LATAM Country Classifier

Index:

1. Introduction
2. Country Detector Model
3. Tags in Research
4. Limitations

Introduction

Learn how the classifier that enhances the possibility of data collection in Spanish works.

STILINGUE has developed the LATAM country classifier, a feature that enables the collection of data from Spanish publications (that have not yet been classified) using Artificial Intelligence to deliver increasingly accurate results. 

With this feature, it is possible to extract reports, understand trends, and gain insights from each Latin American country individually by segmenting publications from the Listening. Through this action, it becomes possible to better understand the opinions and behaviors of LATAM consumers. 

Country Detector Model

STILINGUE's Artificial Intelligence processes the text of desired publications and provides the location to which that publication belongs. This location, delivered by the model, is presented to the client in the form of tags in the publications. When the model lacks sufficient confidence, the tag "Unidentified Country" is added.

The location of the publications, delivered by the AI model, will appear in the search through Tags in the publications. Therefore, it is necessary for the client to create the Tags in the Search Configuration so that they can be used in filters and in any other areas of the War-Room.

Once enabled, the model will automatically apply location Tags to the desired publications in the search. 

This way, the client will be able to use all the filtering and data aggregation functionality of the War-Room to extract reports based on the location of the publications.

Tags in Research

Currently, the model only works with publications from Latin American countries (LATAM), specifically with publications in Spanish. Publications in Portuguese, English, and other languages will not receive Tags. 

The Tags used are:

• Argentina;
• Chile;
• Colombia;
• Mexico;
• Non-LATAM;
• Other LATAM;
• Unidentified country.

Observation: "Non-LATAM" refers exclusively to Spain, "Other LATAM" refers to other countries in Latin America, and "Unidentified Country" indicates that the model did not have sufficient confidence to make an assertive decision. New countries can be developed on demand.

Publications receiving the geolocation tag are those that the language detector classified as being in the Spanish language.

Limitations

• Tag does not correspond with the publication: 

If a publication is found with the wrong geolocation tag, the client can remove this tag and add one of the tags mentioned above.

This manual correction helps the responsible team make the Artificial Intelligence even more accurate, as these corrected data will be subsequently used to improve the accuracy of classification. 

• Add or remove geolocation tag:

If the client only removes a geolocation tag, this tag will be automatically added again by the model in a few minutes.

Now, if the customer removes a geolocation tag and replaces it with another geolocation tag, the new tag will be retained.

If a geolocation tag is added to a post where another one already exists, the post will have two geolocation tags. In this case, it is recommended to remove the unwanted tag to avoid inconsistencies in the reports. 

To access this functionality, please get in touch with your account manager. The subscription to this package is an additional feature offered separately.

For more information, visit the discussion on the subject at our community or videos on our channel. 😃