nlp-tools/nlp-tools

NlpTools is a set of php 5.3+ classes for beginner to semi advanced natural language processing work.

v0.1.3 2016-11-03 15:34 UTC

Requires

• php: >=5.3

WTFPL

• Angelos Katharopoulos

• nlp
• machine learning

README

NlpTools is a set of php 5.3+ classes for beginner to semi advanced natural language processing work.

Documentation

You can find documentation and code examples at the project's homepage.

Contents

Classification Models

1. Multinomial Naive Bayes
2. Maximum Entropy (Conditional Exponential model)

Topic Modeling

Lda is still experimental and quite slow but it works. See an example.

1. Latent Dirichlet Allocation

Clustering

1. K-Means
2. Hierarchical Agglomerative Clustering
   • SingleLink
   • CompleteLink
   • GroupAverage

Tokenizers

1. WhitespaceTokenizer
2. WhitespaceAndPunctuationTokenizer
3. PennTreebankTokenizer
4. RegexTokenizer
5. ClassifierBasedTokenizer This tokenizer allows us to build a lot more complex tokenizers than the previous ones

Documents

1. TokensDocument represents a bag of words model for a document.
2. WordDocument represents a single word with the context of a larger document.
3. TrainingDocument represents a document whose class is known.
4. TrainingSet a collection of TrainingDocuments

Feature factories

1. FunctionFeatures Allows the creation of a feature factory from a number of callables
2. DataAsFeatures Simply return the data as features.

Similarity

1. Jaccard Index
2. Cosine similarity
3. Simhash
4. Euclidean
5. HammingDistance

Stemmers

1. PorterStemmer
2. RegexStemmer
3. LancasterStemmer
4. GreekStemmer

Optimizers (MaxEnt only)

1. A gradient descent optimizer (written in php) for educational use. It is a simple implementation for anyone wanting to know a bit more about either GD or MaxEnt models
2. A fast (faster than nltk-scipy), parallel gradient descent optimizer written in Go. This optimizer resides in another repo, it is used via the external optimizer. TODO: At least write a readme for the optimizer written in Go.

Other

1. Idf Inverse document frequency
2. Stop words
3. Language based normalizers
4. Classifier based transformation for creating flexible preprocessing pipelines