Please use this identifier to cite or link to this item:
Title: Augmenting the interpretation of emoji in translation using machine learning : a study in the local context
Authors: Agius, Keith
Keywords: Emojis
Machine learning
Discourse analysis -- Data processing
Language and emotions -- Malta
Issue Date: 2018
Citation: Agius, K. (2018). Augmenting the interpretation of emoji in translation using machine learning: a study in the local context (Bachelor's dissertation).
Abstract: Emoji are graphically-rich characters embedded in nearly every form of online textual communication. They act as a replacement for non-verbal cues that occur in face-to-face communication, which are extremely important elements in a normal conversation. When emoji accompanying a message are misinterpreted, the meaning and sentiment of the message are incorrectly translated, and this can potentially have negative effects on the relationship of the involved parties. With businesses increasingly using online chat environments to interact with their customers, such misinterpretation could prove to be detrimental to customer relationships. Obvious differences exist between emoji renderings of varying platforms, and this is a well-known cause for misinterpretation. In contrast, this dissertation explores interpretations that are directly attributed to an individual’s demographic and personal characteristics, defined as within-platform misinterpretation. Through a quantitative survey distributed in the Maltese local context, an instantmessaging environment was simulated. This allowed for the collection of emoji interpretation data, paired with the respondents’ characteristics. The simulation of such an environment provided the optimal scenario for within-platform misinterpretation to be studied in. After optimizing the collected data, machine learning techniques were applied to the dataset in an attempt to predict interpretations based on the given characteristics. Testing was done on different versions of the dataset, with each iteration possessing a varying attribute structure. This identified the granularity with which characteristics must be collected in order to obtain more reliable predictions. Results provide valuable insight into how such misinterpretation can occur realistically. This research also concludes that emoji interpretations can be reliably augmented using machine learning techniques, and hence, there is solid groundwork for future work to incorporate such augmentation within various applications. This can prove useful to businesses in achieving optimal customer experiences; with emoji being appropriately used for every situation depending on the customer profile, misinterpretation can be significantly reduced, resulting in the strengthening of business-to-customer relationships.
Description: B.SC.BUS.&I.T.
Appears in Collections:Dissertations - FacEma - 2018
Dissertations - FacEMAMAn - 2018

Files in This Item:
File Description SizeFormat 
  Restricted Access
2.56 MBAdobe PDFView/Open Request a copy

Items in OAR@UM are protected by copyright, with all rights reserved, unless otherwise indicated.