Forum Discussion

chandramouli's avatar
chandramouli
Helper I
9 years ago
Solved

identifying themes and doing sentiment analysis in text analytics using Power BI

Suppose if there are about 8000 ought records and the responses do not have a central theme ...such that some of the responses aare about a particular product, some about the infrastructure, some abo...
  • Eric_Zhang's avatar
    Eric_Zhang
    9 years ago

    chandramouli

     

    In this case, try to create a calculated column. If the real data is not as cleaning as the sample, you may need some extra data cleaning work(not in Power BI). The best option would be classifying the comment when inputing, eg using some droplist having the themes and save the themes along with comment.

    themes = 
SWITCH (
    TRUE (),
    TRUE ()
        && SEARCH ( "manager", Table[Comment],, 0 ) > 0, "manager",
    TRUE ()
        && SEARCH ( "infrastructure", Table[Comment],, 0 ) > 0, "infrastructure",
    TRUE ()
        && SEARCH ( "R&D", Table[Comment],, 0 ) > 0, "R&D"
)

chandramouli's avatar
chandramouli
Helper I
9 years ago

Hi Eric,

 

Please below the sample file that I am talking about 

 

Comment Employee ID 
My manager has good knowledge on statistics 101
The  infrastructure of my company is sufficient for all purposes312
Managers in my business unit have a lot of work132
The R&D business unit of my company is pretty good101
My manager gives me a lot of work345
My company has good infrastructure for research456
Managers in my business unit follow processes strictly 836
My Company has excellent infrastructure for development 98
The R&D business unit of my company has a lot of opportunity to develop new patents906
All Managers at my company are well trained567

 

Here I want PowerBI to classify the topics infrastructure, manager and R&D andd provide me sentiment analysis on the same...Please let me know how should I  go about it 

Eric_Zhang's avatar
Eric_Zhang
Microsoft Employee
9 years ago

chandramouli

 

In this case, try to create a calculated column. If the real data is not as cleaning as the sample, you may need some extra data cleaning work(not in Power BI). The best option would be classifying the comment when inputing, eg using some droplist having the themes and save the themes along with comment.

themes = 
SWITCH (
    TRUE (),
    TRUE ()
        && SEARCH ( "manager", Table[Comment],, 0 ) > 0, "manager",
    TRUE ()
        && SEARCH ( "infrastructure", Table[Comment],, 0 ) > 0, "infrastructure",
    TRUE ()
        && SEARCH ( "R&D", Table[Comment],, 0 ) > 0, "R&D"
)

  • chandramouli's avatar
    chandramouli
    Helper I
    9 years ago

    Hi All.

     

    I am trying to do the sentiment analysis for the following statements but trying to categorize them first by keywords and then do the analysis. Please find below the sample.

     

    Comment Employee ID 
    My manager has good knowledge on statistics 101
    The  infrastructure of my company is sufficient for all purposes312
    Managers in my business unit have a lot of work132
    The R&D business unit of my Company is pretty good101
    My manager gives me a lot of work345
    My Company has good infrastructure for research456
    Managers in my business unit follow processes strictly 836
    My Company has excellent infrastructure for development 98
    The R&D business unit of My Company has a lot of opportunity to develop new patents906
    All Managers at My Company are well trained567

     

    So I want to categorize by keywords "manager", "R&D", "Infrastructure" and then do the sentiment analysis . I am trying to use the nested if statement for the same... So I need to search for these words for each line and then do the sentiment analysis. Can some one please help me with the same...

  • chandramouli's avatar
    chandramouli
    Helper I
    9 years ago

    Hi Eric,

     

    Thanks a lot for your help!.. This works.:) Excited....Also can you please let me know if it is possible to create Word clouds in Power BI..

     

     

    Thanks,

    Arun