Maiden-Markov

About

This provides a fast and compact Markov chains implementation. Markov chains are stored efficiently in memory and on disk, allowing for huge dictionaries without loss of resources or performance.

It also implements its own binary format in order to save the dictionary fast and compact.

How To

Create a generator instance, and train it with learn or learn-from-file

(defvar *g* (make-instance 'generator))
(learn "Hey this is a sentence you can learn." *g*)
(learn-from-file "~/king-james.txt" *g*)
(learn-from-file "~/sicp.txt" *g*)
(make-sentence *g*)
(find-sentence *g* "lambda")

You can also make it learn from a chat channel by activating the module on that channel. See maiden-activatable and the commands in the symbol index.

Copyright

maiden-markov is licensed under the Artistic license. © Nicolas Hafner <shinmera@tymoon.eu> . This library can be obtained on https://github.com/Shinmera/maiden.

Package Index

MAIDEN-MARKOV (ORG.SHIRAKUMO.MAIDEN.AGENTS.MARKOV)

class

`GENERATOR`

This contains all the state to allow you to generate new sentences using Markov chains.

The chains are stored in an efficient indexed format
in order to take up as little memory as possible.

See WORDS
See WORD-MAP
See CHAINS
See VISUALIZE
See WORD
See WORD-INDEX
See CHAIN
See ENSURE-CHAIN
See ADD-CHAIN
See NEXT-WORD-INDEX
See RANDOM-TOKEN
See MAKE-SENTENCE
See FIND-SENTENCE
See TOKEN-SCORE
See FIND-TOPIC
See LEARN-SENTENCE
See LEARN
See LEARN-FROM-FILE

class

`MARKOV`

This agent implements a markov-chain based chat bot for amusement. It can learn from the actual chat messages of a channel.

class
RAMBLE
```
 
```
class
RAMBLE-ABOUT
```
 
```
class
RAMBLE-CHANCE
```
 
```
class
SET-RAMBLE-CHANCE
```
 
```
class
STATS
```
 
```

accessor (

`CHAIN`

GENERATOR FIRST SECOND)

This accesses a vector of possible follow-up words for the given first and second words.

The chains tables are automatically extended as necessary.

See WORD-INDEX
See CHAINS
See GENERATOR

accessor (

`RAMBLE-CHANCE`

C)

Accessor to the chance (in percent) that the bot might reply to a message with random rambling.

See MARKOV
See MAIDEN-STORAGE:WITH-STORAGE

accessor (

`WORD`

THING GENERATOR)

Accessor to a word in the generator.

THING can be a number, symbol, or string, and returned is
the word's index.
If you set this place with a string, the string is added
to the generator and gets its own index, which is returned.

See GENERATOR
See WORD-MAP
See WORDS

function (

`ADD-CHAIN`

GENERATOR FIRST SECOND &REST POSSIBILITIES)

Add a new chain for the given first and second words and follow-up possibilities.

See ENSURE-CHAIN
See WORD-INDEX
See GENERATOR

function (

`ENSURE-CHAIN`

GENERATOR FIRST SECOND)

This ensures that a chain for the given first and second words exists.

See CHAIN
See GENERATOR

function (

`FIND-SENTENCE`

GENERATOR WORDS &OPTIONAL START (MAX-ATTEMPTS 10000))

Attempt to find a sentence that somehow relates to the words given.

This proceeds by generating completely random sentences
until one is found that contains one of the given words.
Since this is not always guaranteed to work --luck may not
be on your side-- it might also return NIL even though it
is theoretically possible to build a chain. If nothing has
been found after MAX-ATTEMPTS generations, it simply gives
up.

See MAKE-SENTENCE
See GENERATOR

function (

`LEARN`

STRING GENERATOR)

Learn the string.

The string may contain multiple sentences at once.

See FIND-SENTENCES
See LEARN-SENTENCE

function (

`LEARN-FROM-FILE`

FILE GENERATOR)

This reads a file line-by-line and feeds each line to LEARN.

See LEARN

function (

`LEARN-SENTENCE`

SENTENCE GENERATOR)

Learn the sentence.

This splits the sentence into tokens and builds Markov
chains from the tokens, thus "learning" the sentence.

See FIND-TOKENS
See ADD-CHAIN

function (

`MAKE-SENTENCE`

GENERATOR &OPTIONAL START)

Build a new sentence generated by markov chains in the generator.

START can be a token to begin the chain with. If not
given, a random starting token is picked.
Chains are generated by using two words, selecting
the chain the two words designate, and picking a random
follow-up word from that chain. The second and new word
then build the input for the next iteration until an END
token is encountered as the next token.

See *START*
See *END*
See WORD-INDEX
See NEXT-WORD-INDEX
See WORD
See RANDOM-TOKEN
See GENERATOR

function (

`NEXT-WORD-INDEX`

GENERATOR FIRST SECOND)

Return the index of a follow-up word for the given first and second words.

This index is chosen randomly from the possible indexes
of follow-up words in the corresponding chain. If no
follow-up can be found, NIL is returned.

See CHAIN
See GENERATOR

function (

`RANDOM-TOKEN`

GENERATOR)

Find a random starting token to start a chain with.

This searches through all possible starting chains in
a hopefully somewhat random manner.

See CHAINS
See GENERATOR

function (

`READ-GENERATOR`

SOURCE)

Reads a complete Markov generator instance from the source.

Source may be a pathname-designator, a stream, or a
fast-io::input-buffer.

function (

`WORD-INDEX`

GENERATOR WORD)

Return the index of the word.

This adds the word to the generator if it doesn't exist in
its dictionary yet.

See WORD
See GENERATOR

function (

`WRITE-GENERATOR`

GENERATOR TARGET)

Writes a complete Markov generator instance to the source.

Source maybe be a pathname-designator, a stream, or a
fast-io::output-buffer.

generic (

`GENERATOR`

MARKOV)

Returns the generator of the markov client.

If the client does not have a generator yet, a new one is
created and set into the client's slit.

See MARKOV
See GENERATOR

command
ramble
```
Generates a random sentence.
```

command (

`ramble about`

&OPTIONAL TOPIC)

Generates a random sentence that should be related to the given topic.

command

`ramble chance`

Display the chance [0,1] for the bot to reply to a message.

command (

`set ramble chance`

NEW-VALUE)

Update the chance [0,1] for the bot to reply to a message.

command

`markov stats`

Display some statistics about the dictionary of markov chains.

maiden-markov

About

How To

Copyright

Package Index

MAIDEN-MARKOV (ORG.SHIRAKUMO.MAIDEN.AGENTS.MARKOV)

`GENERATOR`

`MARKOV`

`RAMBLE`

`RAMBLE-ABOUT`

`RAMBLE-CHANCE`

`SET-RAMBLE-CHANCE`

`STATS`

`CHAIN`

`RAMBLE-CHANCE`

`WORD`

`ADD-CHAIN`

`ENSURE-CHAIN`

`FIND-SENTENCE`

`LEARN`

`LEARN-FROM-FILE`

`LEARN-SENTENCE`

`MAKE-SENTENCE`

`NEXT-WORD-INDEX`

`RANDOM-TOKEN`

`READ-GENERATOR`

`WORD-INDEX`

`WRITE-GENERATOR`

`GENERATOR`

`ramble`

`ramble about`

`ramble chance`

`set ramble chance`

`markov stats`