Abstract: Techniques are described for dynamic phase generation and load reduction for a query. A query, for instance, is based on user input of a query in a natural language (NL) form, e.g., an NL query. Generally, an NL query may include multiple terms and/or phrases that make up a complex query, such as a sentence in a human-readable language. Accordingly, to enable a query result to be generated, the NL query is parsed into multiple logical sections and query contexts are determined for the logical sections. A set of search phases is generated based on the logical sections and the query contexts. The search phases can then be executed in a specific execution order to generate a query result for the NL query.

Abstract: Techniques are described for differentiation of search results for accurate query output. Generally, such techniques provide intelligent grouping and output of search results by considering contextual factors that relate different search results. For instance, data records that are aggregated in response to a query are inspected to identify attributes (e.g., data types) and attribute values to determine contextual relationships between attributes. The contextual relationships are utilized to differentiate the data records into different groups for query output, thus enabling an accurate characterization of query output not enabled by conventional search technologies.

Abstract: Techniques are described for automated summarization of extracted insight data. Insight data, for instance, is summarized via headlines that include content describing insight data, such as text, images, animations, and so forth. In at least some implementations, headlines are generated in response to trigger events, such as time-based and/or user behavioral events that indicate that headlines are to be generated. Further, headlines are selected to cause insight data represented by the headlines to be presented. Implementations include headline ranking to rank and present headlines based on their relevance to different metrics, and headline deduplication to identify and/or remove duplicate headlines.

Abstract: Techniques described herein provide automated generation of a narrated analytics playlist. Various implementations curate data from various data sources, where curating the data includes identifying attributes and relational data models. One or more implementations base the curating upon anecdotal data associated with a user. In response to receiving a trigger event to perform a query analysis, one or more implementations identify keywords to use in the query analysis, and extract information from the curated data based, at least in part on the one or more keywords. The extracted information is then analyzed to identify insights. In turn, one or more implementations generate a narrated analytics playlist using the insights. Some implementations utilize machine-learning algorithms to curate, extract and/or process data to generate insights. Various implementations abstract the data used to teach the machine-learning algorithms and share the abstracted data to other devices.

Abstract: Some implementations generate logical queries from a canonical query, where the logical queries each reflect a modified scope of the canonical query. Implementations receive, via a personalized analytics system, a canonical query that is associated with a user. The canonical query can be analyzed to determine an intent of the canonical query. In turn, one or more implementations generate, based on the intent an anecdotal information associated with the user, a logical query that reflects a modified scope of the canonical query. In implementations multiple logical queries are generated and are processed to remove a duplicate logical query. A logical query can be used to extract data from a database associated with the personalized analytics system based on a modified scope.

Abstract: Some implementations generate logical queries from a canonical query, where the logical queries each reflect a modified scope of the canonical query. Implementations receive, via a personalized analytics system, a canonical query that is associated with a user. The canonical query can be analyzed to determine an intent of the canonical query. In turn, one or more implementations generate, based on the intent an anecdotal information associated with the user, a logical query that reflects a modified scope of the canonical query. The logical query can be used to extract data from a database associated with the personalized analytics system based on the modified scope.

Abstract: Techniques described modify playback of a narrated analytics playlist in a personalized analytics system. In some implementations, audible input is received during playback of the narrated analytics playlist. The audible input can be used to control the behavior of a playback module playing out the narrated analytics playlist. Alternately or additionally, user input can be received, where the user input corresponds to modifying an original scene included in the narrated analytics playlist. Some implementations generate synchronized audible output that be output with the modified original scene of the narrated analytics playlist. Alternately or additional, implementations can automatically determine to visually apply an auto-pointer to portions of the narrated analytics playlist.

Abstract: Techniques are described for differentiation of search results for accurate query output. Generally, such techniques provide intelligent grouping and output of search results by considering contextual factors that relate different search results. For instance, data records that are aggregated in response to a query are inspected to identify attributes (e.g., data types) and attribute values to determine contextual relationships between attributes. The contextual relationships are utilized to differentiate the data records into different groups for query output, thus enabling an accurate characterization of query output not enabled by conventional search technologies.

Abstract: Techniques are described for dynamic phase generation and load reduction for a query. A query, for instance, is based on user input of a query in a natural language (NL) form, e.g., an NL query. Generally, an NL query may include multiple terms and/or phrases that make up a complex query, such as a sentence in a human-readable language. Accordingly, to enable a query result to be generated, the NL query is parsed into multiple logical sections and query contexts are determined for the logical sections. A set of search phases is generated based on the logical sections and the query contexts. The search phases can then be executed in a specific execution order to generate a query result for the NL query.

Abstract: Techniques are described for automated summarization of extracted insight data. Insight data, for instance, is summarized via headlines that include content describing insight data, such as text, images, animations, and so forth. In at least some implementations, headlines are generated in response to trigger events, such as time-based and/or user behavioral events that indicate that headlines are to be generated. Further, headlines are selected to cause insight data represented by the headlines to be presented. Implementations include headline ranking to rank and present headlines based on their relevance to different metrics, and headline deduplication to identify and/or remove duplicate headlines.

Abstract: Techniques described herein provide automated generation of a narrated analytics playlist. Various implementations curate data from various data sources, where curating the data includes identifying attributes and relational data models. One or more implementations base the curating upon anecdotal data associated with a user. In response to receiving a trigger event to perform a query analysis, one or more implementations identify keywords to use in the query analysis, and extract information from the curated data based, at least in part on the one or more keywords. The extracted information is then analyzed to identify insights. In turn, one or more implementations generate a narrated analytics playlist using the insights. Some implementations utilize machine-learning algorithms to curate, extract and/or process data to generate insights. Various implementations abstract the data used to teach the machine-learning algorithms and share the abstracted data to other devices.

Abstract: Techniques described modify playback of a narrated analytics playlist in a personalized analytics system. In some implementations, audible input is received during playback of the narrated analytics playlist. The audible input can be used to control the behavior of a playback module playing out the narrated analytics playlist. Alternately or additionally, user input can be received, where the user input corresponds to modifying an original scene included in the narrated analytics playlist. Some implementations generate synchronized audible output that be output with the modified original scene of the narrated analytics playlist. Alternately or additional, implementations can automatically determine to visually apply an auto-pointer to portions of the narrated analytics playlist.

Abstract: Some implementations generate logical queries from a canonical query, where the logical queries each reflect a modified scope of the canonical query. Implementations receive, via a personalized analytics system, a canonical query that is associated with a user. The canonical query can be analyzed to determine an intent of the canonical query. In turn, one or more implementations generate, based on the intent an anecdotal information associated with the user, a logical query that reflects a modified scope of the canonical query. The logical query can be used to extract data from a database associated with the personalized analytics system based on the modified scope.