Topic Detection in Group Chat Based on Implicit Reply
Topic detection in group chat has become a promising research due to the widely usage of Instant Messaging (IM) systems. Previous works mainly focus on improving the text similarity between two related messages by utilizing different weighting factors. Ho
- PDF / 584,458 Bytes
- 8 Pages / 439.37 x 666.142 pts Page_size
- 18 Downloads / 190 Views
Abstract. Topic detection in group chat has become a promising research due to the widely usage of Instant Messaging (IM) systems. Previous works mainly focus on improving the text similarity between two related messages by utilizing different weighting factors. However, the text similarity of related texts is likely to be zero (or near zero) due to the characteristics of short text messages in group chat. To solve this problem, an innovative topic detection method based on implicit reply which indicates chat messages interact with each other is proposed in this paper. The comparative experiments results on the datasets gathered from QQ groups demonstrate the superiority of the proposed method as compared to the baseline approaches. Keywords: Topic detection
1
· Group chat · Multi-topic window
Introduction
IM systems are becoming more and more popular among netizens from all walks of life. Online group chat as an import service supported by them brings great convenience for the communication among multiple people. However, due to the high speed of message released and meaningless chatting, group chat logs are filled with large amount but not necessarily useful messages. Topic detection in group chat becomes a significant but challenging research task. Although topic detection techniques have been well studied in traditional texts [1], they are not applied to group chat texts because of the brief of the chat texts. Meanwhile, unlike the asynchronous nature of microblog texts, chat text streams are synchronous, which result in topic detection methods for microblog mainly based on word frequency [2] are also not applied for group chat texts. The biggest challenges to topic detection in group chat is the sparse eigen-vector of short text messages. Existing algorithms mainly focus on improving the text similarity between two related messages by utilizing different weighting factors to alleviate the sparsity. However, the contexts in group chat considered as weighting factors are not always reliable due to the uncertain changes of group chat features in different groups. In this paper, an innovative topic detection method based on implicit reply features is proposed to solve the above problems. Messages with reply relations c Springer International Publishing Switzerland 2016 R. Booth and M.-L. Zhang (Eds.): PRICAI 2016, LNAI 9810, pp. 673–680, 2016. DOI: 10.1007/978-3-319-42911-3 56
674
X. Zhang et al.
judged by the features are grouped together as a long text to overcome the challenge that the text similarity of two related messages is too low.
2
Related Work
In recent years, many studies have been done for group chat texts analysis. Uthus et al. [3] did a survey on this. According them, topic detection in group chat is one of the high-level researches. Existing methods for topic detection in group chat can be roughly divided into two categories: supervised [4] and unsupervised. In this paper, unsupervised methods are focused on. Shen et al. [5] represented the messages by a vector space model. They used the si
Data Loading...
Recommend Documents
A User Group Classification Model Based on Sentiment Analysis Under Microblog Hot Topic
171 103 90MB
A deceptive detection model based on topic, sentiment, and sentence structure information
163 40 1MB
