Free distribution of NetMiner Plug-in(an Automatic analysis program) (2/3)

○   Understanding of Analysis output

①     Document classification : # of documents classified by each topic and how much each topic shares in total number of documents

 ※ The label of each topic can be named subjectively by the researcher after figuring out is included words

 When you open the workfile_“Filtered Words” and then “LDA” session, you will catch up followings.

-   Proportion of topics occupied in each document(Actually the probability of topic distributed in a document) at ‘[T] Document Classification’ tab at the bottom)

-   Which topic shares the most proportion in each document

-   Number of documents classified by topic. It can be regarded as the density of each topic.

º Interpret the result

- When we classify tweets regarding Bitcoin into 6 topics, each topic occupies a similar proportion of about 17% in the total 6,000 tweets 

º Where to find out the output

- Workfile_Topic Proportion >> Frequency(Vector)

 ③     Major words per topic(Word cloud) : based on allocation probability

-   The word with higher allocation probability in a topic displays bigger than others.

-   Only the meaningful one is the size of word. Others like color or position are meaningless.

º Where to find out the output

-   Workfile ‘Filtered Words’ >> ‘Tools’ at main menu >> ‘Plug-ins’ >> run ‘Word Cloud’ >> in ‘Select Node Attribute’, pick up ‘Prob_Topic_(number)’ and then ‘OK’

* Prob_Topic_(number) : The topic number to check

④     Words network map per topic(+see original text)

<Words network map per topic>

       -   The words network consists of top 100 words which shares higher allocation probability in a topic. Dots in red( ) correspond to each word.

-   Words network is created according to how close a word appears to other words and documents(Word distance – window size)

-   Link is created between words which appear proximately.

-   The higher allocation probability a word has in a topic, the bigger the size of dot(  ) becomes.

º Where to find out the output

-   Under Workfile Topic_(number), Result[Documents, 3, 1, D] >> Spring 2D

<See original text where the word(Venezuela) appears>

  -   You can check in which sentences or documents a specific word appears.

          º Where to find out the output

   -   Choose one or more specific keywords within network map >> ‘Tools’ at main menu >> ‘Plug-ins’ >> ‘Word in a sentence’

* Note that it is executable only when network map is open

* If 2 or more words are selected and run, only the sentences display which include them both. Nothing will return on the result screen in case no sentences meet the condition. 

* When you select 2 or more words, click them while pushing Ctrl key. 

⑤     Topic-Keywords network: Visualize the network between the extracted topic and its main constituent words

 -  The words of higher influence on topic are visualized. Words are selected in topic analysis.

 -  Dots  in red(  ), in blue(  ) mean topic and its main constituent words respectively. Letters in yellow describes the topic name subjectively assigned by the researcher after referring to constituent words.

 -  Line width corresponds to the allocation probability of main words(  ) to its topic( ). The higher it is, the thicker it displays

        º Interpret the result

-   USA Senate hearing about Tether: On February 6, 2018, the US Senate Banking Committee held a hearing session on the virtual currency by inviting the chairman of the Securities and Exchange Commission (SEC), and chairman of the Commodity Futures Trading Commission (CFTC). Therefore many tweets were created that include related keywords regarding the subject. 

-  Venezuela’s crypto platform development: As the news come out that Venezuelan President has proposed to develop a platform for oil-based crypto trading, many tweets were created that include related keywords to the news.

-   European commission Loves Block chain’s Transparency: The news that the European Union is paying attention to the block chain technology has produced a lot of tweets that contain related keywords.

º Where to find out the output

-  Workfile LDA_Result >> Spring 2D

Continued in the next post.

>> Check out the next post!