A transcription of a query for content discovery is generated, and a context of the query is identified, as well as a first plurality of candidate entities to which the query refers. A search is performed based on the context of the query and the first plurality of candidate entities, and results are generated for output. A transcription of a second voice query is then generated, and it is determined whether the second transcription includes a trigger term indicating a corrective query. If so, the context of the first query is retrieved. A second term of the second query similar to a term of the first query is identified, and a second plurality of candidate entities to which the second term refers is determined. A second search is performed based on the second plurality of candidates and the context, and new search results are generated for output.

The methods and systems described herein leveraging acoustic features of a user to generate and present a personalized content to a user. In one example, the method receives a voice query and determines that the query refers to either a first content item or a second content item. The first content item is associated with a first type assigned with a first score and the second content method also determines whether the query is from the second entity type. The method ranks the first and the second content items based on this determination and generates for presentation of the first and the second content items based on the ranking. The method also changes the first or the second scores based on this determination and selects one of the first or the second content item for presentation.

The present disclosure provides systems and methods for selecting and presenting content items based on user input. The method includes receiving first input intended to identify a desired content item among content items associated with metadata, determining that an input portion has an importance measure exceeding a threshold, and providing feedback identifying the input portion. The method further includes receiving second input, and inferring user intent to alter or supplement the first input with the second input. The method further includes, upon inferring intent to alter the first input, determining an alternative query by modifying the first input based on the second input, and, upon inferring intent to supplement the first input, determining an alternative query by combining the first input and the second input. The method further includes selecting and presenting a subset of content items based on comparing the alternative query and metadata associated with the subset.

A transcription of a query for content discovery is generated, and a context of the query is identified, as well as a first plurality of candidate entities to which the query refers. A search is performed based on the context of the query and the first plurality of candidate entities, and results are generated for output. A transcription of a second voice query is then generated, and it is determined whether the second transcription includes a trigger term indicating a corrective query. If so, the context of the first query is retrieved. A second term of the second query similar to a term of the first query is identified, and a second plurality of candidate entities to which the second term refers is determined. A second search is performed based on the second plurality of candidates and the context, and new search results are generated for output.

A query fingerprint of a set of frames of video content captured at a client device may be generated. Multiple patches of the set of frames of video content may be selected and a value calculated for each of the selected multiple patches. The value for each patch may be indicated as a single bit along with an additional 1-bit value to indicate whether the patch value is weak. A database of known reference fingerprints may be queried using the generated query fingerprint. Matches between the query fingerprint and the reference fingerprints may be identified. Weak bits may be given reduced weight in identifying the match of fingerprints. Based on the matches, an identifier for the video content may be returned to the client device. The client device may use the received identifier to access the supplemental content.

In some implementations, a computing device may create a semantic mapping that includes identified features that appear in a particular percentage of assets in a subset of assets of a media library. Also, the computing device may analyze assets of the media library using the semantic mapping to generate semantic scores, which may be used to determine a first tier of assets from the media library that rate highest for semantic score out of all assets. The computing device may present at least one of the first tier assets prominently in a user interface when viewing assets of the media library.

A system for assisting sharing of information includes circuitry to: input a plurality of sentences each representing a statement made by one of a plurality of users, the sentence being generated by speaking or writing during a meeting or by extracting from at least one of meeting data, email data, electronic file data, and chat data at any time; determine a statement type of the statement represented by each one of the plurality of sentences, the statement type being one of a plurality of statement types previously determined; select, from among the plurality of sentences being input, one or more sentences each representing a statement of a specific statement type of the plurality of types; and output a list of the selected one or more sentences as key statements of the plurality of sentences.

Concepts and technologies disclosed herein are directed to increasing misspelling, typographical, and partial search tolerance for search terms. A system can obtain media content metadata associated with media content. The system can identify a plurality of names within the media content metadata and remove a plurality of stop words from the media content metadata. The system can determine a plurality of unigrams and a plurality of bigrams from the media content metadata. The system can determine a plurality of misspellings of the media content metadata. The system can compile the plurality of names, the plurality of stop words, the plurality of unigrams, the plurality of bigrams, and the plurality of misspellings as results. The system can determine a plurality of incorrect search strings based upon the results and can create a dictionary including the plurality of incorrect search strings mapped to a plurality of correct search strings.

A system and method for searching and displaying multimedia search results is disclosed herein. An embodiment operates by determining that an interface including one or more previously saved searches is displayed, each of the previously saved searches corresponding to a set of one or more search terms. An updated plurality of search results is received for each of the previously saved searches from a remote server. A grouping of the plurality of search results for each of a plurality of different search terms is displayed across a plurality of individual time periods.

Feedback information is an important aspect in all machine learning systems. In robot systems that are picking objects from a plurality of objects this has been arranged by acquiring images of objects that have been picked. When images are acquired after picking they can be imaged accurately and the information about picking and also dropping success can be improved by using acquired images as a feedback in the machine learning arrangement being used for controlling the picking and dropping of objects.