Abstract: Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for encoding and decoding delta values. In one aspect, a method includes accessing a compression buffer having a start position, a sentinel position, and a data storage region; obtaining a first value; determining that a second value stored in the sentinel position does not match a first sentinel value; determining that a third value stored in the start position matches a second sentinel value; and storing the first value at the start position of the compression buffer.

Abstract: A document analysis system analyzes a corpus of documents and automatically generates a dictionary of specialized phrases not already in conventional dictionaries. The dictionary generation process involves a series of operations on the phrases to identify the phrases most suitable for inclusion in a dictionary, such as phrase scoring and phrase clustering. The dictionary generation process also comprises the identification of one or more corresponding definitions for the various phrases identified for inclusion in the specialized dictionary.

Abstract: Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for encoding and decoding delta values. In one aspect, a method includes accessing a compression buffer having a start position, a sentinel position, and a data storage region; obtaining a first value; determining that a second value stored in the sentinel position does not match a first sentinel value; determining that a third value stored in the start position matches a second sentinel value; and storing the first value at the start position of the compression buffer.

Abstract: Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for providing display suggestions. In one aspect, a method includes accessing a resource that includes multiple terms, obtaining one or more prefixes that are derived from the multiple terms and, for each prefix, one or more actual suggestions, wherein each actual suggestion is a term from the resource that includes the prefix, obtaining one or more display suggestions, wherein each display suggestion includes two or more successive terms from the resource that are identified as related, and outputting, in response to receiving a user input of a particular prefix, a representation of a particular display suggestion that includes a term that is an actual suggestion for the particular prefix.

Abstract: Disk-backed array techniques can, in some implementations, help ensure that the arrays contain consistent data. An alert can be provided if it is determined that the data in the array is, or may be, corrupted.

Abstract: The present disclosure describes a data indexing and search service that resides locally on a computing device (e.g., a mobile phone) and that can host data for multiple applications on the device. By centralizing the storage of data as well as the search and query functions, unified search queries can be performed by the service.

Abstract: A data structure includes three arrays. A first array includes a root node, one or more intermediate nodes, and one or more leaf nodes. A second array includes one or more slices, each of which includes one or more elements; a third array also includes one or more elements. Each root node and intermediate node in the first array points to a respective slice in the second array. Each element in the second array stores a byte and has an associated pointer that points either to a respective intermediate node or to a respective leaf node in the first array. Each leaf node in the first array points to a respective element in the third array, and each element in the third array stores a respective sub-string of bytes. The trie can be particularly advantageous, for example, in applications on resource-constrained computing devices, such as mobile phones, tablets, and other hand-held computing devices.

Abstract: Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for storing data on in a storage medium. In one aspect, a method includes receiving a key-value pair including a key k and a value v. The method further includes encoding the key-value pair as (i) a first key-value pair including a first key k1 and first value v1, and (ii) a second key-value pair including a second key k2. The method further includes inserting the first key-value pair and the second key-value pair in a trie.

Abstract: A system and method for discovering content. The system includes a feature extraction application, an indexed server, and a classification application. The feature extraction application is configured to use training examples specific to a first domain type to process content objects and to generate an object vector. The feature extraction application is also configured to apply the feature extraction to a second domain type and to generate another object vector. The indexed server is coupled to the feature extraction application and configured to maintain feature-extracted content objects and to communicate the content objects with a client. The classification application is coupled to the indexed server and configured to cluster content objects and implement at least one level of classification. The classification application is also configured to generate summary vectors formed of weighted sums of object vectors.

Abstract: A system and method for feature extraction of a content object. The system includes a database and a feature extraction application. The database is configured to database to store a plurality of content objects. The feature extraction application is coupled to the database and configured to process each content object, extract a core set of features, and generate an object vector. The object vector includes a vector of numbers representative of a frequency of a superset of features potentially found in the content object.

Abstract: A system and method for indexing content. The system includes a crawler, a crawl database, an indexer, a classification application, and an indexed data server. The crawler is configured to crawl the internet for content objects. The crawl database is coupled to the crawler and configured to cache the content objects. The indexer is coupled to the crawl database and configured to perform feature extraction on the content objects and cluster the content objects by generating an object vector. The classification application is coupled to the indexer and configured to cluster the object vectors and generate a summary vector. The indexed data server is coupled to the indexer and configured to communicate the content objects with a client.

Abstract: A system and method for peer to peer interest sharing. The system includes a plurality of clients and a peer to peer collaboration application. Each client is in communication over the internet with at least one other of the plurality of clients. The peer to peer collaboration application is coupled to at least one of the clients and configured to share content objects of a shared interest of a user among selected peers. The peer to peer collaboration application is further configured to apply a positive and a negative content rating from one client to classifiers of each of the plurality of clients.

Abstract: A system and method for classifying content objects. The system includes a database and a classification application. The database is configured to store a plurality of content objects. The classification application is coupled to the database and configured to cluster content objects and implement at least one level of classification, including generating summary vectors formed of weighted sums of object vectors. The object vector includes a vector of numbers representative of a frequency of a superset of features potentially found in the content object.

Abstract: A system and method for advertisement selection. The system includes a client and an advertisement selection application. The client is capable of accessing content objects over the internet. The advertisement selection application is coupled to the client and configured to maintain at least a partial copy of advertisements. The advertisement selection application is also configured to classify the advertisements and to generate an object vector for each advertisement. The advertisement selection application is also configured to update a user interest profile in response to positive and negative feedback, and to display advertisements according to the user interest profile.

Abstract: A system and method for searching content objects. The system includes a client, a cache, and a content searching application. The client is capable of accessing content objects over the internet. The cache is coupled to the client and configured to store at least a partial copy of object vectors representative of content in the content objects. The content searching application is coupled to the cache and configured to compare the object vectors to a user interest profile to find content similar to the user interest profile.