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

[chandramouli]

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 about the process...the requirement is to seperate the themes and then do the sentimental analysis on them...can someone please throw light on how is to be done in power BI

[Eric_Zhang · Accepted solution]

@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"
)

[Eric_Zhang]

@chandramouli

Is there any column indicating what the row is about in your records? Please give some sample and be more specific.

[chandramouli]

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 purposes	312
Managers in my business unit have a lot of work	132
The R&D business unit of my company is pretty good	101
My manager gives me a lot of work	345
My company has good infrastructure for research	456
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 patents	906
All Managers at my company are well trained	567

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 · Accepted solution]

@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]

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

[Eric_Zhang · Accepted solution]

@chandramouli

You can download the WordCloud visual.

[chandramouli]

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 purposes	312
Managers in my business unit have a lot of work	132
The R&D business unit of my Company is pretty good	101
My manager gives me a lot of work	345
My Company has good infrastructure for research	456
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 patents	906
All Managers at My Company are well trained	567

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...