Our article on Sentiment analysis

Recently me and my colleague made a prototype tool for Sentiment analysis of web reviews in Russian language.

It was quite successful, since it took the second place on the ROMIP Sentiment Analysis Track in 2012.

Here is a link to the article and presentation we wrote for the conference Dialogue-2013 about the prototype.

Common approach to the Named-entity recognition task

In the further explanation I shall provide examples from the task of the extraction of geography mentions from texts (e.g. mentions of cities, countries, districts, streets, etc.).

We may distinguish three steps in the NER:

1. select objects (hypotheses) in the text; object is a word or group of words that probably is a mention of the entity
2. represent each object as a set of features that could be used for classification
3. classification itself