Software, datasets and online services developed by my research group, most of them as outputs of funded research projects. Source code and further details are available from each tool’s page.
SocialNetCrawler is a desktop application, written in Java with a Swing interface, for collecting data from social networks. To use it, obtain an access token from the developer page of each supported network, enter it in the corresponding tab of the application and start crawling.
Access here
SocialDictionary is a Python package that improves sentiment analysis of texts containing emojis. Emojis collected from GetEmoji are stored with their Unicode code and description; each description is scored and labelled as positive, neutral or negative, and assigned an emotion using EmoRoBERTa (trained on GoEmotions). Given an input text, the package returns the text with each emoji replaced by its emotion, the overall sentiment of the text, and details about the emojis it contains.
Access here
SENTAWEB is an online tool for the automatic extraction of multiword expressions: sequences of lexical units (characters, words or punctuation marks), contiguous or not, that behave as syntactic-semantic units with their own meaning.
Access here
Multilingual Text Parser is a web tool for the syntactic and morphological analysis of text, built with the Python libraries spaCy and NLTK. It currently supports Portuguese, French, Spanish and English.
Access here
HultigCrawler is a web crawler that recursively collects all the text from a given website. Each page is stored as an item with its URL, title, tags and text, and saved to a database through Scrapy pipelines.
Access here
HULTIG-C is a cloud platform that provides computational linguistics services built around a large multilingual corpus, created to support research in information retrieval and human language technologies. The corpus covers several languages and includes annotations for keywords, sentences, named entities and multiword expressions, which can be explored through an online demo and accessed programmatically through an API.
Access here
ExtremeSentiLex is a lexicon of extreme sentiments built from SentiWordNet and SenticNet. Three resources are available for download: the lexicon, the datasets classified to validate it, and a bundle with both. The lexicon can also be queried through an online platform and an API.
Access here
Categorium is a Python package for building and training BERT-based text classification models, with preset parameters suited to small datasets. It includes six pre-trained models covering 27 categories each, and tools for testing language models.
Access here