Analysing sports news - an integrated solution

Posted on 2 min.

Tranzpress, the international language & media intelligence company needed a solution to analyze news about sport events in Hungarian and English. They decided to choose Neticle Text Analysis API as the perfect solution to their needs. Why? Check out the details from the blog post.


PROBLEM

Tranzpress, the international language & media intelligence company - providing full-scale localisation services from translation to dtp, from software localisation to testing in over 40 language pairs - needed a solution to analyse sports news from 2016-2017 in Hungarian and English.

It was important to perform human-level precise sentiment analysis on their news database. They needed to make a difference between very positive, positive, neutral, negative and very negative mentions.

As there were some key sports events it was important to analyse only the opinion that is directly targeting these events in the news on both languages and to not analyse the related sports results just the event opinion and experience. Reviews and more personal summaries of events were collected mainly.

Here is an English example. And a Hungarian one.

SOLUTION

They decided to choose Neticle Text Analysis API and they integrated both document and entity-level sentiment analysis for the news.

Neticle's methodology for quantifying opinion phrases fitted well to their purpose. The 7-point scale from -3 to +3 covers 3 stages of tone intensity in each polarity direction: very negative, negative, slightly negative, neutral, slightly positive, positive, very positive.

They converted Neticle's sentiment score to the sentiment categories needed:

  • positive: 1, 2
  • very positive: 3
  • neutral: 0
  • negative: -1, -2
  • very negative: -3

On some of the output charts you can see the ratio of positive, negative and neutral mentions, reach numbers and key authors of sports articles:

Besides checking the most active authors, they defined keywords to cover the sports events and also analysed the key locations (cities) in the news to understand the perceptions of the organizer cities.

We needed 1 week to integrate Neticle's API and 8 hours for net development tasks and fine-tuning.

OUTCOME

Tranzpress has have been using the integrated solution for 3 years and no extra work needed to be done. News are analyzed real-time with human-level precision and showed on their dashboard.

As for future steps in our cooperation, we plan to involve further news feeds in their system.


Share: