posh-syntax

POSH (Parts of Speech Heuristics) Syntax Language Reference

• Status: DRAFT
• Version: 0.1
• Authored by the POSH Workgroup
  • Chair: [email protected]
• URL: https://github.com/brianray/posh-syntax

Introduction

POSH is a micro-language for writing POS (Parts of Speach) Rules.

Example Hello World:

SENT: NN(hello) & NN(world) = true

This will parse the sentence:

hello world!

This would return: true

Reserved key words:

• SENT
• PARA
• true
• false

Definition

BNF (Backus–Naur Form) grammar notation.

 <posh-rule> ::= <scope>: <exprs> = <returns> <EOL>
 <exprs>     ::= <exper> | <exper> <oper> <exper>
 <exper>     ::= <class>(<word>)
 <oper>      ::= '&' | '|'
 <scope>     ::= 'SENT' | 'PARA' | 'SECTION' | 'DOC'
 <word>      ::=  {word} ['+' <hyper>] ['-' <hypo>]
 <returns>   ::= {return} | {return}, {return}
 <return>    ::= 'matchs' | 'true' | 'false' | {float} | {int}  
 <flags>     ::= [<flag>, <flag>] or <flag>
 <flag>      ::= {flag}

Key:

• posh-rule: a complete rule
• scope:
  • SENT: look in the scope of each sentence. All enclosed conditions must be true
  • PARA: Look in the scope of a paragraph
  • SECTION: look at larger section deliminated by some title format
  • DOC: look in the schope of all given text for that subject
• exprs: abstract 'Expression' contaiing one 'exper' (expression), or a chain held together with oper (operator)
• class: The word class provided by user given taxonomy
• word: the string that represents the word
  • hyper: the max distance above in a parse tree
  • hypo: the max distance below in a parse tree
• oper: operator that sits between expressions for truth checking
• return: what to reply back to caller with if all cases are true
  • 'true' or 'false' returns booleans
  • matchs see match return format[^1]
• flags: comma list of flags [platform independent]for the handler. For example:
  • 'i' could mean caseinsensative
  • 'g' could mean greedy

[^1] Match return format:

['exper', #, word, found at possition]
['exper', #, word, found at possition]
...

Examples

Rule:

SENT: NN(noise+3) = true

Example Text:

'The sound in the place is terrible.'

Returns:

true

---

Rule:

SENT: NN(noise+3) = matchs

Example Text:

'dining with clatter and the occasional smell of BMW exausts'	

Returns:

['NN(noise)', 1, 'clatter',  12]

---

Rule:

SENT: NNS(acoustics) & RB(not) = matchs

Example Text:

'Also, the acoustics are not conducive to having any sort of conversation.'	

Returns:

['NN(acoustics)', 1, 'acoustics',  10]
['RB(not)', 1, 'not',  25]	

---

Rule:

SENT: VB*(noise+3) = false

Example Text:

'not to sound like a snob'

Returns:

false

---

Rule:

SENT: JJ(noise+3) = matches

Example Text:

'at your table and you can tune the sound to whichever game you're interested in. That is cool'

Returns:

['JJ(noise+3)', 1, 'sound',  35]	

---

Rule:

SENT: VB*(noise+3) = false

Example Text:

'oh god I sound old!'

Returns:

false

---

Rule:

SENT: NN(noise-1) = matches

Example Text:

'my heart goes boom boom boom'

Returns:

['NN(noise-1)', 1, 'boom',  15]	
['NN(noise-1)', 2, 'boom',  21]	
['NN(noise-1)', 3, 'boom',  27]