You are on page 1of 4

REAL - TIME INCREMENTAL SHORT

TEXT SUMMARIZATION OF
COMMENT STREAMS IN SOCIAL
NETWORKS
Abstract
This paper presents RISTS, a summarization system to support real-time short text
summarization of comment streams in social networks. Now-a-days, the popularity
of social networking services has increased tremendously, so the quantity of
comments can increase at a high rate immediately after a social message is
published. The users of the social sites always desire to get a brief understanding of
a comment stream without reading the whole comment list. Here the system group
comments with content similarity, semantic similarity and generate a concise
opinion summary for the message. Since different users can request the summary
at any moment, existing clustering methods cannot be directly applied because
they cannot meet the real-time need of such application. To provide immediate and
instant summary of real time comment streams, makes use of IncreSTS algorithm
which can incrementally update clustering results with latest incoming comments
in real time and simultaneously outputting signicant comments that are closest to
the center of each cluster. Moreover the comments and summaries are ranked
based on their relevance. Then design a visualization that consists of basic
information, key-term clouds, and representative comments. This at-a-glance
presentation enables users to get an overview understanding of a comment stream
easily and quickly get an overview summary.
Introduction
In recent years, social network services (SNS) are prevalent and have become
important communication platforms in our daily life. Due to the popularity and
convenience of these platforms, celebrities, corporations, and organizations also set
up social pages to interact with their fans and the public. For each message, users
are able to express their opinions by forwarding, giving a like, and leaving
comments on it. As per studies and observations, not only the quantity of
comments is large, but also the generation rate is remarkably high. Users
unnecessarily and almost impossibly go over the whole comment list of each
message. However, some users may still desire to know what are other users
talking about and what are the opinions of these discussion participants. With these
motivations, develop an advanced summarization technique called RISTS targeting
at comment streams in SNS. RISTS, is a summarization system to support real-
time short text summarization of comment streams in social networks. It target at
comment streams in SNS that are in short text style with casual language usage.
For each social message, the main objective is to cluster comments with content
similarity, semantic similarity and generate a concise opinion summary for this
message. There is a need to discover how many different group opinions exist and
provide an overview of each group to make users easily and rapidly understand.
Therefore, here the goal is developing an efficient and effective technique to
identify the clusters of these comments. This report describe the various related
works which contribute the development of the RISTS: real-time short text
summarization of comment streams in social networks. The rest of the report is
organized as Chapter 2 summarizes the related works, Chapter 3 present problem
definition, chapter 4 give the details of RISTS and chapter 5 is the status of the
project.

System Architecture
Figure shows the system architecture of RISTS: Real-Time Incremental Short Text
Summarization of Comment Streams in Social Networks. Once a message is
posted on SNS, users can leave comments immediately and the number of
comments may rise quickly and continuously. Moreover, readers are usually
unwilling to go over the whole list of comments, but they may request to see the
summary at any moment. This indicates that the RISTS approach should be able to
generate the summary result at any time point of a dynamic data stream. To satisfy
this requirement, here model this problem as an incremental clustering task.

The System architecture of RISTS adopt the term vector model, and therefore each
comment is transformed into a set of n-gram terms by the NLP module. Since
informal and unstructured texts are widely used on SNS, and also apply some
heuristics to enhance the quality of n-gram terms that can better represent each
comment. Here a semantic similarity finder is used to check the semantic similarity
between comments. To decide whether two words are semantically similar, it is
important to know the semantic relations that hold between the words. For
example, the words horse and cow can be considered semantically similar because
both horses and cows are useful animals in agriculture. Similarly, a horse and a car
can be considered semantically similar because cars, and historically horses, are
used for transportation. To compute the semantic similarity between two words a
relational model is used. First, using snippets retrieved from a web search engine,
an automatic lexical pattern extraction algorithm is used to represent the semantic
relations that exist between two words.

Whenever a request is received, the IncreSTS algorithm incrementally producing


latest clustering results and simultaneously outputting signicant comments that are
closest to the center of each cluster. Moreover the comments and summaries are
ranked based on their relevance. Finally, for the visualization interface,
representative terms will be extracted to form a key-term cloud for each group.
Thus, users will be provided a concise, informative, and at-a-glace presentation
that can help them easily comprehend the main points of responses to one message
on SNS.

You might also like