LanguageModel

lm. LanguageModel

Class to interact with the remote Language Model server. Builds and maintains a forward-looking cache of all word predictions and character probabilities that are possible after one future selection. Formats the word predictions and character probabilities so that they are usable in Nomon's inference module.

Constructor

new LanguageModel(parent)

Source:
Properties:
Name Type Description
word_predict_base_url string

The api address for querying word predictions for a given left context and prefix.
word_predict_char_future_url string

Used for forward caching. The api address for querying word predictions for a given left context and prefix, for each possible next character that could be added to the prefix.
word_predict_word_future_url string

Used for forward caching. The api address for querying word predictions for a given left context and prefix, for each possible next word prediction that could be added to the prefix.
char_predict_base_url string

The api address for querying character probabilities for a given left context and prefix.
char_predict_char_future_url string

Used for forward caching. The api address for querying character probabilities for a given left context and prefix, for each possible next character that could be added to the prefix.
char_predict_word_future_url string

Used for forward caching. The api address for querying character probabilities for a given left context and prefix, for each possible next word prediction that could be added to the prefix.
lm_prefix string

The prefix for the word currently being typed. For example: If the user has typed "I am using Nomo", the lm_prefix would be "Nomo".
left_context string

The left-context for the word currently being typed. For example: If the user has typed "I am using Nomo", the lm_prefix would be "I am using ".
word_cache Object.<Object.<Object>>

A nested Object maintaining a cache of all possible future word and character probabilities that are one selection away. First level keys are the letters a-z, punctuation, current word predictions, and corrective options:

Properties
Name Type Description
a Object.<Object>

An object containing the word predictions and character probabilities if the next typed character is 'a'

Properties
Name Type Description
keys Array.<number>

An array containing the probabilities of the next characters (following 'a')
word Object.<Array>

An object containing the word predictions and their associated probabilities (following 'a')

Properties
Name Type Description
probs Array.<number>

An array containing the probabilities of the next word predictions (following 'a')
word Object.<string>

An array containing the strings of the next word predictions (following 'a')
b Object.<Object>

All entries follow the schema above...
z Object.<Object>
word_1 Object.<Object>
word_2 Object.<Object>
word_n Object.<Object>
Undo Object.<Object>
word_predictions Array.<Array.<string>>

2D Array containing the word predictions for the currently typed left-contect and prefix. Of the form: [[word_a1, word_a2, word_a3], [word_b1, word_b2, word_b3], ...]
word_prediction_probs Array.<Array.<number>>

2D Array containing the probabilities of the current word predictions.
Parameters:
Name Type Description
parent Keyboard

The instance of the main Keyboard class.

Methods

construct_url(base, parameters) → {String}

Description:

  • Helper function to add the parameters in an object to a URL string.

Source:
Parameters:
Name Type Description
base String

The base URL.
parameters Object

An object with parameter keys and values to add to the base URL string.
Returns:
  • Returns a new URL string with the added parameters.
Type
String

foramt_words(words_li, temp_prefix) → {Object}

Description:

  • Helper function. Formats the word predictions and probabilities into Arrays useable by the Nomon inference module.

Source:
Parameters:
Name Type Description
words_li Array.<String>

An array of word predictions from the language model.
temp_prefix String

The prefix to store the words and probabilities under in the word cache.
Returns:
  • Returns a formatted Object with the word predictions and probabilities.
Type
Object

format_chars(chars_li) → {Array.<Number>}

Description:

  • Helper function. Formats the probabilities of the main characters into an Array useable by the Nomon inference module.

Source:
Parameters:
Name Type Description
chars_li Array.<String>

An Array of the possible main characters (kconfig.main_chars).
Returns:
  • The probabilities of the main characters from the language model.
Type
Array.<Number>

on_cor_load_function(output, cache_type)

Description:

  • Runs when an API query Promise is completed. Adds the results to their corresponding place in the future word cache.

Source:
Parameters:
Name Type Description
output JSON

The data returned from the API call.
cache_type String

The type of API query submitted: "word_base", "char_base", "word_future_char", "word_future_word", "char_future_char", or "char_future_word"

update_cache(left, prefix, selection)

Description:

  • Update the future word cache given the newest selection in the keyboard. Creates Promises to complete when the API returns results.

Source:
Parameters:
Name Type Default Description
left String

The current left context from the keyboard.
prefix String

The current prefix from the keyboard.
selection String | null null

The currently selected option that triggered the cache update (null for cache init)