Mini Project

You might also like

Download as pptx, pdf, or txt
Download as pptx, pdf, or txt
You are on page 1of 16

g e

ra s
v i ve e w
D e - C o f N
id e a la ig h
n g o
a t h
E
u m t H g i a n S
G .K a e n a g s w - C
ic l T V i a r
. M Effi c ia e s ta l ye
rs r : S o
ti c l ka na
M ge a r e n F i
ag .V h.
hT CH ec
a s B .T
H
Timeline flow

Abstract Problem Model Conclusio Future


statement n work
Abstract :
News and social media now play a synergistic role and
neither domain can be grasped in isolation. On one hand, platforms
such as Twitter have taken a central role in the dissemination and
consumption of news. On the other hand, news editors rely on
socialmedia for following their audience’s attention and for crowd-
sourcing news stories.
Twitter hashtags function as a key connection between
Twitter crowds and the news media, by naturally naming and
contextualizing stories, grouping the discussion of news and
marking topic trends.
Problem Statement :
The aim is to map a stream of news articles to a stream of
Twitter hashtags, in real-time. Given a news article, how
quickly can we recommend hashtags which are relative to that
news article.
A word or phrase preceded by a hash mark (#), used within a message to
identify a keyword or topic of interest and facilitate a search for it.
Whenever a user adds a hashtag to their post, it's able to be indexed
by the social network and becomes searchable/discoverable by other users.
Did you know that 
71 percent of social media users regularly make
use of hashtags?
1. To start a topic or discussion.
2. To start a competition.
3. Host and Join chats.
4. Participating in trending topics.
5. Raising awareness about a subject.
6. Community building.
7. Gain more followers etc…
The recommendation model has two phases .

1.Data Collection

2. The choice of
Modelling
Data Collection :

1. Query Generation - Take a Pseudo-Article for this step

POS+T RAKE Textran Alchem


F.IDF k y
API
Rapid Graph Tool by
Python Automatic based IBM
Keyword keyword
Extraction extraction
2. Cold Start Search Algorithms

Recent Past Similar No


Tweets Stories Articles ColdStart

Streaming
Search Search k
Search Live Tweets
past most
recent fromTwitter
articles similar Streaming
tweets from articles API
clusters
Algorithm 2. ColdStart1: Searching Recent Tweets

Input: Query Qa extracted from article a, database of past articles D


Output: Relevant tweets Ta search, relevant hashtags Ha search

Method:
Ta_search = { } ;
w = 60 * 12 //Timewindow for past tweets: 12h
Tw = SubsetTweets(D,w) //Retrieve all tweets within time window w.
for tweet belongs Tw do
for phrase belongs Qa do
If phrase belongs tweet then
//If tweet contains the keyphrase phrase.
Ta_search = Ta_search U (tweet)
break
Return Ta_search ;
The Choice of Modelling includes the following steps :

1. After gathering relevant tweets for a given article using


Twitter
Streaming API, calculate Feature-Vectors for each
Article,Tweet pair.
2. The Time-Aware features are considered to get relevant
hashtags to the article considered. The time-aware features are
:
a) Local Similarity
b) Global Similarity
c) Local Hashtag frequency
d) Trending Hashtags etc…
Conclusion and Results :

The conducted experimental results show that for a given


article, the model provides hashtags from streaming twitter api with
high accuracy and high coverage. High-precision refers to the
quality of recommendations. High Coverage refers to how many
articles get recommended within considered time window.
For the query “EU referendum”, this approach returns
related hashtags #Brexit, #EURef, #EU, #VoteLeave, and #Remain.
These five hashtags are the most used in the discussions relevant to
the EU referendum on Twitter. Both #Brexit and #EURef are
specifically used for this story, but because #Brexit is more
self-explanatory, it is popular in both mainstream and social media.
Future Work : This implementation is used in Insight4News.com

Query Generation has always been an important area of study


and a lot of enhancements are going on in that field. The data
collection for computing features plays a key role in the overall
efficiency and effectiveness of the recommender model. Many
models for KeyPhrase extraction for Document Summarizarion
have been implemented like MAUI, supervised ML based
keyphrase extractions etc…
Thank You

You might also like