Amazon Technologies, Inc.

Rishita Rajal Anubhai

Sravan Babu Bodapati

Yahor Pushkin

Yaser Al-Onaizan

Sunil Mallya Kasaragod

Katrin Kirchhoff

Muhammad Bilal Zafar

Vidya Sagar Ravipati

Dimitrios Soulios

Pu Paul Zhao

Miguel Ballesteros Martinez

Sameer Karnik

Georgiana Dinu

Kasturi Bhattacharjee

Abhinav Goyal

V Divya Bhargavi

Graham Vintcent Horwood

Shuai Wang

Srikanth Doss Kadarundalagi Raghura

Jie Ma

Smaranda Muresan

Yinxiao Zhang

Lin Lee Cheong

Ravindra Manjunatha

Alessandra Brusadin

Jonathan Steuck

Haibo Ding

Federico Torreti

Siddharth Chaitanyakumar Varia

Huzefa Rangwala

Techniques for implementing and using a customizable framework for natural language processing explainability are described. A user of a machine learning (ML) service can generate an explainability configuration identifying a user-selected segmented or a user-selected granularity level, from multiple granularity levels supported by the ML service, at which to use for segmenting text during an ML explainability analysis. The ML service can execute a user-configurable explanation pipeline as part of the explainability analysis of at least an input text, including segmenting the input text using the user-selected segmented or a segmentation algorithm corresponding to the user-selected granularity level to yield candidate segments, generating a mask corresponding to the candidate segments, and executing an explanation algorithm based at least on use of the candidate segments and the mask to yield a result.

Customizable framework for natural language processing explainability

A global segmenting and analysis service of a provider network may receive documents (e.g., posts, product reviews) from different applications. The service may analyze the documents to identify target entities and sentiment. The service may generate different levels of sentiment data and store data into a segmented database. For example, the service may store within-document level sentiment, document-level sentiment, and multi-document level sentiment for a target entity. The service may also update the entity taxonomy automatically or with only a small number of sample documents. The client may query the service for the segmented sentiment data.

Global segmenting and sentiment analysis based on granular opinion detection

Methods, systems, and computer-readable media for event extraction from documents with co-reference are disclosed. An event extraction service identifies one or more trigger groups in a document comprising text. An individual one of the trigger groups comprises one or more textual references to an occurrence of an event. The one or more trigger groups are associated with one or more semantic roles for entities. The event extraction service identifies one or more entity groups in the document. An individual one of the entity groups comprises one or more textual references to a real-world object. The event extraction service assigns one or more of the entity groups to one or more of the semantic roles. The event extraction service generates an output indicating the one or more trigger groups and one or more entity groups assigned to the semantic roles.

Event extraction from documents with co-reference

Techniques for aligning (merging) knowledge graphs is described. Graph entity alignment is the problem of 'joining' two knowledge graphs based on common entities. Most approaches in literature solve this using some annotated seed entity pairs and train a supervised model to rank entities in one knowledge base against another. In some examples, unsupervised alignment approach that uses graph-to-text summaries to encode entities in two or more distinct graphs into the same representation space, encodes those summaries into a common space, and then uses similarity analysis to determine when graph entities should be aligned (merged).

Knowledge graph alignment

Techniques for multi-label document classification are described. Clustering is used to cluster labels in a set. A machine learning model including a multi-label classifier for each cluster is created, the multi-label classifier for a given cluster to classify a document with one or more of the labels in the cluster.

Multi-label document classification for documents from disjoint class sets

Techniques for creating a text classifier machine learning (ML) model are described. According to some embodiments, a language processing service finetunes a language ML model on unlabeled documents of a user, and then trains that finetuned language ML model on labeled documents of the user to be a text classifier that is customized for that user's domain, e.g., the user's documents. Additionally, the finetuned language ML model may be trained on labeled documents of the user, for prediction objectives for unlabeled data, before being trained as the text classifier.

Creating text classification machine learning models

Techniques for data lake-based text generation and data augmentation for machine learning training are described. A user-provided dataset including documents and corresponding label information can be automatically supplemented by creating additional high-quality document samples, with labels, via a large repository of documents in a data lake. Documents from the data lake may be identified as being semantically similar to the user-provided documents but different enough to allow a resulting model to learn from the variation in these documents. New documents can be generated from user-provided document samples or data lake sample documents by identifying and replacing slots within the samples and rewriting adjunct tokens.

Data lake-based text generation and data augmentation for machine learning training

Techniques for extending sensitive data tagging without reannotating training data are described. A method for extending sensitive data tagging without reannotating training data may include hosting a plurality of models at a model endpoint in a machine learning service, each model trained to identify a different sensitive data type in a transcript of content, adding a new model to the model endpoint, the new model trained to identify a new sensitive data entity in the transcript of content, identifying sensitive entities in the transcript by each of the plurality of models and the new model, merging inference responses generated by each of the plurality of models and the new model using at least one inference policy, and returning a merged inference response identifying a plurality of sensitive entities in the transcript.

Extending sensitive data tagging without reannotating training data

Techniques are described for a de-obfuscation framework that utilizes image recognition of text. A word input by a user is received by the de-obfuscation service. Visual feature data associated with an image corresponding to each character of the word is generated. Word embeddings are generated using the visual feature data and each character of the word using a character encoder layer. Feature vectors are generated from the word embedding by combining the generated word embeddings and a provided word embedding using a second neural network. The generated feature vector is classified. Potential text obfuscation is detected from the classified generated feature vector using a lexicon to determine de-obfuscated text closet to the user text.

Text de-obfuscation with image recognition of text

Growing community of inventors

Seattle, WA, United States of America

Rishita Rajal Anubhai