[Air-L] Tool request: sentiment analysis for Cyrillic

[Sergei Pashakhin]

Some insights for the Russian language

It seems that the state of the art for Russian is LSTM with word2vec (http://www.dialog-21.ru/media/3380/arkhipenkoetal.pdf). However, from our experience, a dictionary approach with SentiStrength (http://sentistrength.wlv.ac.uk) yield comparable results. Our lab develops a dictionary for texts on social and political topics, and it's available here http://www.linis-crowd.org (in Russian). You can learn more about the dictionary from this paper http://www.dialog-21.ru/media/3400/koltsovaoyuetal.pdf.


Best regards,

Sergei Pashakhin
Laboratory for Internet Studies,
National Research University Higher School of Economics
https://linis.hse.ru/en/ <https://linis.hse.ru/en/>

> On 29 Oct 2018, at 09:22, Xanat Meza via Air-L <air-l at listserv.aoir.org> wrote:
> 
> There is a benchmark that automatically translates text to English and then does sentiment analysis: https://www.researchgate.net/publication/261959618_iFeel_a_system_that_compares_and_combines_sentiment_analysis_methods
> 
> Xanat V. Meza
> 
> Ph.D. candidate - Kansei, Behavioral and Brain SciencesUniversity of Tsukuba
> M.A. Media and Communication
> Yeungnam University
> B.D. Graphic Communication Design
> Universidad Autonoma Metropolitana
> 
> 
>    El sábado, 27 de octubre de 2018 8:11:58 p. m. GMT+9, Stuart Shulman <stuart.shulman at gmail.com> escribió:  
> 
> DiscoverText.com works with everything we have tested it on, including
> Hebrew, Arabic, Mandarin, and others. I would like to hear from you if we
> can add Russian to the list. I will send you a sponsored license to test it
> out. Please let us know if it works. We are happy to sponsor anyone working
> on research to protect democratic societies from authoritarian assaults on
> the ballot box.
> 
> From our product description:
> 
> "Most text analytics software packages work well with English text and a
> handful of other languages; however, many of these tools fail when
> analyzing non-Latin, multilingual texts, such as Arabic, which appears
> correctly only in a right-to-left format. Further, many software solutions
> have additional problems tokenizing text when it is an ideograph-based
> language (e.g. Chinese or Korean). Texifter’s software, DiscoverText, is
> unique in that it is capable of effective operations on multilingual texts
> and the coding platform builds effective custom machine classifiers on the
> fly and at scale for these corpora."
> 
> Stu Shulman <https://twitter.com/StuartWShulman>NEFC-West
> <https://www.nefc.us/west>
> 2008 Boys Head Coach
> 
> 
> 
> 
> 
> On Fri, Oct 26, 2018 at 4:37 PM John P. Bell <John.P.Bell at dartmouth.edu>
> wrote:
> 
>> Hi all,
>> 
>> I’m looking for tools to do sentiment analysis and general mining on
>> Russian language tweets. I see there are some options out there, but if
>> anyone has experience trying to do this I’d appreciate it if you could
>> share some insight on the software you used. While I’d be more interested
>> in something I can set up and run myself than subscribing to a service, I’m
>> not absolutely committed to that idea.
>> 
>> Any suggestions?
>> 
>> Thanks,
>> 
>> - John
>> 
>> —
>> John P. Bell, PhD
>> Lead Application Developer (Digital Humanities), Dartmouth Research
>> Computing
>> Asst. Prof. of Digital Curation, University of Maine
>> http://johnpbell.info
>> 
>> _______________________________________________
>> The Air-L at listserv.aoir.org mailing list
>> is provided by the Association of Internet Researchers http://aoir.org
>> Subscribe, change options or unsubscribe at:
>> http://listserv.aoir.org/listinfo.cgi/air-l-aoir.org
>> 
>> Join the Association of Internet Researchers:
>> http://www.aoir.org/
> _______________________________________________
> The Air-L at listserv.aoir.org mailing list
> is provided by the Association of Internet Researchers http://aoir.org
> Subscribe, change options or unsubscribe at: http://listserv.aoir.org/listinfo.cgi/air-l-aoir.org
> 
> Join the Association of Internet Researchers:
> http://www.aoir.org/  
> _______________________________________________
> The Air-L at listserv.aoir.org mailing list
> is provided by the Association of Internet Researchers http://aoir.org
> Subscribe, change options or unsubscribe at: http://listserv.aoir.org/listinfo.cgi/air-l-aoir.org
> 
> Join the Association of Internet Researchers:
> http://www.aoir.org/