| 主权项 |
1. A computerized text analytics method comprising using at least one hardware processor for:
receiving: (a) a proposition, wherein the proposition is a text sentence, and (b) a plurality of claims, wherein each of the plurality of claims is a text sentence; and performing sentiment analysis of the proposition and the plurality of claims, to determine a polarity of each of the plurality of claims with respect to the proposition, by:
providing a textual proposition template which comprises: (a) a placeholder for a topic, (b) a placeholder for one of: a sentiment verb and a sentiment adjective, and (c) a text in between the two placeholders,analyzing the text sentence of the proposition, to identify a proposition topic and a proposition sentiment towards the proposition topic, wherein said analyzing is by matching the text sentence of the proposition with the textual proposition template,identifying a local claim polarity of each of the received claims with respect to the proposition, by:
extracting a claim topic from the respective claim by applying, to the text sentence of the respective claim, one of: a manually-set function, and a classifier trained using machine learning,identifying a claim sentiment towards the extracted claim topic, at least by applying a technique selected from the group consisting of: Natural Language Processing (NLP), sentiment matching, and sentiment composition, to at least one of (a) the text sentence of the respective claim, and (b) text extracted from a content resources from which the text sentence of the respective claim originated, wherein the identified claim sentiment towards the extracted claim topic comprises one of: a negative claim sentiment and a positive claim sentiment, andcalculating the local claim polarity of the respective claim based at least on: the identified proposition topic, the identified claim sentiment towards the extracted claim topic, the identified proposition topic, and the identified proposition sentiment towards the proposition topic,calculating a pairwise claim polarity agreement score of every pair of the received claims, by detecting at least one of: (a) a discourse connective appearing between the claims of the respective pair, and (b) textual entailment appearing between the claims of the respective pair,determining a global claim polarity of each of the received claims, based on the calculated local claim polarity of the respective claim, and on the calculated pairwise claim polarity agreement score of the respective pair of the respective claim; and outputting a list of the received claims and of an indication of the global claim polarity of each of the received claims. |
