Collecting and utilizing user data is essential for effective recommender systems to personalize content. However, privacy and compliance regulations protect personal user data. With strict regulations such as the General Data Protection Regulation (GDPR) or California Privacy Rights Act (CPRA) in effect, one may ask: how can a recommender system be both compliant and effective? This paper aims to answer this question, demonstrating privacy-compliant personalization for the Recommended Documents service within Microsoft 365 (M365), particularly Microsoft Feed. It outlines the development of an exemplary L-Profile personalization feature from conception to productionization, covering offline and online evaluations.

This paper presents the first probabilistic parsing results for French, using the recently released French Treebank. We start with an unlexicalized PCFG as a baseline model, which is enriched to the level of Collins' Model 2 by adding lexicalization and subcategorization. The lexicalized sister-head model and a bigram model are also tested, to deal with the flatness of the French Treebank. The bigram model achieves the best performance: 81% constituency F-score and 84% dependency accuracy. All lexicalized models outperform the unlexicalized baseline, consistent with probabilistic parsing results for English, but contrary to results for German, where lexicalization has only a limited effect on parsing performance.