Enabling chatbots by detecting and supporting affective argumentation
First Claim
1. A method for determining argumentation in text, the method comprising:
- accessing text comprising fragments;
creating a discourse tree from the text, wherein the discourse tree comprises a plurality of nodes, each nonterminal node representing a rhetorical relationship between at least two of the fragments and each terminal node of the nodes of the discourse tree is associated with one of the fragments;
matching each fragment that has a verb to a verb signature, thereby creating a communicative discourse tree;
determining whether the communicative discourse tree represents text that comprises affective argumentation by applying, to the communicative discourse tree, a classification model trained to detect a presence of affective argumentation; and
responsive to determining that the text comprises affective argumentation, accessing a response that corresponds to the text and outputting the response.
1 Assignment
0 Petitions
Accused Products
Abstract
Systems, devices, and methods of the present invention detect affective argumentation in text. In an example, an application executing on a computing device accesses text comprising fragments. The application creates a discourse tree from the text. The discourse tree includes nodes, each nonterminal node representing a rhetorical relationship between two of the fragments and each terminal node of the nodes of the discourse tree is associated with one of the fragments. The application matches each fragment that has a verb to a verb signature, thereby creating a communicative discourse tree. The application determines whether the communicative discourse tree represents text that includes affective argumentation by applying a classification model trained to detect affective argumentation to the communicative discourse tree.
71 Citations
20 Claims
-
1. A method for determining argumentation in text, the method comprising:
-
accessing text comprising fragments; creating a discourse tree from the text, wherein the discourse tree comprises a plurality of nodes, each nonterminal node representing a rhetorical relationship between at least two of the fragments and each terminal node of the nodes of the discourse tree is associated with one of the fragments; matching each fragment that has a verb to a verb signature, thereby creating a communicative discourse tree; determining whether the communicative discourse tree represents text that comprises affective argumentation by applying, to the communicative discourse tree, a classification model trained to detect a presence of affective argumentation; and responsive to determining that the text comprises affective argumentation, accessing a response that corresponds to the text and outputting the response. - View Dependent Claims (2, 3, 4, 5, 6, 7, 8, 9)
-
-
10. A method for determining argumentation in text, the method comprising:
-
accessing a set of training data for text in a first domain, wherein the set of training data comprises (i) a first set of communicative discourse trees each comprising an expected classification of a presence of affective argumentation and (ii) a second set of communicative discourse trees each comprising an expected classification of an absence of affective argumentation, wherein each communicative discourse tree comprises a discourse tree with a plurality of nodes, each nonterminal node representing a rhetorical relationship between at least two fragments of text, each terminal node of the nodes of the discourse tree associated with one of the fragments, and wherein each fragment that has a verb is matched to a verb signature; training a classification model to identify affective argumentation by iteratively; providing a communicative discourse tree of either the first set of communicative discourse trees or the second set of communicative discourse trees to the classification model; receiving, from the classification model, a determined classification; calculating a loss function by calculating a difference between the determined classification and the expected classification; and adjusting internal parameters of the classification model to minimize the loss function; creating a second communicative discourse tree for a second body of text in a second domain; obtaining a second classification for the second body of text by applying the trained classification model to the second communicative discourse tree; and responsive to determining that the second body of text comprises affective argumentation, accessing a response that corresponds to the second body of text and outputting the response. - View Dependent Claims (11, 12, 13, 14, 15, 16)
-
-
17. A method of building a dataset of argumentation features with identified classes, the method comprising:
-
accessing text from a first domain, the text comprising fragments; creating a communicative discourse tree from the text by; creating a discourse tree comprises a plurality of nodes, each nonterminal node representing a rhetorical relationship between two of the fragments and each terminal node of the nodes of the discourse tree is associated with one of the fragments; and matching each fragment that has a verb to a verb signature, thereby creating a communicative discourse tree; accessing a positive communicative discourse tree from a positive training data set and a negative training data communicative discourse tree from a negative training data set, wherein the positive training data set comprises communicative discourse trees representing text containing affective argumentation, and the negative training data set comprises communicative discourse trees representing text without affective argumentation and wherein the communication discourse trees for the positive training data set and negative training data set are each based on text from a second domain; identifying whether the communicative discourse tree is from the positive training data set or the negative training data set by applying a classification model to the communicative discourse tree; adding the communicative discourse tree to either the positive training data set or the negative training data set based on the identified class; training a classification model, with the positive training data set and the negative training data set; applying the trained classification model to additional text; and outputting a classification received from the trained classification model. - View Dependent Claims (18, 19, 20)
-
Specification