Available at

Parliament Interpreting Corpus), an electronic parallel corpus of source and target 
speeches in Italian, English and Spanish (Monti et al., 2005; Bendazzoli & 
Sandrelli, 2005). The plenary sittings of the European Parliament were chosen as 
source material in this corpus because they show a high level of homogeneity, in 
that all speeches are produced in the same formal setting (Marzocchi & Zucchetto, 
1997). EPIC allows both parallel and comparable analyses and contains about 
180,000 words. It is available online for the whole interpreting community to help 
to share our knowledge of interpreting and even enhance its teaching.
As for 
DIRSI, i.e. Directionality in Simultaneous Interpreting (Bendazzoli & Sandrelli, 
2009), it includes interpreters’ output into both their native language and their 
foreign working language. This corpus was compiled thanks to audio recordings 
from international conferences about health-related subjects held in Italy between 
2005 and 2008 and includes recordings from different sessions (i.e. opening 
statements, presentations, and closing sessions). The language pair is English 

---

Corpus-based interpreting studies 
 
page 16 
and Italian and five professional interpreters have accepted to contribute data. 
Debates were excluded from the corpus due to their degree of interactivity. The 
creation of this corpus has been possible thanks to the experience previously 
gained with the EPIC project. 
It is important to add that interpreting corpora created ad hoc by individual 
researchers for manual analysis are still used today, and complement the realm of 
CIS. The notion of corpus in interpreting studies has clearly initially been linked to 
empirical research based on authentic data (i.e. from real-life interpreting 
assignments) and because of the difficulty to compile electronic corpora that 
notion still applies to data sets that continue to be analysed manually and not 
electronically.
In the following paragraphs, I will describe the main characteristics of interpreting 
corpora (i.e. interpreting mode and setting, corpus size, languages and data 
accessibility) that have already been compiled recently, and that are currently 
being compiled. 
The first CIS projects focused on professional simultaneous interpreting 
performed in conference settings. Two specific sources of data have been 
predominant, namely TV broadcasting and the European Parliament (EP). At the 
EP, for instance, source speeches are interpreted simultaneously into as many as 
23 languages (sometimes through relais interpreting) and these speeches can be 
used for research purposes (Bendazzoli, 2010). Other fields have yet been 
explored such as festivals, medical conferences, and football press conferences. 
Simultaneous interpreting is nevertheless not the only interpreting mode that can 
be analysed. Asian research centres focus more on consecutive interpreting 
because of data accessibility, their data source being televised press conferences 
of Chinese political representatives (Wang, 2015). More recent projects also focus 
on short consecutive interpreting in community settings or on dialogue interpreting. 
It is also important to note that efforts are being made to develop sign-language 

---

Corpus-based interpreting studies 
 
page 17 
corpora despite the difficulty to collect data due to anonymity issues (see Metzger 
& Roy, 2011). 
If they are compared to the spoken part of the British National Corpus (10 million 
words), interpreting corpora are quite small (Dembry & Love, 2015). Projects 
based on EP data still have to reach the size of general reference corpora, and it 
might just be a matter of time and of labour force. However, it is hoped that CIS 
projects will develop in other international organizations than the EP to diversify 
interpretation settings. It is already considered in some organisations such as the 
European Commission (Spinollo, 2018; Scardulla, 2016). Even though current 
projects carried out in Asia are expected to generate pretty large resources, 
compiling very large corpora in the near future is not likely to happen.
In terms of languages used, we can see a wide range of language combinations, 
which confirms one of the “special challenges” (Setton 2011: 68) of CIS - 
multilingualism. English is represented in many studies, but it is really encouraging 
to see that non-European languages such as Hebrew, Japanese or Chinese are 
represented as well.
The last feature of interpreting corpora I would like to focus on is data accessibility, 
which has always been an issue in CIS. In the oldest studies mentioned in 
Setton’s (2011) overview (see Appendix 1), transcripts were rarely made available 
and sound files were recorded on tape and not in digital form. Transcript files were 
thus hard to access (Diriker, 2004) and the analysis had to be carried out manually. 
In the same period, some studies were however based on machine-readable 
corpora (Cencini, 2002; Fumagalli, 2000) but the transcripts were “not available for 
outside use” (Setton 2011: 40). In some other cases, transcripts are available on 
CDs (e.g. Vuorikoski, 2004; Monacelli, 2009) or on the web as it is the case for the 
two corpora I mentioned earlier, EPIC and DIRSI. In the future, data accessibility 
should be facilitated at least among the research community. 

---

Corpus-based interpreting studies 
 
page 18