The LegalNLP library is a powerful tool for anyone working with the Brazilian legal language. This user-friendly guide will walk you through the installation process, how to access and utilize the language models, and provide troubleshooting tips for common issues.

1. Accessing the Language Models

All the language models you need can be found here. If you encounter any trouble accessing the models, feel free to contact felipemaiapolo@gmail.com.

2. Installing the Package

Before diving into the specific models, the first step is to install the required libraries. LegalNLP is designed for Python, a popular programming language for machine learning.

Run the following command in your terminal to install the HuggingFaceHub library:

pip install huggingface_hub

Next, import the necessary module:

from huggingface_hub import hf_hub_download

Then, you can download the Word2Vec and Doc2Vec models using these commands:

w2v_sg_d2v_dbow = hf_hub_download(repo_id='ProjetoLegalNLP', filename='w2v_d2v_dbow_size_100_window_15_epochs_20')\n
w2v_cbow_d2v_dm = hf_hub_download(repo_id='ProjetoLegalNLP', filename='w2v_d2v_dm_size_100_window_15_epochs_20')

3. Understanding Word2Vec and Doc2Vec Models

The Word2Vec and Doc2Vec models are like a sophisticated recipe book that transforms text into numerical data, much like cooking ingredients into a delicious dish. Here’s a breakdown:

• Word2Vec: Think of it as a chef who understands the significance of each ingredient based on its context. It creates a representation of words (tokens) by looking at the context in which they appear. This method captures the essence and meaning of words in a multidimensional space.
• Doc2Vec: Now imagine our chef not only recognizing ingredients but also understanding entire dishes. Doc2Vec extends Word2Vec, providing representations of full sentences or texts, allowing it to grasp the bigger picture or context of the content.

Both models require Gensim version 3.8.3 for optimal performance. Here’s how to install it:

!pip install gensim==3.8.3

4. Using Word2Vec and Doc2Vec

Now let’s load and utilize the models. Below are the steps for using both models:

Using Word2Vec

from gensim.models import KeyedVectors\n\n# Loading a W2V model\nw2v = KeyedVectors.load(w2v_cbow_d2v_dm)\nw2v = w2v.wv\n\n# Viewing the first 10 entries of the "juiz" vector\nprint(w2v['juiz'][:10])

This command allows you to view the vector representation of the word “juiz”. You can also find the closest tokens to “juiz” by using the following command:

print(w2v.most_similar('juiz'))

Using Doc2Vec

from gensim.models import Doc2Vec\n\n# Loading a D2V model\nd2v = Doc2Vec.load(w2v_cbow_d2v_dm)\n\n# Inferring vector for a text\ntext = "direito do consumidor origem : bangu regional xxix juizado especial civel ação : [processo] - - recte : fundo de investimento em direitos creditórios"\ntokens = text.split()\ntxt_vec = d2v.infer_vector(tokens, epochs=20)\nprint(txt_vec[:10])

5. Demonstrations and Tutorials

To see the models in action, explore our demonstration notebooks, where we apply them to legal datasets using various classification models:

• BERT Notebook – A demonstration of the BERT model
• Word2Vec Notebook – Explore the Word2Vec model
• Doc2Vec Notebook – Application of the Doc2Vec model

Troubleshooting and Support

If you encounter issues while using the LegalNLP library, here are some troubleshooting tips:

• Ensure that you have installed the correct version of Gensim (3.8.3). Check your installation using pip show gensim.
• If you are having trouble downloading models, confirm your internet connection is stable and retry the download commands.
• For specific errors, consult the official documentation of Hugging Face or Gensim, as they may provide insights into resolving common issues.
• For more insights, updates, or to collaborate on AI development projects, stay connected with fxis.ai.

At fxis.ai, we believe that such advancements are crucial for the future of AI, as they enable more comprehensive and effective solutions. Our team is continually exploring new methodologies to push the envelope in artificial intelligence, ensuring that our clients benefit from the latest technological innovations.

References

• Mikolov, T. et al. (2013). Distributed representations of words and phrases and their compositionality.
• Mikolov, T. et al. (2013). Efficient estimation of word representations in vector space.
• Le, Q. & Mikolov, T. (2014). Distributed representations of sentences and documents.
• Bojanowski, P. et al. (2017). Enriching word vectors with subword information.
• Devlin, J. et al. (2018). BERT: Pre-training of deep bidirectional transformers for language understanding.
• Souza, F. et al. (2020). BERTimbau: pretrained BERT models for Brazilian Portuguese.