31:[] 2f:{"title":"Compromise-Making-a-bot","cells":[{"id":0,"value":"md`# Compromise-Making-a-bot\n\n[compromise](http://compromise.cool) is a great way to make a simple *rule-based* chat-bot, where you can create easily-readable templates for things to respond to, without getting caught on things like punctuation or contractions, or messed-up regular expressions.\n`","pinned":false,"hidden":false,"mode":"ojs"},{"id":14,"value":"html`

JSConf Asia

\n \n \n \n

\n`","pinned":false,"hidden":false,"mode":"ojs"},{"id":83,"value":"md`Briefly:`","pinned":false,"hidden":false,"mode":"ojs"},{"id":82,"value":"{\n //'simon says' chat-bot\n function myBot(input){\n let doc = nlp(input)\n //look for known-templates\n if(doc.has('^simon says .')){\n let todo = doc.after('^simon says')\n let action = todo.verbs(0).normalize() //doing what?\n return `OK, ${action.conjugate()[0].Gerund} now` //'read' -> 'reading'\n } else {\n return 'doing nothing.'\n }\n }\n return myBot('simon says shoot the puck') // 'shooting'\n //return myBot('shoot the puck') // 'do nothing'\n}","pinned":true,"hidden":false,"mode":"ojs"},{"id":119,"value":"md`or here is a compromise-version of the amazing [ELIZA bot](https://en.wikipedia.org/wiki/ELIZA), by Joseph Weizenbaum:`","pinned":false,"hidden":false,"mode":"ojs"},{"id":124,"value":"{\n function elizaBot(input){\n let doc = nlp(input)\n //our canned-templates\n if(doc.has('i #Adverb? (am|feel) #Adverb? #Adjective')){\n let feeling = doc.match('i #Adverb? am #Adverb? [#Adjective]').out('normal')\n return `When did you become ${feeling}?`\n } else if(doc.has('(father|mother|dad|mom)')){\n let whichParent = doc.match('(father|mother|dad|mom)').out('normal')\n return `Why don't you tell me about your ${whichParent}.`\n } else {\n return 'can you elaborate on that?'\n }\n }\n // return elizaBot(`well, i'm kinda upset about it.`)\n return elizaBot(`well, i don't really talk with my dad much anymore...`)\n}","pinned":true,"hidden":false,"mode":"ojs"},{"id":23,"value":"md`\nyou can push things further, by ensuring the compromise tagger knows about the things you want it to, by combining it with the [plugin scheme](https://beta.observablehq.com/@spencermountain/compromise-plugins):`","pinned":false,"hidden":false,"mode":"ojs"},{"id":5,"value":"{\n const myPlugin = {\n //add words to the internal lexicon\n words:{\n 'facebook': 'Software',\n 'google': 'Software',\n 'salesforce': 'Software'\n },\n compute: {\n loginTypes:(doc)=>{\n doc.match(\"how do? i (can|should)? (log|sign|get) in to my? #Software\").tag('LoginIssue')\n doc.match(\"i can't (log|sign|get) in to my? #Software\").tag( 'LoginIssue')\n doc.match(\"#Software won't let me (log|sign|get) in\").tag('LoginIssue')\n }\n }\n }\n nlp.plugin(myPlugin)\n\n //ok, now this should be tagged as a LoginIssue.\n let doc=nlp('umm, do you know how I should sign-in to my salesforce account?')\n\n //respond appropriately\n if(doc.has('#LoginIssue')){\n return 'OK, press the Log-in button.'\n } else {\n return 'talk to an operator.'\n }\n}","pinned":true,"hidden":false,"mode":"ojs"},{"id":31,"value":"md`if you want to make a stateful, conversational bot, maybe it would look something like this:`","pinned":false,"hidden":false,"mode":"ojs"},{"id":33,"value":"{\n let state = {}\n \n function doReply(input){\n //parse the input\n let doc = nlp(input, {hi:'Greeting', hey:'Greeting', yo:'Greeting'})\n \n //build-up our state object from the parsed text\n let name = doc.match(`i'm [#Person+]`)\n //or try these versions of it:\n name = name.found ? name : doc.match('my name is [#Person+]')\n name = name.found ? name : doc.match('they call me [#Person+]')\n\n if(name.found === true){\n state.name = name.normalize().toTitleCase().out('text')\n } \n //begin input-based branching..\n if(doc.has('#Greeting')){\n //some state-based branching..\n if(!state.name){\n return `Hi! What's your name?`\n }else{\n return `oh hey ${state.name}`\n }\n }\n }\n return doReply(`oh hey, I'm maurice. They call me a space cowboy.`)\n //return doReply(`hello, is this thing on?`) // \"Hi! What's your name?\"\n}","pinned":true,"hidden":false,"mode":"ojs"},{"id":130,"value":"md`this can get unweildly pretty fast. For making a complex, stateful bot, i recommend using a [bot-framework](https://www.marutitech.com/complete-guide-bot-frameworks/), like **[superscript](https://github.com/superscriptjs/superscript/blob/82e46877931f28e75eec63b034209c877b7f1f5b/example/eliza.ss)** (which uses compromise under the hood!)\n\nlike, just check out the superscript version of the **ELIZA chatbot**,\n\n[\\`./eliza.ss\\`](https://github.com/superscriptjs/superscript/blob/82e46877931f28e75eec63b034209c877b7f1f5b/example/eliza.ss):\n\n> > + [*] (sorry|apologize|apology) [*]\n> > - Please don't apologize.\n> > - Apologies are not necessary.\n\n> > + [*] perhaps [*]\n> > - You don't seem quite certain.\n> > - Why the uncertain tone?\n\nyeah, right? that's gonna be a lot better of a bot in the long-run, without the million ***'if' statements*** .\n >> see [superscriptjs](https://github.com/superscriptjs/superscript)\n\n`","pinned":false,"hidden":false,"mode":"ojs"},{"id":10,"value":"nlp = require('compromise')","pinned":false,"hidden":false,"mode":"ojs"},{"id":244,"value":"nlp.version","pinned":false,"hidden":false,"mode":"ojs"}],"files":"$W31","stdlib":"1","titling":"auto"}