Tone/Sentiment-Correct Synonym Suggestion Using NLP
Publication Date: 2014-Jun-19
The IP.com Prior Art Database
AbstractDisclosed is a system that determines the tone that an author is trying to convey in a document via Natural Language Processing (NLP) (annotators) and machine learning techniques. When the system detects that an entered word does not match the current tone, it highlights the word and offers a synonymous suggestion.
Page 01 of 2
Disclosed herein is a system to help a user achieve an even tone in a piece of writing by providing alternate word suggestions that match the tone of the rest of the current document. Current approaches are manual; the user selects a tone from a predetermined list of options prior to beginning writing.
The disclosed system determines the tone that an author is trying to convey in a document via Natural Language Processing (NLP) (annotators) and machine learning techniques. The derived tone is used to provide alternate suggestions for words as the author enters text. In addition, it can also help detect errors in verb tense changes in the same paragraph, suggesting proper tense verbs when an abrupt change in diction is detected.
For initial tone settings, when an author uses several words normally associated with formal writing and writes the word, "small"; the system highlights the word. When the user clicks on the word, the system presents a sorted list of suggestions for synonyms in decreasing order of formality. In this example, if the alternatives are, "minuscule, tiny, diminutive", then the order of the suggestion list might be, "diminutive, minuscule, tiny".
For initial tone settings, in a preferred embodiment of the present invention, the system gathers information learned from past emails with a given recipient as a starting value until the analysis of the document determines the tone.
The system uses NLP to analyze the new document and determine a tone for it. Tone is defined as having the majority of words located in the same list of words mainta...