INHouse

App

• Visualisation
  • Parsing Linkedin profiles
  • Controlled audio-record
  • Uncontrolled audio-record
  • Additional functionality
  • Parsing documents (PDF)
• Admin Panel
  • User section
  • Manual adding sales Questions / Answers
  • Sales Questions / Answers

Backend

• AUTH flow
• Dialog
• Files
• HumanticAPI
• Metrics
• Parser
• SpeechRecognition
• TextAnalyzer
• User

App

Visualisation

Parsing Linkedin profiles

• user puts a person's Linkedin profile into dedicated input
• user receives all the parsed information about a person gradually. The first portion comes instantly, the second one comes after 20 seconds interval

Controlled audio-record

• user (admin) selectes another related user
• user presses Start button
• voice-recognition starts working and records users's speech
• user presses Stop button
• front-end sends the speech to the server
• server sends a response that contains content for Question / Answer / Transcript
• Transcript is a pure speech without any kind of transpiling
• Questions / Answers are dedicated types of content (columns) that get generated from our AlgoliaDB (on the basis of previously parsed PDF docs)

Uncontrolled audio-record

• user (admin) selectes another related user
• user presses Auto record button
• voice-recognition starts working and records users's speech
• after every 15-seconds interval front-end sends the speech to the server
• server sends a response that contains content for Question / Answer / Transcript
• Transcript is a pure speech without any kind of transpiling
• Questions / Answers are dedicated types of content (columns) that get generated from our AlgoliaDB (on the basis of previously parsed PDF docs)
• the operation can be interrupted, by pressing Stop button

Additional functionality

• user is able to manipulate every piece of text, by removing them

Parsing documents (PDF)

• user (admin) selectes another related user, this info will be related to
• user presses I am a customer checkbox to enable customer's view
• user starts uploading his own PDF (front-end sends a payload => to the back-end => back-end sends payload to Python AI)
• Python AI parses the entire PDF and generates Questions / Answers sets on the basis of this document
• Python AI sends response to the server => server stores the data in the AlgoliaDB & sends the response to the front-end
• these sets get automatically displayed on the view
• these sets come in handy when a user records an audio file

Admin Panel

User section

• a user (admin) is able to search for another users
• a user is able to remove users
• a user is able to see related Humantic API data by clicking on a particular user
• a user is able to see total amount of Algolia metrics
• a user is able to see total amount of Humantic API metrics
• a user is able to see total amount of parsed Linkedin profiles

PRE section

• a user is able to see total amount of users recored text / their owners / date
• a user is able to download each recorded text

Manual adding sales Questions / Answers

• a user is able to manually generate required Questions / Answers with related user and context

Sales Questions / Answers

• a user is able to see total amount of all the available Questions / Answers and their owners

Backend

Auth

AUTH procedures are made via JWT.

Dialog

For storing Questions/Answers we use Algolia, for search - Algolia Answers

Files

For storing files we use GridFS.

HumanticAPI

For getting information about a person through Linkedin profile - we use HumanticAPI

Metrics

We use metrics for:

• tracking usage of external API's: Algolia and HumanticAPI
• tracking usage of HumanticAPI, that a user does
• tracking recorded messages, that were received through SpeechRecognition

Parser

We use Parser for:

• parsing PDF files
• parsing questions from recognized message

SpeechRecognition

For speech recognition we use Microsof Cognitive Services Speech.

TextAnalyzer

For text analyze we use API, that makes questions from received text.

User

Users and the rest of the app's information is being stored into our DB - MongoDB.