“There’s No Way I Would Ever Buy Any Mp3 Player with a Measly 4gb of Storage”: Mining Intention Insights about Future Actions

Abstract

In this paper we present a method for the automatic detection of userstated intentions in terms of desires, purposes and commitments as specific insights deriving from the semantics of the intention expressions. The method is based on a linguistic data-driven and domain-independent framework for textual intention analysis and achieves substantial levels of accuracy in detecting future intention expressions and their structural components. Furthermore, we demonstrate several usage scenarios in the business intelligence context showing that the introduced insights can be interpreted from various perspectives and serve as variables in predictive or decision making models in any domain of interest.

Split Over-Training for Unsupervised Purchase Intention Identification

“…Cognitive or robotic research often adopt one or both of the first two renditions, and focus on the process of becoming aware of the intention of another agent or inferring an agent's intention through its actions and their effects on the environment [6], [12]- [14]. Social media or web-based research usually take the last rendition, and try to detect intention as practical attitude from texts [1], [5], [15]. Our research complies with the last rendition.…”

“…For an utterance, there are usually "primary" and "secondary" illocutionary acts simultaneously, with the primary also called an indirect speech act and the secondary a direct one [17]. Some research classify emails or sentences directly into different speech acts [7], [18], [19], [20],while others and our present work focus on the speech acts that communicate future intentions [1], [5], [15]. Specifically, we regard linguistically instantiated or textual intentions as indirect illocutionary speech acts with special reasons or practical attitudes.…”

“…Existing research on intention identification or analysis from natural language involve different social media platforms, various textual genres, and a diversity of intention categories. Mostly covered social media includes Facebook [22], Yahoo [8], Twitter [15]and Amazon [5]; textual genres include emails [7], chat rooms [23] and message boards [18]; intention categories vary from two simple ones as commercial versus non-commercial [24] to very complicated 136 intention types from a social-psychological framework [8].…”

“…GoJek and Grab) [4] etc. The ability of recognizing and understanding intentions from texts is also crucial for many domains like business intelligence [1], cyber security [5], and industrial robotics [6]. For instance, the tweet "Anyone have a suggestion in Paris for a 2-night getaway idea?…”

“…Existing research about intention identification are either based on acquired linguistic knowledge of syntactic and/or semantic patterns [7], [8] or manually labeled intention corpus [1], [5]. To our knowledge, there is no large-scaled labeled social media intention corpus or linguistic knowledge database publicly available.…”