Abstract
Short text representation is one of the basic and key tasks of NLP. The traditional method is to simply merge the bag-of-words model and the topic model, which may lead to the problem of ambiguity in semantic information, and leave topic information sparse. We propose an unsupervised text representation method that involves fusing word embeddings and extended topic information. Following this, two fusion strategies of weighted word embeddings and extended topic information are designed: static linear fusion and dynamic fusion. This method can highlight important semantic information, flexibly fuse topic information, and improve the capabilities of short text representation. We use classification and prediction tasks to verify the effectiveness of the method. The testing results show that the method is valid.
Highlights
With the rise and the widespread use of social media platforms, huge amounts of text data are generated every day
We propose a short text representation method, which is based on weighted word embeddings (WWE) and extended topic information (ETI)
This paper proposes a short text representation method based on the weighted word embedding vector and extended topic information, which consists of three parts: short text semantic feature representation based on WWE and extended topic feature representation based on ETI, and their fusion strategy
Summary
With the rise and the widespread use of social media platforms, huge amounts of text data are generated every day. The text usually contains a lot of information, such as emotions and positions. Text is unstructured data, which leads to timeconsuming and laborious manual analysis. Figuring out how to represent unstructured text as a distributed vector that can be recognized by a computer is very important [1]. Text representation has become more and more important in natural language processing (NLP). A good representation method should fully learn the grammatical and semantic information in natural language and lay a solid foundation for downstream tasks, such as text classification and sentiment analysis [2]. Training deep learning models of text representation through labeled datasets usually requires a lot of manual work [3]. We will focus on the unsupervised learning of short text representation, which includes abstracts, instant messaging, social reviews, etc. We will focus on the unsupervised learning of short text representation, which includes abstracts, instant messaging, social reviews, etc. (the short text studied in this paper mainly refers to the text with a length of no more than 512 words)
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
