Amazon Technologies, Inc.

Tanvi Milind Bhadbhade

David R Azari

Alan Kipust

Divya Mahalingam

Lee M Miller

Nadia C Payet

Denis V Batalov

Theresa M Hollis

Cynthia A Prentice

Techniques are described for analyzing a submitted manuscript through the use of an entity/action relationship matrix determined for the manuscript. The entity/action relationship matrix may describe one or more entities, such as characters, locations, objects and so forth, which are identified in the manuscript using a natural language analysis such as a part-of-speech analysis. The entity/action relationship matrix may also describe actions or events associated with each of the entities, as determined using the natural language analysis. The entity/action relationship matrix may be employed to determine whether the manuscript complies with rules constraining the use of entities in the manuscript.

Event detection based on entity analysis

Systems, devices, and processes for classifying a digital item are described. In some examples, a first classifier determines whether a digital item, such as an electronic book (eBook), includes content of a first category that is acceptable for publication by a publisher. A second classifier determines whether the digital item includes content of a second category that is acceptable for publication by a publisher. In response to determining that the digital item includes content of the first category or content of the second category, a third classifier may determine whether the digital item includes a phrase that is indicative of content of a third category that is unacceptable for publication.

Classifier for classifying digital items

One or more computing systems can implement a classifier to classify content publishers as being likely to provide appropriate content or as being likely to provide inappropriate content. The classifier can gather information from previously classified publishers. The information from the previously classified publishers can used to train the classifier. Based on the training, the classifier can learn about traits, characteristics, and/or behavioral patterns, etc., associated with publishers that have been previously classified as being good as well as publishers previously classified as being bad. The classifier can then process information about an unclassified publisher to determine a classification for the unclassified publisher, as being good (and likely to provide appropriate content) or bad (and likely to provide inappropriate content).

Machine-assisted publisher classification

When a digital item is submitted for publication, an automated system may determine whether the digital item includes content from other digital items. In some implementations, when the digital item is an electronic book (eBook), the automated system may select sets of words from the eBook and compute hash codes, such that each hash code corresponds to a set of words. The automated system may compare the computed hash codes with retained hash codes associated with other electronic books to determine whether the digital item includes duplicate content.

Growing community of inventors

Seattle, WA, United States of America

Tanvi Milind Bhadbhade