(1/3) Free distribution of NetMiner Plug-in(an Automatic analysis program)
SNA newsletter from Cyram Inc.,
Since established in year 2000, Cyram has been specialized in Social Network Analysis. In this edition, we’ll introduce Cyram’s ‘Plug-in Service’. It is a tool on NetMiner to perform ‘Automatic Topic Analysis’ for the text data of SNS, News and Papers, and to
produce the output in various form of visualization. It is now being delivered free of charge for users.
Topic Modeling is one of the machine learning techniques. It allows you to identify topics embedded in document data and then to classify the documents by them.
In addition, it enables you to extract words cluster that make up a topic. Because this method can be utilized in conjunction with SNA's 2-mode network data, many researchers and analysts are utilizing it in order to analyze unstructured texts in richer way.
□ What is “Automatic
Topic Analysis Plug-in”?
With just 2 clicks, you can process
all the following complex topic analysis at once.
Automatic Topic Analysis Plug-in extracts topics using the Latent Dirichlet Allocation(LDA) algorithm, one of the topic modeling techniques, and then provides it with a variety of visualizations including main words per topic, document clustering and more.
▶ Download Automated Topic Analysis Plug-in
□ Analysis Process of “Automatic Topic Analysis plug-in”
□ How to add Plug-in onto NetMiner
At first, you need to resister the Plug-in onto NetMiner in order to utilize it.
① Download [Automatic Plugin] Topic modeling.nmx at ‘Download Automatic Topic Analysis Plug-in’ and then copy it to Plug-in Basic folder [C:Usersuser made accountNetMiner].
② After opening NetMiner, click Tools >> Plug-in Manager
③ Click ‘Add Plug-in’ and register the downloaded file ([Automatic Plugin] Topic modeling.nmx)
④ Confirm the registration at Plug-in List and then click ‘Close’ to end
□ Analysis case by utilizing ‘Automatic Topic Analysis Plug-in’
○ Data
As for the word ‘bitcoin’ that caused a major issue from the end of 2017, we collected 6,088 tweets as of February 27, 2018 that
includes the word 'bitcoin'.
○ Analysis Process
① Importing text data(License of Semantic Network edition and above is mandatory)
i. File >> Import Unstructured Text
ⅱ. Import text data and set dictionary (We’ll use the attached excel and dictionary data)
The data to be analyzed are Bitcoin_Tweets.xls, and the dictionaries utilized are Document Filtering, Thesaurus, Defined Words and Exception List. Only nouns are extracted.
② Run Plug-in
i. Tools >> Plug-ins >> [Automatic Plugin] Topic Modeling
ⅱ. Input value of analysis option(Filtering word)
TF-IDF Threshold: The option to remove the words below the value of TF-IDF ⇒ Remove commonly used words
Word length: The option to remove short length words ⇒ Remove 1~2 spell-length words
We entered the above in order to extract only the keywords with TF-IDF threshold value 0.5 or more and the words with more than two letters length. Pressing ‘OK’ button will execute ‘Filtering Word’ and pop up the next option window.
ⅲ. Input the value of analysis option(Discover Topics in your text)
# of Topics : number of topics to be extracted
# of words by topic for visualization: number of words by topic to visualize at Topic–Word network
# of top words in each topic for analysis: number of top words in each topic to be analyzed.
We input above so that we can extract 6 topics with 7 key words each, which are to be displayed at Topic-Word network. And as
for the number of words to be analyzed, we set top 100 per each topic according to the allocation probability.
ⅳ. Final Result
Upon completion of automatic analysis by Plug-in, you will see at first two visual outputs. One is ‘word cloud’ for topic 1 and the
other is topic-words network.
<Word cloud for topic 1> <topic-words network>
As a result of automatic analysis, you see a newly created Workfile at the right-bottom area of screen.
Continued in the next post.