A Semantic Triplet Based Story Classifier

Abstract

A story is defined as "an actor(s) taking action(s) that culminates in a resolution(s)." In this paper, we investigate the utility of standard keyword based features, statistical features based on shallow-parsing (such as density of POS tags and named entities), and a new set of semantic features to develop a story classifier. This classifier is trained to identify a paragraph as a "story," if the paragraph contains mostly story(ies). Training data is a collection of expert-coded story and non-story paragraphs from RSS feeds from a list of extremist web sites. Our proposed semantic features are based on suitable aggregation and generalization of <Subject, Verb, Object> triplets that can be extracted using a parser. Experimental results show that a model of statistical features alongside memory-based semantic linguistic features achieves the best accuracy with a Support Vector Machine (SVM) classifier.

Utilisation of Open Intent Recognition Models for Customer Support Intent Detection

“…OIR is a new field in NLP focusing on the extraction and categorise of intentions from natural language statements using semi-supervised or unsupervised methods; this extends the intent analysis field which often requires expert domain knowledge and supervised training data for models to produce accurate results, creating inflexible models tailored to specific domains [6]. Extracted intents could then be used in dialogue systems to categorise statements [6,7] and summarise large corpuses of natural language [8] without reliance on prior subject knowledge or supervised model training, pushing the boundaries of natural language understanding by machines.…”

“…Utterance pruning has to occur prior to extract, in which pronouns are resolved and linked to subjects to remove subject ambiguity and all verbs are converted to their root forms and similar meaning verbs are reduced using a VerbNet [8]. Extraction methods commonly iterate through utterances with grammatical rules to extract triplets, aiming to identify singular or multiple triplets based on identified subjects within the utterance.…”

“…Extraction methods commonly iterate through utterances with grammatical rules to extract triplets, aiming to identify singular or multiple triplets based on identified subjects within the utterance. Ceran, et al [8]'s work identified different events linked to each subject and extracted triplets using a semantic role labeller (SRL) and triplet matching rules, allowing for complex sentence analysis and multiple events contextually linked to multiple subjects. Rusu, et al [11] used Treebank parsers which linked each word contextually to the subject, predicate or object in a sentence in a tree structure, and formatted triplets based on the resulting tree for each utterance, only extracting triplets from simple sentences with singular subjects.…”

“Climate Change” Frames Detection and Categorization Based on Generalized Concepts

“…Others, [7] employed POS-tags [47] and named entities [23] as features to detect and classify frames. Ceran et al [16] experimented with fsubject, verb,objectg based features and benchmarked``paragraph level" classi¯er for story detection against standard keyword based features, which showed signi¯cant improvement in classi¯cation accuracy. More advanced conceptual features engineering was developed in [15] as they showed how generalized concepts performed better in detecting stories in paragraphs.…”

Folktale Classification Using Learning to Rank

“…We aim to capture these using verb(subject, object) (SVO) triplets, such as lives(princess, castle) or partial triplets such as disappear(driver,). Recently, triplets have been explored to distinguish between stories and non-stories [8]. Triplets are much sparser than just words; we therefore explore allowing partial matches, and abstraction of verbs to a higher semantic level using VerbNet [14].…”