Intent Classification of Short-Text on Social Media

Purohit, Hemant; Dong, Guozhu; Shalin, Valerie L.; Thirunarayan, Krishnaprasad; Sheth, Amit P.

doi:10.1109/smartcity.2015.75

Cited by 51 publications

(35 citation statements)

References 27 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…After the speech signal is converted into text by a Automatic Speech Recognition (ASR) system [11], identifying the intention of the user is called Intent Classification. Reference [12] addressed the problem of Intent Classification in a social media set up. They used a Hybrid Feature Representation method to handle data ambiguity.…”

Section: A Intent Classificationmentioning

confidence: 99%

Parallel Intent and Slot Prediction using MLB Fusion

Bhasin

Natarajan

Mathur

et al. 2020

2020 IEEE 14th International Conference on Semantic Computing (ICSC)

View full text Add to dashboard Cite

Intent and Slot Identification are two important tasks in Spoken Language Understanding (SLU). For a natural language utterance, there is a high correlation between these two tasks. A lot of work has been done on each of these using Recurrent-Neural-Networks (RNN), Convolution Neural Networks (CNN) and Attention based models. Most of the past work used two separate models for intent and slot prediction. Some of them also used sequence-to-sequence type models where slots are predicted after evaluating the utterance-level intent. In this work, we propose a parallel Intent and Slot Prediction technique where separate Bidirectional Gated Recurrent Units (GRU) are used for each task. We posit the usage of MLB (Multimodal Low-rank Bilinear Attention Network) fusion for improvement in performance of intent and slot learning. To the best of our knowledge, this is the first attempt of using such a technique on text based problems. Also, our proposed methods outperform the existing state-of-the-art results for both intent and slot prediction on two benchmark datasets

show abstract

Section: A Intent Classificationmentioning

confidence: 99%

Parallel Intent and Slot Prediction using MLB Fusion

Bhasin

Natarajan

Mathur

et al. 2020

2020 IEEE 14th International Conference on Semantic Computing (ICSC)

View full text Add to dashboard Cite

show abstract

“…Ambiguity and sparsity are two particular issues while dealing with mining intentions from social media text that should be addressed along with the health-specific context of the intention of use. 62…”

Section: Future Research Challenges To Extract Relevant Information Fmentioning

confidence: 99%

Challenges to Transforming Unconventional Social Media Data into Actionable Knowledge for Public Health Systems During Disasters

Chan¹,

Purohit

2019

Disaster med. public health prep.

Self Cite

View full text Add to dashboard Cite

Every year, there are larger and more severe disasters and health organizations are struggling to respond with services to keep public health systems running. Making decisions with limited health information can negatively affect response activities and impact morbidity and mortality. An overarching challenge is getting the right health information to the right health service personnel at the right time. As responding agencies engage in social media (eg, Twitter, Facebook) to communicate with the public, new opportunities emerge to leverage this non-traditional information for improved situational awareness. Transforming these big data is dependent on computers to process and filter content for health information categories relevant to health responders. To enable a more health-focused approach to social media analysis during disasters, 2 major research challenges should be addressed: (1) advancing methodologies to extract relevant information for health services and creating dynamic knowledge bases that address both the global and US disaster contexts, and (2) expanding social media research for disaster informatics to focus on health response activities. There is a lack of attention on health-focused social media research beyond epidemiologic surveillance. Future research will require approaches that address challenges of domain-aware, including multilingual language understanding in artificial intelligence for disaster health information extraction. New research will need to focus on the primary goal of health providers, whose priority is to get the right health information to the right medical and public health service personnel at the right time.

show abstract

“…This might require completely new approaches, which should borrow from speech act theory, ontological semantics, and sociolinguistics. Intention analysis (Cui & Mao, 2017;Purohit, Dong, Shalin, Thirunarayan, & Sheth, 2015) is also a very promising field, which combines some of the simplest methods of detecting the modality with the message (affirmation, negation, exclamation, interrogation) with temporal, sentiment, and topic detection (Curth, 2018). Sherin (2013) suggested that to a way to resolve the tension between the computational and manual approaches is to employ computational approaches that verify findings derived from the manual and qualitative approach, not simply to replace manual analysis and reduce labor.…”

Section: Content Analysismentioning

confidence: 99%

Computational communication research

Matei

Kee

2019

WIREs Data Min & Knowl

View full text Add to dashboard Cite

The article advances a new way to think about computational research, in general, and computational communication research, in particular. Contrary to current definitions of “computational science,” which emphasizes its inductive nature, we define computational research as an incomplete inductive process, blending both theoretical and data‐driven methods of discovery. Communication theory needs to be driven by a clear concept of human needs and abilities, recovering and extending known theoretical insights from mass and interpersonal communication research. The definition we propose for computational communication research has a practical implication. Relying on theory, the definition demands to identify specific processes and domains within the field of computational communication research. The processes include communication production, behavior, and effects. The domains include collaboration, trust, and data storytelling and journalism, while the methods include content and network analyses. The article starts with a broad definition of the “computational” approach, using the Johari window. We continue with a typology of computational communication research, which blends reviews of foundational texts with summaries of leading research. In the conclusions, we discuss the strengths and identify new opportunities in the field of computational communication research. This article is categorized under: Fundamental Concepts of Data and Knowledge > Human Centricity and User Interaction Commercial, Legal, and Ethical Issues > Social Considerations Fundamental Concepts of Data and Knowledge > Big Data Mining

show abstract

Intent Classification of Short-Text on Social Media

Cited by 51 publications

References 27 publications

Parallel Intent and Slot Prediction using MLB Fusion

Parallel Intent and Slot Prediction using MLB Fusion

Challenges to Transforming Unconventional Social Media Data into Actionable Knowledge for Public Health Systems During Disasters

Computational communication research

Contact Info

Product

Resources

About