Abstract: Approaches in accordance with various embodiments provide unified training and inference frameworks useful for co-training, and performing inferencing using, models such as language and graphing models. The frameworks can be hierarchical, in that an inner layer can include a language model for generating semantic embeddings from input text content, and these semantic embeddings can be passed as input to an outer layer that can generate graph labels based, at least in part, upon a knowledge graph having these semantic embeddings substituted for textual content. Nodes of a knowledge graph that are determined to be related to the input content can then be used for various purposes, such as to generate tags for the content or determine how to route that content.

Abstract: A model training service of a provider network may receive content items as training data. For example, the content items may be documents with certain portions labeled as stale. The model training service may train one or more different types of models using those content items (e.g., natural language inference model, paraphrasing detection model, named entity recognition model). The model training service may then provide the model(s) to a content staleness check (CSC) service. The CSC service may receive, from a client, a request that indicates one or more content items to be checked for staleness. The CSC service may process the content items by the model(s) to generate one or more indications of staleness of the content item. The CSC service may generate a response based on the indications of staleness. For example, the CSC service may generate a document with certain portions highlighted as stale content.