Inspiration
Initally, idea was closely related to the elder people and their trust in web-shops - some of them does not trust it because of having other customer habits. So, we decided to create a voice-assistant for them, such that they would more preferably would buy things in the web-shops, however after some consideration we understood that it may be useful not only for the elder people.
What it does
When placed on the Shopify shop, it has access to all the goods from it and may help person to find thing which he/she actually needs by analyzing their reactions and specifiactions about given products. In future it hopefully be able to recommend which present should be bought for a person according to info about his/her habits etc
How I built it
At first we needed to create a development shop, so that we have information about it and goods inside it, simulating real shop. After that, we needed to implement Speech-to-Text - NLP - Text-to-Speech pipeline. Speech to Text and reverse were done via IBM tools, as well as part of the NLP. However, the most interesting task was to understand which information from the answers is the most important one
Challenges I ran into
Most interesting challenge was processing NLP data in order to extract user preferences. Now it is done via several heuristics. For example, during the first request from the bot only noun information is extracted.
What's next for Online shopping voice assistant
We will get rid of the heuristics and create something much stronger maybe with the use of GPT-3 license, or more basic methods like TF-IDF in order to find important word into the request.
Log in or sign up for Devpost to join the conversation.