When it comes to mining and analyzing text data, text classification plays an important role. Categorizing text based on sentiment, genre, status, or intent is useful for tasks like language detection, customer feedback analysis, and fraud detection. However, arriving at these data insights can be both time and labor intensive when done manually. Fortunately, with the development of machine learning and natural language processing, much of the process can now be automated.
Below, we’ve compiled a list of open source tools for developing your own text classification system. We’ve also listed available services and platforms that include text classification as part of their suite of text analysis tools.
Open Source Tools
-
Apache OpenNLP: OpenNLP supports common NLP tasks such as tokenization, sentence segmentation, named entity extraction, and language detection. It also offers text classification through its Document Classifier, which allows you to train a model that categorizes text based on pre-defined categories.
-
The Natural Language Toolkit: Commonly referred to as NLTK, the Natural Language Toolkit is an open-source, community driven project for natural language processing tasks. The creators have written a guidebook that walks through the fundamentals of writing Python programs for tasks including text classification, analyzing linguistic structure, and more.
-
Orange: Specializing in building data analysis workflows and visualizations, Orange offers a host of NLP and analytics tools. These include text classification, social media data analysis, and sentiment analysis. Their team also offers online training courses in data mining to help people understand data exploration without the coding and the math.
-
TextFlows: This online platform is designed for the composition, execution, and sharing of text mining and NLP workflows for text analysis tasks. It uses visual programming to simplify complex procedures and is cloud-based, meaning you can work anywhere without installing it on your local hard drive.
-
Textable: Built on top of the Orange framework, Textable is built specifically for analyzing and processing texts visually. By adding blocks to create data processing “recipes”, you can create data analysis workflows and gain visual insights into them quickly.
-
DatumBox: The DatumBox API currently offers 14 different functions as part of its machine learning platform, including topic classification, subjectivity analysis, keyword extraction, and more. It supports a variety of different methods and algorithms that can be found on their official website.
Text Classification Services
-
MeaningCloud: MeaningCloud is a set of APIs (application programming interfaces) for text analytics, including text classification. Its flexibility makes it a great option for developers, but the coding requirements make it a more difficult option for non-technical users. However, a free version is also available for processing up to 20,000 requests per month if you’d like to try it out.
-
MonkeyLearn: The MonkeyLearn platform can be used to build a custom text classifier to categorize your text data as per your programmed specifications. The process involves uploading your data, defining your tags, and training the model by tagging data for it to learn from. You can then test it, improve it as necessary, and put it to work.
-
Google Cloud NLP: If your data is already stored on Google’s cloud, their NLP service may be an easy way to smoothly transition into text analysis. The AutoML Natural Language platform allows you to upload documents based on specific keywords and phrases, then train a model and evaluate it.
-
IBM Watson: The Watson Natural Language Classifier is part of a suite of text analysis tools available with IBM Watson. If you have your training data ready, the classifier is easy to train, and the system is built to make it easy to integrate into applications. Do keep in mind however that coding may be necessary to really get the most out of their classifier.
-
Aylien: Specializing in the analysis of news articles, Aylien’s text analysis allows you to create a custom text classification model without leaving your browser. They boast a simple process that doesn’t require coding, and a database of documents from which to start building a dataset.
-
Rosette: Part of Basis Technology, Rosette’s text classification system comes pre-trained on the IAB Tech Lab Content Taxonomy, but can also be customized through keyword-based training or a training dataset.
Text Classification Datasets
To make the most of the tools above, you’ll need a dataset of annotated text data to train your model to accurately classify text per your specifications.
If you’re looking for text classification datasets to help with the training of a customized machine learning model, we’ve compiled datasets from across the web. You can find datasets for product reviews, online content evaluation, news classification, and available dataset repositories. They should provide a good starting point for machine learning projects.
The Lionbridge Text Classification Tool
There are a variety of approaches you can take to data labeling, but if you’re unsure of where to start, get in touch to learn about our own text classification tools and services.
Lionbridge provides data services to collect, clean, and annotate text data for a wide range of use-cases. You can set up text classification projects on our dedicated data annotation platform with your own internal team. Alternatively, you can work with our community of 1,000,000+ qualified annotators, data scientists, and project managers to help complete your next big project.