Rosette API 1.7 ReleaseJune 15, 2017
Great news! Yesterday we released Rosette API v. 1.7. We added support for Arabic sentiment analysis (beta), confidence scores for all extracted entities, pronominal resolution in targeted relationship extraction, and a new /transliteration endpoint for transforming romanized Arabic chat text (“Arabizi”) to standard Arabic script. We also introduced specialized linguistic analysis for emojis, emoticons, hashtags, @mentions, urls, and email addresses in all supported languages.
Sentiment Analysis: We now support both document and entity-level sentiment in Arabic, with particular emphasis on Modern Standard (MSA) and Egyptian dialect.
Entity Extraction: Rosette API now returns a confidence score alongside every entity extracted using our statistical algorithm. This makes it easier to tune your results to achieve the optimal balance of precision and recall.
Relationship Extraction (LABS): Pronouns are now extracted and resolved as components of targeted relationships. For example, if given the text
“Queen Elizabeth II is the current queen of England. She is the mother of Prince Charles”
Rosette will return the relationship
<Prince Charles, PER-PARENTS, Elizabeth II>.
Transliteration (LABS): Our new endpoint /transliteration takes in Arabizi (a romanized form of Arabic used primarily in online chat and social media contexts), and transliterates it into standard Arabic script. It can also transliterate standard Arabic into Arabizi. For example, the Arabizi text
“ana r2ye7 el gam3a el sa3a 3 el 3asr”
would be translated into the Arabic
“ أنا رايح الجامعة الساعة ٣ العصر”
Morphological Analysis and Tokenization: Rosette API now tokenizes, normalizes and part-of-speech tags emojis, emoticons, hashtags, @mentions, urls, and email addresses in all supported languages, greatly improving the analysis accuracy further downstream.
Already a Rosette API user? Make sure you update your client binding! This should ensure that none of the new technical changes break your code — look for version 1.7.x in the package manager of your choice. In addition, if you use Postman, check out our updated collection.
Did you catch our recent blog posts about the Basis Technology R&D team’s presentation at the ICLR conference in Toulon, France, a recipe for summarization with Rosette API, and tweet analysis related to the recent elections in France? Check them out. Have a cool project you’d like us to feature on our blog? Let us know.