Online discussion platforms are often flooded with opinions from users across the world on a variety of topics. Many such posts, comments, or utterances are often sarcastic in nature, i.e., the actual intent is hidden in the sentence and is different from its literal meaning, making the detection of such utterances challenging without additional context. In this paper, we propose a novel deep learning-based approach to detect whether an utterance is sarcastic or non-sarcastic by utilizing the given contexts in a hierarchical manner. We have used datasets from two online discussion platform -Twitter and Reddit 1 for our experiments. Experimental and error analysis shows that the hierarchical models can make full use of history to obtain a better representation of contexts and thus, in turn, can outperform their sequential counterparts.
Tbeadvauesiocompuler hardwammmbinedwitbinnovaliveArtwdal Intelllgen~ (AI) techniques can be a powerful "odologv to perform lotelllgent cognitive hskg We have Inmtlgated Speeeh R a o g n " techniques using Hidden Markov M a s and rueees%Uy clasrWed speakers based on tbelr utterancrr This paper proposes a dlraete ProbabUW HMMS based appmach to rurarsrwb. daasily speakers based w their utteram The results dmw that we have got hi& ao~uracy io IdeuMylng the speakererr. This leads us to conclude that Uoaertaio Reasoning and Leadug are vital components of AI that Fwld kad to the development of automated Mdlfgent roluIlonr to various complex and ioterestiog problems Keywords: Dlrrrete RobablUty HMM; Vlterbl Algorttbm; HTK IntrodudionHuman beings have a lot of Criteria that they use to recognize and categorize different kinds of sounds, voices and songs. We are interested to know how much of these can be "taught" to a computer. How much can we teach a computer to recognize and categorize sounds? By extending this idea can a computer system listen to sounds made by a speaker and tell us who the speaker is? Here, we show how the probability of correctness of our classification of speaker utterances depends on many parameters. If there hss to be a good solution, we recognized from the start that the following two conditions must hold 1. 2.There must be a mathematical model of the system. We need tohave a large number of sample sounds with which to train the computer to recognize different sources of sounds.Once we identified these two requirements, we see that the Markov Model is most appropriate for the problem because the Markov Property holds true in our case. The Markov Property states that transition probability from any given state depends only on the current state and not on the previous history If a computer system has heard a large number of sounds made by a speaker, we found that it can then do better at recognizing the speaker when it hears a new sound made by another speaker. This is the marvel of learning, an important component of AI. In fact, learning is a thread that traverses the entire breadth and depth of AI. For the second requirement of having a large number of samples, we have recorded speech samples for five different speakers. The speech samples were recorded in Microsoil Wave format and were then parameterized to be in HTKs special format. We considered the approach of using the Hidden Markov Model Toolkit (HTK) for our problem. HTK is a portable toolkit for building and manipulating Hidden Markov models. HTK is primarily used for speech recognition research although it has been used for numerous other applications including research into speech synthesis, character recognition and DNA sequencing. HTK is in use at hundreds of sites worldwide. We chose HTK as our development system due to its specific features for speech recognition. Discrete Probability HMMsAlthough HTK was designed primarily for builtlug continuous density HMM system$ it also supports discrete density HMMs. Discrete HMMs are particu...
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.
hi@scite.ai
10624 S. Eastern Ave., Ste. A-614
Henderson, NV 89052, USA
Copyright © 2024 scite LLC. All rights reserved.
Made with 💙 for researchers
Part of the Research Solutions Family.