The ParlSpeech V2 dataset provides access to parliamentary debates in nine democracies, namely Austria, Czech Republic, Germany, Denmark, Spain, Netherlands, New Zealand, Sweden and the United Kingdom. Data ranges from 1987 to 2018 and comprises more than 6 million individual speeches by 12,368 unique speakers. The dataset was created by Christian Rauh and Jan Schwalbach. It provides variables for the data of the speech, the party of the speaker and in many cases the agenda item under discussion. The dataset can be accessed here. We present the dataset below. While reading, please keep in mind the questions you see below and answer them once you reached the end. At the end, we provide a link to a platform with an interactive version of the dataset and additional tasks.

Table 1: General Tasks for the Dataset

Tasks
What kind of information do parties reveal in parliamentary speeches?
How do parties react to their role in parliament (e.g. being in government or the opposition)?
How do new and radical parties debate once they enter parliament?

Dataset Description

Overall Information

The dataset can be downloaded as separate R Data files, one for each country. Below, in Table 2, you can see an example of the dataset. The column “debate” shows the debate title (if available), “party” gives the party name, “text” contains the text of a speech as recorded by the parliament and “SPEAKER” gives the name of the speaker.

Table 2: Example from the Austria dataset

	debate	party	text	SPEAKER
1706	Punkt Erste Lesung: Antrag der Abgeordneten Mag. Alev Korun, Kolleginnen und Kollegen betreffend ein Bundesgesetz, mit dem das Bundesgesetz über die österreichische Staatsbürgerschaft, BGBl. Nr. 311/1985, zuletzt geändert durch BGBl. I Nr. 188/2013, geändert wird (242/A)	SPÖ	Herr Präsident! Hohes Haus! Der von Kollegin Korun geschilderte Fall ist ein Paradebeispiel dafür, dass Grenzwerte, Richtlinien menschliche Härtefälle verursachen können. Ich habe deshalb Sympathie, in diesem Bereich notwendige Änderungen herbeizuführen. Tatsache ist jedoch, dass zwei Tagesordnungspunkte später ein konträrer Antrag der FPÖ zur Beratung vorliegt und es deshalb notwendig ist, die Frage der Grundlagen zur Erreichung der österreichischen Staatsbürgerschaft, zu denen auch ein gesicherter Lebensunterhalt zählen muss, intensiv zu diskutieren. Ich möchte mit erwähnen – es wurde indirekt angesprochen –, dass es die SPÖ war, die in die Novelle 2013 hineinreklamiert hat, dass es für Behinderte und dauerhaft Erkrankte Ausnahmebestimmungen gibt. Wir werden uns in den nächsten Wochen nicht nur die internationalen Beispiele genau anschauen, sondern im Hinblick auf die Vereinbarung zwischen mittlerweile allen Fraktionen, dass es im Herbst ein Hearing im Innenausschuss geben soll, beide Anträge sehr detailliert beraten. Ich möchte nur dazusagen, dass die Staatsbürgerschaft ein sehr hohes demokratisches Gut ist. Zweifelsohne eignet sie sich nicht für tagespolitische Polemik, und ich würde wirklich ersuchen, dass wir hier in einer sehr sachlichen Diskussion eine einvernehmliche Lösung herbeiführen. – Herzlichen Dank. (Beifall bei der SPÖ.)	Hannes Weninger

Sentiment Analysis

One possibility to analyse legislative conflict in parliament is through sentiment analysis of the speeches. The basics of quantitative text analysis are outlined in Box 8.1 in chapter 8. As we saw in Chapter 7, legislative behaviour is an important form of political representation. Through their speeches in parliament, parties represent the views of their voters and communicate also with other members of parliament and parties. Governments and the cabinet parties typically control the legislative proceedings through agenda control. Therefore, much of what is discussed in parliament is government business, such as bills and motions. We therefore expect the government speakers to be more positive in debates about these items. Figure 1 shows the sentiment expressed by Austrian parties over time. Sentiment is calculated as the logged ratio of positive to negative words expressed by all speakers of a party. The measurement of the tone of debates, or sentiment, is based on a dictionary approach. Which words are considered positive and negative is based on the Lexicoder Sentiment Dictionary (Young and Soroka 2012, Proksch et al 2019). The parties ÖVP and SPÖ formed a coalition during the 2013-17 term and are clearly the most positive parties, as expected. Within the opposition, the FPÖ, a radical right party, is the most negative.

Figure 1: Sentiment of Austrian parties over time

However, we might think that some parties are just generally more positive than others and these parties might also be more likely to enter governments (for example because they are mainstream parties with more moderate views). A good test case of this is whether parties become more negative once they leave government and more positive when they enter. Figure 2 shows the sentiment of Swedish parties over time from 2010 to 2018. Consequently, the data covers two electoral periods with two different governments (Reinfeldt II and Lofven I). Reinfeldt II was a coalition of the parties Moderata samlingspartiet (M), Folkpartiet liberalerna (FP), Centerpartiet (C) and Kristdemokraterna (KD). These four parties are also the most positive parties throughout the government period. In contrast, the Lofven I government consisted of the Socialdemokraterna (S) and Miljöpartiet de Gröna (MP), which are the most positive throughout that particular legislative period. The radical right Sweden Democrats are the most negative party under both governments, especially at the beginning of the 2010 term when they entered parliament for the first time.

Figure 2: Sentiment of Swedish parties over time