Semantic Search Engine

Given a query, return applications from the Aptoide Mobile Application Dataset.

The search engine leverages RoBERTapp in a bi-encoder architecture:

The textual fields of the applications were encoded and index offline. At run-time, the query is encoded, and the representations are compared.

The model considers the name and description textual fields for retrieval, which can be given different weights.

Endpoint

https://apprecommender.caixamagica.pt/api/search

Parameters

1. query - A query for which applications will be retrieved.
   • For example social networks
2. n - Number of applications to return
   • Default value: 5
3. w_name - Weight given to the name textual field within the retriever.
   • Default value: 0.5
4. w_description - Weight given to the description textual field within the retriever.
   • Default value: 0.5

Returns

JSON with the following fields:

1. status

   • 'ok'
   • 'nok'

2. data

   • if status is 'ok', returns a list of ids of applications.
   • The Aptoide Mobile Application Dataset can be used to get more information for each id.

Example

import requests

endpoint = "https://apprecommender.caixamagica.pt/api/search"
params = {'query': "social networks" , 'n': 5, 
          'w_name':0.5, 'w_description': 0.5}

r = requests.get(endpoint, params = params)
print(r.json())

Output:

{'status': 'ok', 'data': [56965434, 56966709, 56970406, 56970650, 56490993]}

The Aptoide Mobile Application Dataset can be used to get information for the returned ids. For instance, using the API:

import requests

endpoint = "https://apprecommender.caixamagica.pt/api/appsdataset"

for app_id in [56965434, 56966709, 56970406, 56970650, 56490993]:
    params = {'app_id': app_id}

    r = requests.get(endpoint, params = params)
    print(r.json()['data']['name'])

Output

Facebook
WhatsApp Messenger
Instagram 
Messenger - Text and Video Chat for Free
Twitter