Jump to the main content block
TITLE

[2020-Dec-23] Network Embedding with Textual Information

ImgDesc
ImgDesc
ImgDesc

Speaker:

Prof. Chuan-Ju Wang 王釧茹教授

Associate Professor, Research Center for Information Technology Innovation (CITI), Academia Sinica

Topic:

Network Embedding with Textual Information

Date:

13:20-15:00 Wednesday 23-Dec-2020

Locate

Delta R103

QR code:

Link:

https://teams.microsoft.com/l/channel/19%3a9af56d066d6343c2a28be7f3529466e3%40thread.tacv2/%25E4%25B8%2580%25E8%2588%25AC?groupId=f3cb8647-c58d-48dd-accc-c1bcf9dd0984&tenantId=6c3bc511-43c7-4596-baeb-2335c69c41f1

Hosted by:

Prof. Min-Chun Hu

Abstract

This talk focuses on network embedding with the incorporation of textual information. Specifically, two of our recent related studies will be introduced. The first one is related to item concept modeling, for which an item concept embedding (ICE) framework is proposed to model item concepts via textual information. With the proposed carefully designed ICE networks, the resulting embedding facilitates both homogeneous and heterogeneous retrieval, including item-to-item and word-to-item retrieval. Moreover, as a distributed embedding approach, the proposed ICE approach not only generates related retrieval results but also delivers more diverse results than traditional keyword-matching-based approaches. The second part of this talk introduces a network embedding learning framework for constructing domain-specific sentiment dictionaries from online customer reviews. The proposed framework has three main advantages. First, no additional annotations of words or external dictionaries are needed for the proposed framework; the only resources needed are the review texts and entity ratings. Second, the framework is applicable across a variety of user-generated content from different domains to construct domain-specific sentiment dictionaries. Finally, each word in the constructed dictionary is associated with a low-dimensional dense representation and a degree of relatedness to a certain rating, which enable us to obtain more fine-grained dictionaries and enhance the application scalability of the constructed dictionaries as the word representations can be adopted for various tasks or applications, such as entity ranking and dictionary expansion.

All faculties and students are welcome to join.

Click Num: