Abstract: A candidate attribute combination of a first data set is identified, such that the candidate attribute combination meets a data type similarity criterion with respect to a collection of data types of sensitive information for which the first data set is to be analyzed. A collection of input features is generated for a machine learning model from the candidate attribute combination, including at least one feature indicative of a statistical relationship between the values of the candidate attribute combination and a second data set. An indication of a predicted probability of a presence of sensitive information in the first data set is obtained using the machine learning model.

Abstract: A generative adversarial network (GAN) may be implemented to recognize named entity types in detection of sensitive information in datasets. The GAN may include a generator and a discriminator. The generator may be trained to produce synthetic data to include information that simulates named entity types representing the sensitive information. The discriminator may be fed with real data that are known to include the sensitive information (as positive examples), together with the synthetic data that simulate the sensitive information (as negative examples), to train to classify the real vs. synthetic data. In field operations, the discriminator may be deployed to perform named entity type recognition to identify data having the sensitive information. The generator may be deployed to provide anonymous data in lieu of real data to facilitate sensitive information sharing and disclosure.

Abstract: Systems and techniques are generally described for anomaly detection in text. In some examples, text data comprising a plurality of words may be received. An image of a first word of the plurality of words may be generated. A feature representation of the first word may be generated using a variational autoencoder. A score may be generated based at least in part on the feature representation. In various examples, the score may indicate a likelihood that an appearance of the first word in the image of the first word is anomalous with respect to at least some other words of the plurality of words.

Abstract: Systems and methods are described for natural language processing of a text sequence. The system can identify a set of text and location information for the set of text in an image. The set of text may correspond to an input sequence space. The system can project embeddings of the text into a latent space for processing. Further, the system can reproject the processed embeddings from the latent space to the input sequence space. The system may perform multiple stages of projecting the embeddings to the latent space and reprojecting the processed embeddings from the latent space to the input sequence space. The system can route the reprojected embeddings to a neural network that can identify class predictions for elements of the set of text.

Abstract: A candidate attribute combination of a first data set is identified, such that the candidate attribute combination meets a data type similarity criterion with respect to a collection of data types of sensitive information for which the first data set is to be analyzed. A collection of input features is generated for a machine learning model from the candidate attribute combination, including at least one feature indicative of a statistical relationship between the values of the candidate attribute combination and a second data set. An indication of a predicted probability of a presence of sensitive information in the first data set is obtained using the machine learning model.

Abstract: A candidate attribute combination of a first data set is identified, such that the candidate attribute combination meets a data type similarity criterion with respect to a collection of data types of sensitive information for which the first data set is to be analyzed. A collection of input features is generated for a machine learning model from the candidate attribute combination, including at least one feature indicative of a statistical relationship between the values of the candidate attribute combination and a second data set. An indication of a predicted probability of a presence of sensitive information in the first data set is obtained using the machine learning model.

Abstract: Techniques for dynamically provisioning host devices to process requests and other types of received data include receiving traffic data that indicates an amount of data received by the host devices over time and resource data that indicates an amount of computing resources used by the host devices to process the data. Host data is generated that indicates a relationship between received quantities of data and corresponding quantities of computing resources used to process the data. Based on the host data, a number of host devices used to process a predicted amount of data to be received at a future time, using a selected amount of computational resources, may be determined. Based on the determined number of devices, additional host devices are provisioned to process the received data, or diverted from processing the data.

Abstract: Architecture that provides fully automatic generation of a geo-ontology and does not use pre-existing geo-ontologies or other location entity repositories (e.g., a licensed location). The architecture extracts the formal administrative structure of a geographical region of interest (e.g., country) (a geo-ontology of locations with attributes and relations) from a collection of entities with spatial attributes, extracts the informal administrative structure of a geographical region of interest (e.g., country) (informal administrative regions and names and informal neighborhoods and their attributes), and extracts location static rank features for all these entities (attributes used for ranking locations from the geo-ontology that appear in user queries).

Abstract: Architecture that generates local intent suggestions as completion suggestions (auto-suggest solutions) for an incomplete (or partially-entered) search query. The local intent suggestions are “synthetic” in that these suggestions are derived based on the near or total absence of any prior query history. The local intent suggestions can be derived and presented without the typical web-based suggestions or with the web-based suggestions. The web-based suggestions are blended (e.g., placed above, under, or mixed in with) with the local intent suggestions and displayed to the user. Web interfaces include, but are not limited to a full page display interface (e.g., that includes a full page map, pin points with images, sub-intents, title and description), and/or a right pane (sub-pane) auto-suggest interface with mini-map pinpoints, sub-intents, and disambiguation tiles, for example.

Abstract: Embodiments of the present invention dynamically generate customized online advertisements. A dynamic advertisement may take the form of a paid search result. The advertisement is dynamic in the sense that all of or portions of the advertisement are generated using content from a web page operated by an entity associated with the advertisement. The web page used to generate the dynamic ad may be the entity's most responsive web page within normal search results. Embodiments of the invention use content from the most responsive web page to generate one or more dynamic ad features. Possible dynamic ad features include the title of the advertisement, a landing page of the advertisement, a caption on the advertisement, and a dynamic annotation of the advertisement.

Abstract: Architecture that provides fully automatic generation of a geo-ontology and does not use pre-existing geo-ontologies or other location entity repositories (e.g., a licensed location). The architecture extracts the formal administrative structure of a geographical region of interest (e.g., country) (a geo-ontology of locations with attributes and relations) from a collection of entities with spatial attributes, extracts the informal administrative structure of a geographical region of interest (e.g., country) (informal administrative regions and names and informal neighborhoods and their attributes), and extracts location static rank features for all these entities (attributes used for ranking locations from the geo-ontology that appear in user queries).