The Homepage of NLQ.py

(Explanation) (Definition) (Examples) (Download) (Links)

---

[2001-04-15] The current NLQ version is 0.1 - Beware!

---

Explanation

NLQ.py is a quick and dirty (with an emphasis on both) python module that attempts to parse an incoming "Natural Language" query, like "How many keywords does Python have?" and "Why do I get this rash when I pet my cat?" I needed something like this for a project that I was working on and this is what I came up with in the course of an evening.

NLQ.py is implemented as a Python class (class NLQ). For every NLQ you have, you declare an NLQ object and give it the string. The object takes the string, chops it up and attempts classify it and extract the relevant pieces. This is all gone over more in the class definition.

NLQ.py is pre-pre-pre-alpha. I have no AI experience. I have no relevant language experience. The design of this may change at any time. The woman that you think is your mother is really a being from another planet placed there to raise you until you're old enough to have your True Purpose revealed to you. Pro wrestling is real.

---

Definition

The NLQ.py module

...contains a handful of definitions.

• The question types: UNKNOWN, KNOWLEDGE, COMPREHENSION, APPLICATION, ANALYSIS, SYNTHESIS, and EVALUATION. These are lists of words that correspond to the type of question that is being asked. For now, you'll have to be satisfied with the explanations given on this page.
• There are also small lists of words that the program will ignore. "The", "but", "to", etc. A necessary evil. There are about 50 right now. I forsee it becoming anywhere from 100-250 words long eventually.

...and one function - determine_type. D_t is used by the NLQ objects to assess the string and fit it into one of the pigeon-hole types defined above. Absurdly simplistic right now...

---

Class NLQ...

• ...takes 1 argument - a string hopefully containing a query written as a regular question.
• ...has 2 attributes: type and keywords.
  • type indicates what type of query it is, according to the question types defined above. (KNOWLEDGE, COMPREHENSION, etc...)
  • keywords is a list containing what NLQ.py has determined to be the important pieces of the submitted NLQ. These you would use in your application. For instance, if the nlq was "What is the capitol of Vermont?", the keyword list would look like this: ["capitol", "vermont"].
• ...has 0 methods, though print (__repr__) is redefined to print out the keywords and the type.

---

Download

• HTTP
  1. http://gurno.com/adam/dl

---

Examples

abg@stegosaurus ~/src/nlq $ ./nlq.py "Who is the greatest baseball player?"
type: 1
keywords: ['greatest', 'baseball', 'player']
abg@stegosaurus ~/src/nlq $ ./nlq.py "Why should I file my taxes?"
type: 3
keywords: ['should', 'file', 'taxes']
abg@stegosaurus ~/src/nlq $ ./nlq.py "Should I dye my hair pink?"
type: 0
keywords: ['dye', 'hair', 'pink']

---

Links

There are only about a zillion AI/NL links out there. These are only a few. For more than you can shake a stick at, try searching for "natural language query" on google or yahoo.

• NL Processing in Information Retrieval - A wordy guide to the problems parsing NL
• A press release from the Evil Empire
• AI and Python - including an ELIZA clone.

---