US 9,959,328 B2Granted

Analysis of user text

AssigneeMicrosoft Technology Licensing, LLC

InventorsNavendu Jain, Rajeev Dubey, Huy P. M. Dao, Adam J. Betz, Jason M. Deakins

Life of this patent

Estimating the term…

Loading patent…

Abstract

Free-form text in a document can be analyzed using natural-language processing to determine actionable items specified by users in the text or to provide recommendations, e.g., by automatically analyzing texts from multiple users. Words or phrases of the text can be mapped to classes of a model. An actionable item can be determined using the mapped words or phrases that match a selected grammar pattern. Items can be ranked, e.g., based on frequency across multiple documents. In some examples, the classes can include a suggestion-indicator class or a modal-indicator class, and the selected grammar pattern can include one of those classes. In some examples, the mapping can use a dictionary. A new term not in the dictionary can be automatically associated with classes based on attributes of the new term and of terms in the dictionary, e.g., the new term's part of speech or neighboring terms.

Claims17

Dependent claims

A system comprising:

one or more processors;
memory communicatively coupled to the one or more processors;
an interface communicatively coupled to the one or more processors and configured to receive a plurality of documents, individual ones of the plurality of documents including free-form user text;
a mapping module stored in the memory and executable by the one or more processors to map, for individual ones of the plurality of documents, individual words or phrases of the free-form user text of the respective document to individual classes of a model;
an analysis module stored in the memory and executable by the one or more processors to determine, for individual ones of the plurality of documents, corresponding actionable items based at least in part on a comparison of the mapped words or phrases for the respective document to a selected grammar pattern;
a recommendation module stored in the memory and executable by the one or more processors to analyze individual ones of the actionable items to generate a recommendation;
a reporting module stored in the memory and executable by the one or more processors to present or transmit an indication of the recommendation;
and
a classification module stored in the memory and executable by the one or more processors to automatically:
determine a particular word or phrase that is not associated with a dictionary of a knowledge base and that is present in one of the plurality of documents, the dictionary includes one or more respective attributes for individual ones of the words or phrases in the dictionary;
determine one or more respective attributes of the particular word or phrase;
and
associate, in the dictionary, the particular word with a particular class of the model based at least in part on the one or more respective attributes of the particular word or phrase and one or more of the one or more respective attributes for individual ones of the words or phrases in the dictionary.

10 more elements

A method comprising:

receiving, by a computing device, a plurality of documents, individual ones of the plurality of documents including free-form text;
tagging, by the computing device and for individual ones of the plurality of documents, individual words or phrases of the free-form text of the respective document with individual classes of a set of classes, the tagging being based at least in part on a stored dictionary mapping words and phrases to ones of the set of classes;
determining, by the computing device and for individual ones of the plurality of documents, an actionable item based at least in part on the tagged words or phrases for the respective document;
arranging, by the computing device, the actionable items in an order based at least in part on occurrences of individual ones of the actionable items for the plurality of documents;
determining a particular word or phrase in at least one of the plurality of documents that is not associated with the dictionary, the dictionary includes respective attributes for individual ones of the words or phrases in the dictionary;
determining one or more respective attributes of the particular word or phrase;
and
updating the dictionary to include an association between the particular word or phrase and particular ones of the set of classes based at least in part on the one or more attributes of the particular word or phrase and one or more of the attributes of individual ones of the words or phrases in the dictionary.

5 more elements

One or more non-transitory computer-readable media storing computer-executable instructions that, when executed by the one or more processors, cause the one or more processors to perform operations comprising:

mapping, for individual ones of a plurality of documents including free-form user text, individual words or phrases of the free-form user text of the respective document to individual classes of an ontology;
determining a particular word or phrase in at least one of the plurality of documents that is not associated with a stored dictionary, the dictionary including respective attributes for individual ones of the words or phrases in the dictionary;
determining one or more respective attributes of the particular word or phrase;
and
updating the dictionary to include an association between the particular word or phrase and particular ones of a set of classes based at least in part on the one or more attributes of the particular word or phrase and one or more of the attributes of individual ones of the words or phrases in the dictionary;
determining, for individual ones of the plurality of documents, one or more actionable items regarding an update to a software or hardware element based at least in part on individual ones of the mapped words or phrases for the respective document that match a selected grammar pattern;
ranking individual ones of the actionable items;
and
outputting the ranking of the actionable items.

6 more elements

Dependencies are inferred from the claim text.