(MAVEN-Project)
SLaMowJ is a simple n-gram language model framework. With SLaMowJ a user has the possibility to create a database which contains words and their possible (n-1) following words. (For more information about n-gram model see: https://en.wikipedia.org/wiki/N-gram)
- How to use
Before using the language model, you should train the database with text files.
English/German languages are good for training of the database. But you can use any language you want, make sure that every sentence ends with one of these characters " !, ?, , , . ". Every word within the text has to be seperated by a whitespace.
- Train a database
For training the database you have to use the Trainer.java Utility-Class:
Trainer.trainStandardEnglishMode("PATH_TO_DATABASE_WITHOUT_SUFFIX", TextReader.readFromFile("train.set"), TrainerMode.WORD, Trainer.BIGRAM, true);
With this short code, you can train a database with the content of the "train.set" file.
The created database is located within the database folder.
You can use three TrainerMode's:
- SENTENCE
- WORD
- CHAR
For displaying some informations while trainig the database call Trainer.initDebugger();
While training, the following (n-1)-grams of a word are always sorted by its count of occurrence. That means the result you will get by calling the following methods, are always sorted by the most possible word.
- Using the database
After the training you can use the interface LanguageModel to access the trained database.
For instantiating a LanguageModel-Object you have to use this following code:
LanguageModel lm = LanguageModelBuilder.createInstance("PATH_TO_DATABASE_WITHOUT_SUFFIX");
after this line of code you can access the database using the two methods:
public String[] getGramFromModel(String firstWord);
The method String[] getGramFromModel(String first Word); retrieves if exists the
the following (n-1)- words of the given firstWord, otherwise the method returns null.
public String[] getGramsOfGrams(String firstWord, int maxWidth, int maxDepth);
This method returns all possible word combinations from the given word and its maxDepth and maxWidth followers. Take care by using this method with high numbers for maxDepth and/or maxWidth!
Make sure to close the connection to the database by using this method: closeLanguageModel()
- Example
An example code can be found within the main folder.