This is a web-app which employs the citizen science program by tracking their tweets and subsequent data analysis on Google Earth Engine, aimed at effective CyanoHAB detection and monitoring in water bodies around the world.
Currently the tweets are being filtered using a specific set of keywords
However, not all filtered tweets have valid locations - a few of them sometimes have invalid locations such as "Earth" etc. These cannot be geocoded correctly and can result into a false potential CyanoHAB location. To solve this issue update this file.
Identify tweets with such invalid locations and flag them
Attempt retrieving correct locations from the flagged tweets and report the results
NOTE: This issue is congruent with issue #8. All changes here should be reflected in this file as well.
Currently on the login page the focus is always set on the username field
Because of this, when the user clicks on the password field, it again goes back to the username field
Input can be entered in the password field only by using 'Shift' button in the keyboard to navigate to the password field as this does not toggle the focus flag.
The above repo already scrapes the tweets in real time (live) and creates a stream of the filtered tweets
Figure out a way to store this stream - the collection should go on, a file should be saved on disk, appended to every 5 mins or so with the new incoming tweets
Currently the tweets are being filtered using a set of keywords
Update that set of keywords in the tweet_tracker.py file with the one below: BlueGreenAlgae, CyanoBacteria, cyanotracker, anabaena, microcystis, cyanotoxins, toxic algae, algae bloom, algal bloom, '#CyanoBacteria', '#AlgaeBloom', '#CyanoBacteriaBlooms', '#CyanoHABs', '#HABs'
Employ an algorithm similar to page ranking, to manage simultaneous webapp calls to the GCP vm and the consequent API calls for the Google map to avoid timeouts.
Currently links are being provided to relevant tweets (see snapshot)
Find a way to embed these tweets within the webpage (all tweets together in a single stream) instead of the user having to click on the link and open a new webpage
Currently the words are being filtered using keyword matching Code here
This approach simply looks for the exact keywords in the tweets and filters them
It is always possible that some tweets may contain some keywords which might have a similar meaning to algae bloom (such as eutrophication) but are not the exact same
Such tweets while important are not filtered and instead missed out
Use NLP and word embedding concepts to process the tweet content and find similarity with the keywords so that the tweets having words which have similar meaning to keywords but are not exact keywords are not missed out