VNU-UET Repository

Intent Extraction from Social Media Texts Using Sequential Segmentation and Deep Learning Models

Luong, Thai Le and Cao, Minh Son and Le, Duc Thang and Phan, Xuan Hieu (2017) Intent Extraction from Social Media Texts Using Sequential Segmentation and Deep Learning Models. In: The 9th International Conference on Knowledge and Systems Engineering (KSE 2017), October 19-21, 2017, Hue city, Vietnam.

Full text not available from this repository.


Nowadays, users are much more willing to share their daily activities, their thoughts or feelings, and even their intentions (e.g., buy an apartment, rent a car, travel to somewhere, etc.) on online social media channels. Understanding intents of online users, therefore, has become a crucial need in many different business areas like production, finance/banking, real estate, tourism, e-commerce, and online marketing. In this paper, we will present our solutions to extract intent information from online social media texts. This can be seen as an information extraction or sequential segmentation task. In order to perform this task, we have built our machine learning models based on conditional random fields (CRFs), an advanced statistical graphical model for sequence data, and bidirectional long short–term memory (Bi–LSTM), a well-known deep learning model. To evaluate our methods, we have defined two intent extraction tasks in two domains: real estate and cosmetics & beauty. For each task, we have defined the tag set as well as prepared the labeled data set that consists of Vietnamese social media text posts with the annotation of intent words/phrases. Experimental results showed that the proposed methods can effectively extract intent information from online texts with significantly high accuracy.

Item Type: Conference or Workshop Item (Paper)
Subjects: Information Technology (IT)
Divisions: Faculty of Information Technology (FIT)
Depositing User: A/Prof. Xuan Hieu Phan
Date Deposited: 30 Oct 2017 04:19
Last Modified: 27 Nov 2017 04:29

Actions (login required)

View Item View Item