Natural language processing

nlp

#1

Hi Folks,

Me and my few team mates have decide to build a project on NLP using python, it will be sentiment analysis on patient chat conversation to predict their depression level. Hence, we are looking for good source from where we can begin and start learning or any similar project as a reference point would be helpful, even suggestions are most welcome.


#2

I had done similar kind of project which was sentiment analysis of news articles. In order to provide you with suggestions on such type of problem, I need to ask you few questions:
1)Do you have a labelled dataset of patient chat conversation against their depression level?
2)And how large it is ?
3)Approx how long is each conversation(no. of words)
4)Are you considering features other than patient’s chat conversation eg. time of conversation etc?


#3

Best solution i have come across is to build CNN classifier for sentiment analysis. But this is good in predicting +ve or -ve sentiments. Please check out “devashish shankar flipkart” on youtube for some idea. There is coreNlp library available in R operation “getSentiment” built on top of Stanford library to classify sentences into positive and negative. The output will be good for short sentences.


#4

Hi Ashish,

Thanks for reply, sorry i am bit late in responding. My answers in-line

1)Do you have a labelled data set of patient chat conversation against their depression level? - No, we don’t have any chat history of patient as this is new project so we don’t have anything in place. We are suppose to design interactive chat pod which will converse with patient and predict, though we are planning to train corpus based on the data that we will collect from internet or other sources
2)And how large it is ? - NA
3)Approx how long is each conversation(no. of words) - NA
4)Are you considering features other than patient’s chat conversation eg. time of conversation etc? - NA

Note: Company want us to come up with this utility which will converse with patient and predict mood, previously a small question and answer system is there, which ask few specific questions to patient and based on their response it define mood of patient.