Latent Semantic Scaling: A Semisupervised Text Analysis Technique for New Domains and Languages

Abstract

Many social scientists recognize that quantitative text analysis is a useful research methodology, but its application is still concentrated in documents written in European languages, especially English, and few sub-fields of political science, such as comparative politics and legislative studies. This seems to be due to the absence of flexible and cost-efficient methods that can be used to analyze documents in different domains and languages. Aiming to solve this problem, this paper proposes a semisupervised document scaling technique, called Latent Semantic Scaling (LSS), which can locate documents on various pre-defined dimensions. LSS achieves this by combining user-provided seed words and latent semantic analysis (word embedding). The article demonstrates its flexibility and efficiency in largescale sentiment analysis of New York Times articles on the economy and Asahi Shimbun articles on politics. These examples show that LSS can produce results comparable to that of the Lexicoder Sentiment Dictionary (LSD) in both English and Japanese with only small sets of sentiment seed words. A new heuristic method that assists LSS users to choose a near-optimal number of singular values to obtain word vectors that best capture differences between documents on target dimensions is also presented. Many social scientists recognize that quantitative text analysis is a useful research methodology, but its application is still concentrated in documents written in European languages, especially English, and few sub-fields of political science, such as comparative politics and legislative studies. This is not because they are only interested in domestic politics in North America and Europe but because the existing quantitative text analysis toolkit is not suitable for analysis of documents in non-European languages or in other fields. The domination of European languages in quantitative text analysis is partially due to its history: many of the text analysis dictionaries, including the Lexicoder Sentiment Dictionary (LSD) (Young & Soroka, 2012) and the Linguistic Inquiry and Word Count (LIWC), were created based on the General Inquirer dictionaries (Stone et al., 1966), which were developed to analyze English texts during the Cold War; statistical analysis of textual data was also introduced to political science for CONTACT Kohei Watanabe

Judicial diplomacy of the German Federal Constitutional Court: bilateral court meetings as a novel data source to assess transnational communication of constitutional courts

“…These steps ensure that the remaining word-tokens represent the meetings’ most significant and meaningful words. Finally, this corpus is used to create a feature co-occurrence matrix (FCM) that estimates the semantic proximity of each token, which is why the relatively short meeting reports are a suitable data source (Watanabe 2020 ). In general, an FCM represents a network based on co-occurrences of specific features (in this study tokens) in a defined context (here: all meeting reports).…”

Supranational emergency politics? What executives’ public crisis communication may tell us

“…To construct such a measure, I resort to what the recent text-as-data literature has denoted as latent semantic scaling (Rheault et al, 2016;Wantanbe, 2020). It starts from just a few selected words that are intuitively associated with the polarity of interest.…”

The Effect of Exposure to (Non-)Like-Minded Information on the Use of Political Incivility on Twitter

“…The level of the incivility of a word was obtained by summing the cosine similarity of the vector of the word and the seven uncivil seed words, summing the cosine similarity of the vector of the word and the seven civil seed words multiplied by -1, and dividing the sum by 14. The level of incivility of a tweet was obtained by calculating the level for all the words in the tweet using the aforementioned method and dividing the sum by the number of words (for details, see Watanabe (2020)).…”