Back to results
US 11,455,473 B2Granted
Vector representation based on context
AssigneeInternational Business Machines Corporation
InventorsAnkur Tagra, Krishnendu Saha
Life of this patent
Estimating the term…
Abstract
Embodiments relate to a system, program product, and method for use with an intelligent computer platform to create and apply textual data in vector format, and more specifically to apply context to the vector representation. Both context and document vectors are generated and assessed, with a calculated distance between the vectors corresponding to a weight. Word vectors are generated with associated word pairs and frequencies. A word vector generation model is trained. Utilization of the trained model generates one or more context sensitive word vector representations. A summarized sentence document is created and returned through application of the context sensitive word vectors.
Claims20
A computer system comprising:
- a processing unit operatively coupled to memory;
- and
- a tool in communication with the processing unit to create a word vector representation and to apply the word vector representation to context, including:
- a document manager configured to compute document relevance, including generate one or more document vectors for one or more documents, calculate respective distances between the one or more generated document vectors and one or more context vectors generated for context provided by a user or extracted from user behavior, and assign respective weights to the one or more documents based on the calculated distances;
- a word manager operatively coupled to the document manager, the word manager configured to:
- generate one or more word vectors for the one or more documents, including represent the one or more documents as a collection set of word pairs from the one or more documents and respective frequencies associated with the word pairs in the one or more documents;
- and
- train a word vector generation model with at least a calculated weighted frequency, including train the word vector generation model to generate one or more context sensitive word vector representations, the calculated weighted frequency based upon the respective weights assigned to the one or more documents and the frequencies associated with the word pairs;
- and
- a director operatively coupled to the document manager and the word manager, the director configured to apply the trained word vector model to context, including using the word vector generation model to generate the one or more context sensitive word vector representations.
7 more elements
A computer program product to create and apply word vector representation to context, the computer program product comprising:
- a computer readable storage medium having program code embodied therewith, the program code executable by a processing unit to:
- generate one or more context vectors for context provided by a user or extracted from user behavior;
- compute document relevance, including generate one or more document vectors for one or more documents, calculate respective distances between the one or more generated document vectors and the one or more context vectors, and assign respective weights to the one or more documents based on the calculated distances;
- generate one or more word vectors for the one or more documents, including represent the one or more documents as a collection set of word pairs from the one or more documents and respective frequencies associated with the word pairs in the one or more documents;
- train a word vector generation model with at least a calculated weighted frequency, including train the word vector generation model to generate one or more context sensitive word vector representations, the calculated weighted frequency based upon the respective weights assigned to the one or more documents and the frequencies associated with the word pairs;
- and
- apply the trained word vector model to context, including using the word vector generation model to generate the one or more context sensitive word vector representations.
4 more elements
A method comprising:
- generating one or more context vectors for context provided by a user or extracted from user behavior;
- computing document relevance, including generating one or more document vectors for one or more documents, calculating respective distances between the one or more generated document vectors and the one or more context vectors, and assigning respective weights to the one or more documents based on the calculated distances;
- generating one or more word vectors for the one or more documents, including representing the one or more documents as a collection set of word pairs from the one or more documents and respective frequencies associated with the word pairs in the one or more documents;
- training a word vector generation model with at least a calculated weighted frequency, including training the word vector generation model to generate one or more context sensitive word vector representations, the calculated weighted frequency based upon the respective weights assigned to the one or more documents and the frequencies associated with the word pairs;
- and
- applying a word vector model to context, including using the word vector generation model to generate the one or more context sensitive word vector representations.
3 more elements
Dependencies are inferred from the claim text.