingText = totalMatching.displayText;
             const allResults = historicResults
             const passed = allResults.filter(x => x.passes);
-            averageScore = passed.reduce((acc, curr) => acc + curr.scores.compound,0) / passed.length;
-            averageWindowScore = allResults.reduce((acc, curr) => acc + curr.scores.compound,0) / allResults.length;
+            averageScore = passed.reduce((acc, curr) => acc + curr.scoreWeighted,0) / passed.length;
+            averageWindowScore = allResults.reduce((acc, curr) => acc + curr.scoreWeighted,0) / allResults.length;
 
             const firstActivity = allResults[0].activity;
             const lastActivity = allResults[allResults.length - 1].activity;
@@ -253,53 +154,8 @@ export class SentimentRule extends Rule {
         })]);
     }
 
-    protected testActivity(a: (Submission | Comment), criteria: GenericComparison | RangedComparison): ActivitySentimentResult {
-        // determine what content we are testing
-        let contents: string[] = [];
-        if (asSubmission(a)) {
-            for (const l of this.testOn) {
-                switch (l) {
-                    case 'title':
-                        contents.push(a.title);
-                        break;
-                    case 'body':
-                        if (a.is_self) {
-                            contents.push(a.selftext);
-                        }
-                        break;
-                }
-            }
-        } else {
-            contents.push(a.body)
-        }
-
-        const contentStr = contents.join(' ');
-        const scores = SentimentIntensityAnalyzer.polarity_scores(contentStr);
-        const textSentimentScore = scoreToSentimentText(scores.compound);
-
-        if (asGenericComparison(criteria)) {
-            return {
-                passes: comparisonTextOp(scores.compound, criteria.operator, criteria.value),
-                scores,
-                sentiment: textSentimentScore,
-                activity: a
-            }
-        } else {
-            if (criteria.not) {
-                return {
-                    passes: scores.compound < criteria.range[0] || scores.compound > criteria.range[1],
-                    scores,
-                    sentiment: textSentimentScore,
-                    activity: a
-                }
-            }
-            return {
-                passes: scores.compound >= criteria.range[0] || scores.compound <= criteria.range[1],
-                scores,
-                sentiment: textSentimentScore,
-                activity: a
-            }
-        }
+    protected async testActivity(a: (Submission | Comment), criteria: GenericComparison | RangedComparison): Promise<ActivitySentimentTestResult> {
+        return await testActivitySentiment(a, criteria, {testOn: this.testOn});
     }
 }
 

src/Schema/App.json

@@ -2926,7 +2926,7 @@
                     "type": "boolean"
                 },
                 "sentiment": {
-                    "description": "Test the calculated VADER sentiment (compound) score for an Activity using this comparison. Can be either a numerical or natural language\n\nSentiment values range from extremely negative to extremely positive in a numerical range of -0.5 to 0.5:\n\n* -0.5 => extremely negative\n* -0.2 => very negative\n* -0.5 => negative\n*    0 => neutral\n* 0.05 => positive\n*  0.2 => very positive\n*  0.5 => extremely positive\n\nThe below examples are all equivocal. You can use either set of values as the value for `sentiment` (numerical comparisons or natural langauge)\n\n* `>= 0.05` = `is positive`\n* `<= 0.2` = `is very negative`\n* `< 0.05` = `is not positive`\n* `> -0.2` = `is not very negative`\n\nSpecial case:\n\n* `is neutral` equates to `> -0.5 and < 0.5`\n* `is not neutral` equates to `< -0.5 or > 0.5`\n\nMore about VADER Sentiment: https://github.com/cjhutto/vaderSentiment",
+                    "description": "Test the calculated VADER sentiment (compound) score for an Activity using this comparison. Can be either a numerical or natural language\n\nSentiment values range from extremely negative to extremely positive in a numerical range of -1 to +1:\n\n* -0.6 => extremely negative\n* -0.3 => very negative\n* -0.1 => negative\n*    0 => neutral\n*  0.1 => positive\n*  0.3 => very positive\n*  0.6 => extremely positive\n\nThe below examples are all equivocal. You can use either set of values as the value for `sentiment` (numerical comparisons or natural langauge)\n\n* `>= 0.1` = `is positive`\n* `<= 0.3` = `is very negative`\n* `< 0.1` = `is not positive`\n* `> -0.3` = `is not very negative`\n\nSpecial case:\n\n* `is neutral` equates to `> -0.1 and < 0.1`\n* `is not neutral` equates to `< -0.1 or > 0.1`\n\nContextMod uses a normalized, weighted average from these sentiment tools:\n\n* NLP.js (english, french, german, and spanish) https://github.com/axa-group/nlp.js/blob/master/docs/v3/sentiment-analysis.md\n* (english only) vaderSentiment-js https://github.com/vaderSentiment/vaderSentiment-js/\n* (english only) wink-sentiment https://github.com/winkjs/wink-sentiment\n\nMore about the sentiment algorithms used:\n* VADER https://github.com/cjhutto/vaderSentiment\n* AFINN http://corpustext.com/reference/sentiment_afinn.html\n* Senticon https://ieeexplore.ieee.org/document/8721408\n* Pattern https://github.com/clips/pattern\n* wink https://github.com/winkjs/wink-sentiment",
                     "examples": [
                         "is negative",
                         "> 0.2"
@@ -5244,7 +5244,7 @@
                     "type": "string"
                 },
                 "sentiment": {
-                    "description": "Test the calculated VADER sentiment (compound) score for an Activity using this comparison. Can be either a numerical or natural language\n\nSentiment values range from extremely negative to extremely positive in a numerical range of -0.5 to 0.5:\n\n* -0.5 => extremely negative\n* -0.2 => very negative\n* -0.5 => negative\n*    0 => neutral\n* 0.05 => positive\n*  0.2 => very positive\n*  0.5 => extremely positive\n\nThe below examples are all equivocal. You can use either set of values as the value for `sentiment` (numerical comparisons or natural langauge)\n\n* `>= 0.05` = `is positive`\n* `<= 0.2` = `is very negative`\n* `< 0.05` = `is not positive`\n* `> -0.2` = `is not very negative`\n\nSpecial case:\n\n* `is neutral` equates to `> -0.5 and < 0.5`\n* `is not neutral` equates to `< -0.5 or > 0.5`\n\nMore about VADER Sentiment: https://github.com/cjhutto/vaderSentiment",
+                    "description": "Test the calculated VADER sentiment (compound) score for an Activity using this comparison. Can be either a numerical or natural language\n\nSentiment values range from extremely negative to extremely positive in a numerical range of -1 to +1:\n\n* -0.6 => extremely negative\n* -0.3 => very negative\n* -0.1 => negative\n*    0 => neutral\n*  0.1 => positive\n*  0.3 => very positive\n*  0.6 => extremely positive\n\nThe below examples are all equivocal. You can use either set of values as the value for `sentiment` (numerical comparisons or natural langauge)\n\n* `>= 0.1` = `is positive`\n* `<= 0.3` = `is very negative`\n* `< 0.1` = `is not positive`\n* `> -0.3` = `is not very negative`\n\nSpecial case:\n\n* `is neutral` equates to `> -0.1 and < 0.1`\n* `is not neutral` equates to `< -0.1 or > 0.1`\n\nContextMod uses a normalized, weighted average from these sentiment tools:\n\n* NLP.js (english, french, german, and spanish) https://github.com/axa-group/nlp.js/blob/master/docs/v3/sentiment-analysis.md\n* (english only) vaderSentiment-js https://github.com/vaderSentiment/vaderSentiment-js/\n* (english only) wink-sentiment https://github.com/winkjs/wink-sentiment\n\nMore about the sentiment algorithms used:\n* VADER https://github.com/cjhutto/vaderSentiment\n* AFINN http://corpustext.com/reference/sentiment_afinn.html\n* Senticon https://ieeexplore.ieee.org/document/8721408\n* Pattern https://github.com/clips/pattern\n* wink https://github.com/winkjs/wink-sentiment",
                     "examples": [
                         "is negative",
                         "> 0.2"

src/Schema/Rule.json

@@ -1309,7 +1309,7 @@
                     "type": "boolean"
                 },
                 "sentiment": {
-                    "description": "Test the calculated VADER sentiment (compound) score for an Activity using this comparison. Can be either a numerical or natural language\n\nSentiment values range from extremely negative to extremely positive in a numerical range of -0.5 to 0.5:\n\n* -0.5 => extremely negative\n* -0.2 => very negative\n* -0.5 => negative\n*    0 => neutral\n* 0.05 => positive\n*  0.2 => very positive\n*  0.5 => extremely positive\n\nThe below examples are all equivocal. You can use either set of values as the value for `sentiment` (numerical comparisons or natural langauge)\n\n* `>= 0.05` = `is positive`\n* `<= 0.2` = `is very negative`\n* `< 0.05` = `is not positive`\n* `> -0.2` = `is not very negative`\n\nSpecial case:\n\n* `is neutral` equates to `> -0.5 and < 0.5`\n* `is not neutral` equates to `< -0.5 or > 0.5`\n\nMore about VADER Sentiment: https://github.com/cjhutto/vaderSentiment",
+                    "description": "Test the calculated VADER sentiment (compound) score for an Activity using this comparison. Can be either a numerical or natural language\n\nSentiment values range from extremely negative to extremely positive in a numerical range of -1 to +1:\n\n* -0.6 => extremely negative\n* -0.3 => very negative\n* -0.1 => negative\n*    0 => neutral\n*  0.1 => positive\n*  0.3 => very positive\n*  0.6 => extremely positive\n\nThe below examples are all equivocal. You can use either set of values as the value for `sentiment` (numerical comparisons or natural langauge)\n\n* `>= 0.1` = `is positive`\n* `<= 0.3` = `is very negative`\n* `< 0.1` = `is not positive`\n* `> -0.3` = `is not very negative`\n\nSpecial case:\n\n* `is neutral` equates to `> -0.1 and < 0.1`\n* `is not neutral` equates to `< -0.1 or > 0.1`\n\nContextMod uses a normalized, weighted average from these sentiment tools:\n\n* NLP.js (english, french, german, and spanish) https://github.com/axa-group/nlp.js/blob/master/docs/v3/sentiment-analysis.md\n* (english only) vaderSentiment-js https://github.com/vaderSentiment/vaderSentiment-js/\n* (english only) wink-sentiment https://github.com/winkjs/wink-sentiment\n\nMore about the sentiment algorithms used:\n* VADER https://github.com/cjhutto/vaderSentiment\n* AFINN http://corpustext.com/reference/sentiment_afinn.html\n* Senticon https://ieeexplore.ieee.org/document/8721408\n* Pattern https://github.com/clips/pattern\n* wink https://github.com/winkjs/wink-sentiment",
                     "examples": [
                         "is negative",
                         "> 0.2"
@@ -2855,7 +2855,7 @@
                     "type": "string"
                 },
                 "sentiment": {
-                    "description": "Test the calculated VADER sentiment (compound) score for an Activity using this comparison. Can be either a numerical or natural language\n\nSentiment values range from extremely negative to extremely positive in a numerical range of -0.5 to 0.5:\n\n* -0.5 => extremely negative\n* -0.2 => very negative\n* -0.5 => negative\n*    0 => neutral\n* 0.05 => positive\n*  0.2 => very positive\n*  0.5 => extremely positive\n\nThe below examples are all equivocal. You can use either set of values as the value for `sentiment` (numerical comparisons or natural langauge)\n\n* `>= 0.05` = `is positive`\n* `<= 0.2` = `is very negative`\n* `< 0.05` = `is not positive`\n* `> -0.2` = `is not very negative`\n\nSpecial case:\n\n* `is neutral` equates to `> -0.5 and < 0.5`\n* `is not neutral` equates to `< -0.5 or > 0.5`\n\nMore about VADER Sentiment: https://github.com/cjhutto/vaderSentiment",
+                    "description": "Test the calculated VADER sentiment (compound) score for an Activity using this comparison. Can be either a numerical or natural language\n\nSentiment values range from extremely negative to extremely positive in a numerical range of -1 to +1:\n\n* -0.6 => extremely negative\n* -0.3 => very negative\n* -0.1 => negative\n*    0 => neutral\n*  0.1 => positive\n*  0.3 => very positive\n*  0.6 => extremely positive\n\nThe below examples are all equivocal. You can use either set of values as the value for `sentiment` (numerical comparisons or natural langauge)\n\n* `>= 0.1` = `is positive`\n* `<= 0.3` = `is very negative`\n* `< 0.1` = `is not positive`\n* `> -0.3` = `is not very negative`\n\nSpecial case:\n\n* `is neutral` equates to `> -0.1 and < 0.1`\n* `is not neutral` equates to `< -0.1 or > 0.1`\n\nContextMod uses a normalized, weighted average from these sentiment tools:\n\n* NLP.js (english, french, german, and spanish) https://github.com/axa-group/nlp.js/blob/master/docs/v3/sentiment-analysis.md\n* (english only) vaderSentiment-js https://github.com/vaderSentiment/vaderSentiment-js/\n* (english only) wink-sentiment https://github.com/winkjs/wink-sentiment\n\nMore about the sentiment algorithms used:\n* VADER https://github.com/cjhutto/vaderSentiment\n* AFINN http://corpustext.com/reference/sentiment_afinn.html\n* Senticon https://ieeexplore.ieee.org/document/8721408\n* Pattern https://github.com/clips/pattern\n* wink https://github.com/winkjs/wink-sentiment",
                     "examples": [
                         "is negative",
                         "> 0.2"

src/Schema/RuleSet.json

@@ -1280,7 +1280,7 @@
                     "type": "boolean"
                 },
                 "sentiment": {
-                    "description": "Test the calculated VADER sentiment (compound) score for an Activity using this comparison. Can be either a numerical or natural language\n\nSentiment values range from extremely negative to extremely positive in a numerical range of -0.5 to 0.5:\n\n* -0.5 => extremely negative\n* -0.2 => very negative\n* -0.5 => negative\n*    0 => neutral\n* 0.05 => positive\n*  0.2 => very positive\n*  0.5 => extremely positive\n\nThe below examples are all equivocal. You can use either set of values as the value for `sentiment` (numerical comparisons or natural langauge)\n\n* `>= 0.05` = `is positive`\n* `<= 0.2` = `is very negative`\n* `< 0.05` = `is not positive`\n* `> -0.2` = `is not very negative`\n\nSpecial case:\n\n* `is neutral` equates to `> -0.5 and < 0.5`\n* `is not neutral` equates to `< -0.5 or > 0.5`\n\nMore about VADER Sentiment: https://github.com/cjhutto/vaderSentiment",
+                    "description": "Test the calculated VADER sentiment (compound) score for an Activity using this comparison. Can be either a numerical or natural language\n\nSentiment values range from extremely negative to extremely positive in a numerical range of -1 to +1:\n\n* -0.6 => extremely negative\n* -0.3 => very negative\n* -0.1 => negative\n*    0 => neutral\n*  0.1 => positive\n*  0.3 => very positive\n*  0.6 => extremely positive\n\nThe below examples are all equivocal. You can use either set of values as the value for `sentiment` (numerical comparisons or natural langauge)\n\n* `>= 0.1` = `is positive`\n* `<= 0.3` = `is very negative`\n* `< 0.1` = `is not positive`\n* `> -0.3` = `is not very negative`\n\nSpecial case:\n\n* `is neutral` equates to `> -0.1 and < 0.1`\n* `is not neutral` equates to `< -0.1 or > 0.1`\n\nContextMod uses a normalized, weighted average from these sentiment tools:\n\n* NLP.js (english, french, german, and spanish) https://github.com/axa-group/nlp.js/blob/master/docs/v3/sentiment-analysis.md\n* (english only) vaderSentiment-js https://github.com/vaderSentiment/vaderSentiment-js/\n* (english only) wink-sentiment https://github.com/winkjs/wink-sentiment\n\nMore about the sentiment algorithms used:\n* VADER https://github.com/cjhutto/vaderSentiment\n* AFINN http://corpustext.com/reference/sentiment_afinn.html\n* Senticon https://ieeexplore.ieee.org/document/8721408\n* Pattern https://github.com/clips/pattern\n* wink https://github.com/winkjs/wink-sentiment",
                     "examples": [
                         "is negative",
                         "> 0.2"
@@ -2826,7 +2826,7 @@
                     "type": "string"
                 },
                 "sentiment": {
-                    "description": "Test the calculated VADER sentiment (compound) score for an Activity using this comparison. Can be either a numerical or natural language\n\nSentiment values range from extremely negative to extremely positive in a numerical range of -0.5 to 0.5:\n\n* -0.5 => extremely negative\n* -0.2 => very negative\n* -0.5 => negative\n*    0 => neutral\n* 0.05 => positive\n*  0.2 => very positive\n*  0.5 => extremely positive\n\nThe below examples are all equivocal. You can use either set of values as the value for `sentiment` (numerical comparisons or natural langauge)\n\n* `>= 0.05` = `is positive`\n* `<= 0.2` = `is very negative`\n* `< 0.05` = `is not positive`\n* `> -0.2` = `is not very negative`\n\nSpecial case:\n\n* `is neutral` equates to `> -0.5 and < 0.5`\n* `is not neutral` equates to `< -0.5 or > 0.5`\n\nMore about VADER Sentiment: https://github.com/cjhutto/vaderSentiment",
+                    "description": "Test the calculated VADER sentiment (compound) score for an Activity using this comparison. Can be either a numerical or natural language\n\nSentiment values range from extremely negative to extremely positive in a numerical range of -1 to +1:\n\n* -0.6 => extremely negative\n* -0.3 => very negative\n* -0.1 => negative\n*    0 => neutral\n*  0.1 => positive\n*  0.3 => very positive\n*  0.6 => extremely positive\n\nThe below examples are all equivocal. You can use either set of values as the value for `sentiment` (numerical comparisons or natural langauge)\n\n* `>= 0.1` = `is positive`\n* `<= 0.3` = `is very negative`\n* `< 0.1` = `is not positive`\n* `> -0.3` = `is not very negative`\n\nSpecial case:\n\n* `is neutral` equates to `> -0.1 and < 0.1`\n* `is not neutral` equates to `< -0.1 or > 0.1`\n\nContextMod uses a normalized, weighted average from these sentiment tools:\n\n* NLP.js (english, french, german, and spanish) https://github.com/axa-group/nlp.js/blob/master/docs/v3/sentiment-analysis.md\n* (english only) vaderSentiment-js https://github.com/vaderSentiment/vaderSentiment-js/\n* (english only) wink-sentiment https://github.com/winkjs/wink-sentiment\n\nMore about the sentiment algorithms used:\n* VADER https://github.com/cjhutto/vaderSentiment\n* AFINN http://corpustext.com/reference/sentiment_afinn.html\n* Senticon https://ieeexplore.ieee.org/document/8721408\n* Pattern https://github.com/clips/pattern\n* wink https://github.com/winkjs/wink-sentiment",
                     "examples": [
                         "is negative",
                         "> 0.2"

src/util.ts

@@ -2757,3 +2757,11 @@ export const filterByTimeRequirement = (satisfiedEndtime: Dayjs, listSlice: Snoo
 
     return [truncatedItems.length !== listSlice.length, truncatedItems]
 }
+
+export const between = (val: number, a: number, b: number, inclusive: boolean): boolean => {
+    var min = Math.min(a, b),
+        max = Math.max(a, b);
+
+    return inclusive ? val >= min && val <= max : val > min && val < max;
+
+}