Sanremo's Winner Is... Category-driven Selection Strategies for Active Learning
نویسندگان
چکیده
English. This paper compares Active Learning selection strategies for sentiment analysis of Twitter data. We focus mainly on category-driven strategies, which select training instances taking into consideration the confidence of the system as well as the category of the tweet (e.g. positive or negative). We show that this combination is particularly effective when the performance of the system is unbalanced over the different categories. This work was conducted in the framework of automatically ranking the songs of “Festival di Sanremo 2017” based on sentiment analysis of the tweets posted during the contest. Italiano. Questo lavoro confronta strategie di selezione di Active Learning per l’analisi del sentiment dei tweet focalizzandosi su strategie guidate dalla categoria. Selezioniamo istanze di addestramento combinando la categoria del tweet (per esempio positivo o negativo) con il grado di confidenza del sistema. Questa combinazione è particolarmente efficace quando la distribuzione delle categorie non è bilanciata. Questo lavoro aveva come scopo il ranking delle canzoni del “Festival di Sanremo 2017” sulla base dell’analisi del sentiment dei tweet postati durante la manifestazione.
منابع مشابه
Evaluating quantitative stock selection strategies in Tehran Stock Exchange
There are different strategies for selecting stocks, and different investors use different strategies according to their risk tolerance or their expected rate of return. In this study, the profitability of a broad range of stock se-lection strategies in Tehran Stock Exchange over the period 1370-1383, has been examined, and it has been investigated whether the successful strategies in other cou...
متن کاملLanguage Learning Strategy Use and Instruction for the Iranian Junior High School EFL Learners: A Mixed Methods Approach
In order to confirm the effectiveness of language learning strategies in theIranian context in junior high schools, this study was designed to examine thepatterns of strategy use, the effects of strategy instruction on the students’ strategyuse, and the relationship between the participants’ strategy use and their Englishachievement. To achieve this objective, 57 junior high school participants...
متن کاملLearning Active Learning from Data
In this paper, we suggest a novel data-driven approach to active learning (AL). The key idea is to train a regressor that predicts the expected error reduction for a candidate sample in a particular learning state. By formulating the query selection procedure as a regression problem we are not restricted to working with existing AL heuristics; instead, we learn strategies based on experience fr...
متن کاملActive learning and logarithmic opinion pools for HPSG parse selection
For complex tasks such as parse selection, the creation of labelled training sets can be extremely costly. Resource-efficient schemes for creating informative labelled material must therefore be considered. We investigate the relationship between two broad strategies for reducing the amount of manual labelling necessary to train accurate parse selection models: ensemble models and active learni...
متن کاملLearning Active Learning from Real and Synthetic Data
In this paper, we suggest a novel data-driven approach to active learning (AL). The key idea is to train a regressor that predicts the expected error reduction for a candidate sample in a particular learning state. By formulating the query selection procedure as a regression problem we are not restricted to working with existing AL heuristics; instead, we learn strategies based on experience fr...
متن کاملذخیره در منابع من
با ذخیره ی این منبع در منابع من، دسترسی به آن را برای استفاده های بعدی آسان تر کنید
عنوان ژورنال:
دوره شماره
صفحات -
تاریخ انتشار 2017