Integrating OpenNLP in a Spring Boot Application

To incorporate Apache OpenNLP into a Maven-based Spring Boot application, add the following dependency in your pom.xml file:

<dependency>

<groupId>org.apache.opennlp</groupId>

<artifactId>opennlp-tools</artifactId>

<version>2.0.0</version>

</dependency>

Case 1: Tokenization

Tokenization represents one of the foundational steps in natural language processing. This technique divides text into smaller components, which helps in understanding the context of the content. Apache OpenNLP offers an API through the SimpleTokenizer class, which can be used as follows:

SimpleTokenizer tokenizer = SimpleTokenizer.INSTANCE;

String tokens[] = tokenizer.tokenize("Your Service is bad");

The resulting tokens from this code snippet would be [Your, Service, is, bad].

Case 2: Part of Speech Tagging

For ML algorithms to function effectively, reliable training data is essential. This use case employs a pre-trained model known as en-pos-maxent.bin. Through Part of Speech (POS) Tagging, we can identify various parts of speech such as nouns, verbs, and adjectives in a given text. This method is particularly useful in applications like sentiment analysis, where specific parts of speech are analyzed further.

Here’s a simple Java method to extract ADVERB, ADJECTIVE, and VERB from a text:

// Example method to extract parts of speech

If we analyze the text "Your Service is bad" with this method, the POS Tagger would label the tokens as:

Your_PRP$ Service_NN is_VBZ bad_JJ

Case 3: Sentiment Analysis / Document Categorization

The Sentiment Analyzer or Document Categorizer can be utilized to evaluate sentiments in various texts, such as product reviews or tweets. Unlike other use cases, this one requires users to generate their own training data, as Apache OpenNLP does not supply a pre-trained model for this purpose.

To create a training model, you’ll need to compile a file with training data and save it with a .txt extension. For example, the data might categorize sentiments as Angry or Neutral. Any input processed against this model will be classified into one of these categories or default to the first category (Neutral) if it can't be decisively categorized.

After preparing the training data, you will generate a model file with a .bin extension using the following Java program:

// Example program to create a training model

Next, we utilize the created model to categorize an input:

File file = ResourceUtils.getFile("src/main/resources/nlp-model/en-trained-model.bin");

InputStream in = new FileInputStream(file);

DoccatModel m = new DoccatModel(in);

DocumentCategorizerME myCategorizer = new DocumentCategorizerME(m);

String[] inputText = input.split(" ");

double[] outcomes = myCategorizer.categorize(inputText);

String category = myCategorizer.getBestCategory(outcomes);

For an input text like “I will not recommend this product,” the output category would be “Angry.”