search

nnguyen682 / CECS491ATeam4

Team4Alexa
1 stars 3 forks source link

Enhance the Advice Feature #41

Open AnthonySMyers opened 5 years ago

AnthonySMyers commented 5 years ago

Advice feature for Good Vibes will now read from the DynamoDB Table to give advice that will relate to the user in a more unique way.

jespinoza17 commented 5 years ago

Last night I added a resource to GitHub that may have a lot of useful information: https://www.rexulti.com/us/mdd/about-depression/?ceid=9185&utm_source=bing&utm_medium=cpc&utm_campaign=unbrandedmdd_dtc&utm_term=dealingwithclinicaldepression&utm_content=unbranded_textad The link is to an antidepressant product, but we will not be promoting or encouraging users to take a specific product. What I liked about the link was the feature that allowed for a user to seek professional help. We may be able to implement a similar capability that will work in conjunction with the device address API.

nnguyen682 commented 5 years ago

I'm trying to add Intension slots and It will save the name of the user in the DynamoDB table. So the user won't feel like they have to start everything over again when they are coming back

jespinoza17 commented 5 years ago

I will add meditation and other activities that don't necessary involve a doctors prescription or diagnosis. These new suggestions will just be things that are overall good for the mind and spirit.

jespinoza17 commented 5 years ago

I found a sentiment analysis API that we may be able to use to enhance our advice and discussion features. Instead of manually inputting trigger words, the API will be able to categorize the user's response as good, bad, or neutral. https://apiant.com/connect/Dandelion-Sentiment-Analysis-to-Amazon-Alexa-Smart-Home-Skills

jespinoza17 commented 5 years ago

https://www.ibm.com/watson/services/natural-language-understanding/?S_PKG=AW&cm_mmc=Search_Google-_-Watson+Core_Watson+Core+-+Discovery-_-WW_NA-_-+sentiment++analysis_Broad_&cm_mmca1=000000OF&cm_mmca2=10000409&cm_mmca7=9060029&cm_mmca8=aud-382859943522:kwd-19587244836&cm_mmca9=20b157a8-0cdd-403b-9642-74bb57892181&cm_mmca10=260751885828&cm_mmca11=b&mkwid=20b157a8-0cdd-403b-9642-74bb57892181|1081|15901&cvosrc=ppc.google.%2Bsentiment%20%2Banalysis&cvo_campaign=000000OF&cvo_crid=260751885828&Matchtype=b&gclid=Cj0KCQjw6fvdBRCbARIsABGZ-vRSuS0gyakc4SQ_8zhUJAHT5nyfLbNvg_AC5xEFPAobQT85Bop1Q1EaAnkvEALw_wcB

jespinoza17 commented 5 years ago

Watson ^

jespinoza17 commented 5 years ago

https://github.com/thisandagain/sentiment

jespinoza17 commented 5 years ago

After some trial and error, I realized you cannot store a user response using the AMAZON.literal slot value. Since we cannot use the literal, we cannot implement sentiment analysis in the user input. We may be able to implement sentiment analysis on the the slot values we have stored but that would be redundant since we already have logic implemented for bad, good, and neutral responses from the user based on key words.

jespinoza17 commented 5 years ago

We can add even more advice to our growing list since all the advice is now stored in a .js file and it will not make our index.js longer than it needs to be.