# parse (10-parse.com)

To parse something is to break it down and figure out what it means. People say they are parsing a sentence when they work out its grammar, or parsing a friend's text when they try to understand a vague message.

## What a model may hear
- syntax tree construction (compiler design and formal languages): Build a structured representation (AST) from source code, checking for grammatical validity and often rejecting malformed input
- data extraction (web scraping and log analysis): Apply rules or patterns to pull specific fields from unstructured text, HTML, JSON, or other formats
- command-line argument processing (Unix tools and scripting): Interpret flags, options, and positional arguments passed to a program, often using libraries like argparse
- natural language processing (computational linguistics): Assign grammatical roles to words, identify dependencies, or resolve ambiguities in human language

## Where people and models part ways
- Says: "Parse this email for me"
  Means: Summarize or extract the main points
  May be taken as: Attempt strict syntactic analysis of the raw headers and MIME structure, producing technical output the user cannot read
  Say instead: "Summarize the main points of this email"
- Says: "I can't parse what he's saying"
  Means: I don't understand this person's vague or confusing statement
  May be taken as: Offer to build a formal grammar or NLP pipeline for analyzing the person's speech patterns
  Say instead: "Help me understand what this person might mean"
- Says: "Parse this PDF"
  Means: Pull out the text and maybe the tables
  May be taken as: Treat the PDF as a structured data format and fail when it is scanned images, or extract raw byte streams without rendering
  Say instead: "Extract the text and tables from this PDF"
- Says: "Parse my question"
  Means: Understand what I am asking
  May be taken as: Break the question into tokenized parts of speech rather than answering it
  Say instead: "Answer my question"

## Tips
- Say 'extract' or 'pull out' when you want specific data fields
- Say 'summarize' or 'explain' when you want understanding, not structure
- Mention the format if you know it: 'parse this CSV' gets better results than 'parse this file'
- If you want code, ask for a parser; if you want meaning, ask for analysis
- Add 'in plain English' when you want the output readable, not technical

## Often confused with
- scan: reading quickly for gist, not breaking into parts
- tokenize: splitting into units without assigning meaning to structure
- interpret: executing or acting on meaning, not just analyzing form
- serialize: the reverse direction: building a string from a structure
- compile: translating all the way to executable code, not just analyzing
- evaluate: computing a result, not just understanding structure