Before using the service, please read the preliminary information containing a description of steps that enable access to the CLARIN-PL developer interface.
The service allows for the determination of the emotional tone of texts in various languages according to four categories: negative, positive, meta_zero, and ambiguous. Each category is assigned a numerical value corresponding to the level of emotional tone. The highest value determines the emotional tone of a given text.
MultiEmo enables the analysis of text in 109 languages, including 25 particularly supported ones. The list of particularly supported languages can be found here.
The service uses the MultiEmo model for multilingual sentiment analysis, which covers 11 languages and includes consumer reviews in four domains: medicine, hotels, products, and universities. The original Polish reviews consisted of 8 216 documents with 57 466 sentences. The reviews were manually annotated for sentiment at the document level and sentence level (3 annotators per element). We achieved a high Positive Specific Agreement value of 0.91 for texts and 0.88 for sentences. The dataset was then automatically translated into English, Chinese, Italian, Japanese, Russian, German, Spanish, French, Dutch, and Portuguese. It is publicly available under the Creative Commons Attribution 4.0 International license and can be downloaded here.
The service can be run:
type: selection of the level of text processing
text - the entire document is processed, default optionparagraphsentencelanguage:
auto - default option, enables automatic language detection,ca, cs, de, el, en, es, fi, fr, ga, hu, is, it, lt, lv, nl, pl, pt, ro, ru, sk, sl, sv, ta, yue, zh.The service can be run in the Windows system with default values using the following LPMN query: ['any2txt','multiemo']
[['any2txt','multiemo']] - input data in the form of a compressed directory (.zip)['any2txt',{'multiemo':{'lang':'en'}}] - input data in English['any2txt',{'multiemo':{'type':'paragraph'}}] - text segmentation by paragraphs['any2txt',{'multiemo':{'lang':'en','type':'sentence'}}] - input data in English, ext segmentation by sentencesA text file.
A JSON file containing the following information:
labels - categories identifying the emotional tone of the text:
meta_minus_m - negativemeta_plus_m - positivemeta_zero - neutralmeta_amb - ambiguousdecision - the category with the highest value assigned to itlang - the language of the processed textsource - the content of the segment indicated for analysis in the type optionIn Colab: Multiemo - Determining the emotional tone of texts in various languages
Piotr Miłkowski, Marcin Gruza, Przemysław Kazienko, Joanna Szołomicka, Stanisław Woźniak, Jan Kocoń (2022) "MultiEmo: Language-Agnostic Sentiment Analysis", Computational Science – ICCS 2022. ICCS 2022. Lecture Notes in Computer Science, vol 13351, Springer, 72–79.
(C) CLARIN-PL